|
 
 |
| REVIEW ARTICLE |
|
| Year : 2023 | Volume
: 6
| Issue : 2 | Page : 53-68 |
|
Artificial intelligence in respiratory medicine: The journey so far – A review
K Kalaiyarasan, R Sridhar
Sri Venkateshwaraa Medical College Hospital and Research Centre, Puducherry, India
| Date of Submission | 20-Mar-2023 |
| Date of Acceptance | 02-May-2023 |
| Date of Web Publication | 13-Jul-2023 |
Correspondence Address:
Dr. K Kalaiyarasan
Sri Venkateshwaraa Medical College Hospital and Research Centre, Puducherry
India
 Source of Support: None, Conflict of Interest: None
DOI: 10.4103/japt.japt_13_23
The integration of artificial intelligence (AI) and the medical field has opened a wide range of possibilities. Currently, the role of AI in the medical field is limited to image analysis (radiological and histopathology images), identifying and alerting about specific health conditions, and supporting clinical decisions. The future of lung cancer screening, diagnosis, and management is expected to undergo significant transformation with the use of radiomics, radiogenomics, and virtual biopsy. AI can also help physicians diagnose and treat a variety of respiratory illnesses, including interstitial lung diseases, asthma, chronic obstructive pulmonary disease, and pleural diseases such as effusion and pneumothorax, pneumonia, pulmonary artery hypertension, and tuberculosis. AI can also help in the automated analysis and reporting of lung function tests, polysomnography, and recorded breath sounds. Through robotic technology, AI is set to create new milestones in the realm of interventional pulmonology. A well-trained AI may also offer new insights into the genetic and molecular mechanisms of the pathogenesis of various respiratory diseases and may also assist in outlining the best course of action with the horizontal integration of patients' digital health records, digital radiographic images, digital pathology images, and biochemical lab reports. As with any technology, doctors and researchers should be aware of the advantages and limitations of AI, and they should use it responsibly to advance knowledge and provide better care to patients.
Keywords: Artificial intelligence, lung cancer, radiomics, respiratory medicine, robotic-assisted bronchoscopy, virtual biopsy, whole slide imaging
How to cite this article:
Kalaiyarasan K, Sridhar R. Artificial intelligence in respiratory medicine: The journey so far – A review. J Assoc Pulmonologist Tamilnadu 2023;6:53-68 |
How to cite this URL:
Kalaiyarasan K, Sridhar R. Artificial intelligence in respiratory medicine: The journey so far – A review. J Assoc Pulmonologist Tamilnadu [serial online] 2023 [cited 2023 Sep 30];6:53-68. Available from: https://www.japt.in//text.asp?2023/6/2/53/381412 |
| Introduction | |  |
The term “artificial intelligence” (AI) was coined by Professor John McCarthy in 1955. He defined AI as “the science and engineering of making intelligent machines.”[1] AI is the simulation of human intelligence in machines that are programmed to think and learn like humans. It involves the development of algorithms and computer programs that can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation.
The field of AI is constantly evolving and advancing, with new techniques and technologies being developed to enable machines to perform more complex tasks and make more accurate decisions. AI has wide-ranging applications in many industries, including healthcare, finance, transportation, and manufacturing.
| Types of Artificial Intelligence | | |
Based on its functionality, AI can be classified as:
- Reactive machines: AI systems that only respond to the current situation and do not have the ability to remember past events. They cannot draw inferences from the data to evaluate their future actions and can only perform a narrow range of pre-defined tasks. For example, AI chess game
- Limited memory: AI systems that can remember past events and use them to inform current decision-making. For example, self-driving cars
- Theory of mind: AI systems that can understand and infer human emotions and mental states. This type of AI has not yet been fully developed, but rigorous research is happening in this area
- Self-aware: AI systems that possess a sense of self and consciousness. We have yet to achieve this form of AI.
| Branches of Artificial Intelligence | | |
AI can be used to solve real-world problems by implementing the following processes/techniques:
- Machine learning (ML)
- Deep learning (DL)
- Natural language processing (NLP)
- Robotics
- Expert systems
- Fuzzy logic
| Machine Learning | | |
ML is a subset of AI that involves training computer algorithms to learn from data and make predictions or decisions without being explicitly programmed. ML algorithms can be supervised, unsupervised, semi-supervised, or reinforced.
| Types of Machine Learning | | |
- Supervised learning: This is the most common form of ML. It involves training an AI model on a labeled dataset, where the model is provided with inputs and the corresponding correct outputs. The model then makes predictions on new, unseen inputs by generalizing from the examples it was trained on
- Unsupervised learning: In this type of learning, the AI model is not provided with labeled data. Instead, it must discover patterns and structure in the input data on its own. This is often used for tasks such as clustering and dimensionality reduction
- Reinforcement learning: In this approach, an AI agent is trained through trial-and-error interactions with an environment. The agent receives rewards or penalties based on its actions, and it learns to optimize its behavior to maximize the rewards over time
- Semi-supervised learning: This is a combination of supervised and unsupervised learning. The model is trained on a dataset that includes both labeled and unlabeled examples
- Transfer learning: In this approach, a pretrained model is fine-tuned for a new task using a smaller dataset. This is useful when there is a shortage of labeled data for a specific task
- Multi-task learning: In this approach, a model is trained to perform multiple tasks simultaneously. It is a way to leverage the knowledge learned from one task to improve performance on other tasks.
| Deep Learning | | |
DL is a specific type of ML that is based on artificial neural networks (ANNs). ANNs are inspired by the structure and function of the human brain and are composed of layers of interconnected “neurons” that process information. The term “deep” refers to the fact that these networks have multiple layers, allowing them to learn and represent increasingly complex features of the input data.
DL is particularly useful for tasks involving image, audio, or text data, as it can automatically learn features from raw data and improve the performance of the model. DL models are highly accurate and widely used in a variety of applications, including image and speech recognition, NLP, and computer vision.
| Types of Deep Learning | | |
There are several different types of neural networks, each with their own strengths and weaknesses. Some of the most commonly used types of DL networks include:
- Convolutional neural networks (ConvNets or CNNs): These are neural networks designed for image and video recognition. ConvNets consist of multiple layers, including convolutional layers, pooling layers, and fully connected layers
- Recurrent neural networks (RNNs): These are neural networks designed for sequential data, such as text, speech, and time series data. RNNs have loops that allow the network to process information from previous time steps, making them well-suited for tasks such as language translation and speech recognition
- Autoencoders: These are neural networks designed for unsupervised learning. Autoencoders consist of two parts: an encoder that maps the input to a lower-dimensional representation, and a decoder that maps the lower-dimensional representation back to the original input
- Generative adversarial networks (GANs): These are neural networks designed for generative tasks, such as generating new images or music. GANs consist of two networks: a generator network that creates new data, and a discriminator network that determines whether the data are real or generated
- Transformers: These are neural networks designed for NLP tasks, such as machine translation and sentiment analysis. Transformers consist of multiple attention layers, allowing the network to process information from different parts of the input data in parallel.
The choice of network architecture will depend on the task at hand, the available data, and the computational resources available.
In summary, ML is a broader concept that encompasses different types of algorithms and techniques to learn from data and make predictions or decisions, while DL is a specific type of ML that uses neural networks to learn patterns and features in the data.
| Natural Language Processing | | |
NLP is the study of using natural human language to extract information that may be used to communicate with machines. NLP involves the use of computational techniques to process and analyze human language data, such as text, speech, and writing. The goal of NLP is to enable computers to understand and respond to human language in a way that is similar to how a human would.
| Robotics | | |
Robotics is a branch of AI that involves the design, construction, operation, and use of robots. Robotics combines AI with hardware and engineering to create machines (robots) that can perform tasks that would normally require human intelligence.
| Fuzzy Logic | | |
Fuzzy logic is a mathematical framework for dealing with uncertainty and imprecision in decision-making. It was introduced as an alternative to classical (binary) logic, which only allows for decisions to be made on a strict yes or no (true or false) basis. Fuzzy logic, on the other hand, allows for decisions to be made based on degrees of truth, represented by a value between 0 and 1. It is particularly useful in situations where there is uncertainty, ambiguity, or subjectivity in the data and decisions being made.
One of the key benefits of fuzzy logic is that it can handle complex, real-world problems that are difficult to solve using classical logic. Using a continuous representation of concepts and degrees of truth, fuzzy logic is able to make decisions based on partial or uncertain information. This makes it a powerful tool for solving problems in areas such as control, decision-making, and pattern recognition.
Fuzzy logic is used in the medical field to solve complex problems that involve decision-making.
| Artificial Intelligence in Medical Field | | |
AI technology is currently being applied in various fields of medicine, such as radiology, cardiology, respiratory medicine, gastroenterology, nephrology, endocrinology, neurology, and histopathology.[2] The role of AI in the medical field at present is limited to image analysis (radiological and histopathology images), identifying and alerting about specific health conditions such as atrial fibrillation, epileptic seizures, and hypoglycemia, and supporting clinical decisions.[2],[3] AI technology can analyze the electronic health records of patients, and using DL can predict multiple medical events, unplanned admissions, complications, and mortality.[4] AI technology, through ML, can analyze the vast amount of data generated in various genomic, proteomic, and clinical trial studies, thereby playing a role in drug discovery and predicting the efficacy and toxicity of drugs.[5] AI is being increasingly used in cardiology, where AI coupled with smart devices is used to monitor patients' pulse and immediately alert emergency services in case of any abnormalities in the pulse, such as atrial fibrillation. This list is incomplete and will be expanded to other fields in medicine as AI is still in its nascent stage.
With the horizontal integration of patients' digital health records, digital radiographical images, digital pathology images, and biochemical lab reports, a well-trained AI can assist the doctor in arriving at the diagnosis with improved efficiency and accuracy, cutting down the overall cost and waiting time for the patients. A single AI algorithm can scan and analyze thousands of radiological and pathological images for an abnormality in a few minutes, which will not be possible for a single human being. Thus, incorporating AI into the workflow will greatly reduce the work of the doctors by decreasing the time required, and improving the accuracy of diagnosis.
The goal of this review article is to explore the advancements made using AI in the field of respiratory medicine to improve patient care. The broad and specific application of AI to various aspects of respiratory medicine is summarized in [Table 1]. | Table 1: Summary of the applications of artificial intelligence in respiratory medicine
Click here to view |
| Applications of Artificial Intelligence in Respiratory Medicine | | |
| Lung Cancer-Radiology | | |
The application of AI in the field of lung cancer is largely thanks to techniques such as radiomics, radiogenomics, and virtual biopsy. The definitions are given in [Table 2] and the simplified process of radiomics is presented in [Figure 1].
Several studies have shown that early detection of lung cancer using low-dose computed tomography (LDCT) reduces mortality by up to 20%.[6],[7],[8] The US Preventive Services Task Force concludes with moderate certainty that annual screening for lung cancer with LDCT has a moderate net benefit.[9] Although LDCT helps in the early detection of lung cancer in at-risk populations, it also causes false-positive results, leading to unnecessary tests and invasive procedures.[10] Further, due to the limits of human vision and inter-reader variability among radiologists, up to 35% of lung nodules, especially the tiny ones, are missed at the initial screening.[11]
In the 1960s, the first attempt was made to use a computer to identify lung cancer.[12] The computer-aided diagnosis (CAD) system was developed for detecting pulmonary nodules in chest X-rays.[13] Newer CAD systems are integrated with AI algorithms that automatically detect lung nodules on chest X-rays. Overall, they are not superior to radiologists in detecting nodules and suffer from poor sensitivity.[13],[14],[15] One study revealed that the DL-CAD system had a higher detection rate but also suffered from high false-positive results.[16]
AI technology is being increasingly used in LDCT screening for screening lung cancers. A team of bioengineers has made use of AI technology to convert low-dose CT images into images of superior quality by postprocessing the images. Thus, AI can produce high-quality images from low-dose CT scans, which typically have poor resolution and noise artifact, eliminating the need for high-dose radiation CT scans.[17],[18]
An AI trained using an adequate quantity of radiology images can localize and identify the lesions in digital images of chest X-rays and CT scans with reasonable accuracy, sometimes better than radiologists. Coupled with histopathology images, an AI can accurately diagnose lung cancer, assist in treatment decisions, and predict the prognosis.[19] AI technology not only identifies the lesion with improved accuracy compared to reading the CT images by radiologists alone but is also able to identify the lesion as benign or malignant with a decent accuracy of more than 80%.[20],[21] Studies have shown that when AI technology is coupled with radiologist interpretation, it leads to a significant improvement in sensitivity and a reduction in false positive interpretation.[14],[20] Various studies have successfully applied radiomics to predict the malignant transformation of solid, part-solid, or subsolid nodules.[22],[23],[24],[25]
| Lung Cancer-Histopathology | | |
The researchers have successfully used AI-ML and DL to not only identify various types of cancer but also predict cancer recurrence from digital histopathology images.[26],[27],[28],[29],[30],[31],[32] AI can even distinguish between malignant and premalignant lesions.
High-quality digital images of pathology slides, commonly stained with H and E stains, are analyzed by an AI algorithm. The region of interest (ROI) is then detected by the AI, which has been trained using DL on a large number of normal and pathological slide images.[33] A sufficiently trained AI can identify whether the abnormalities in the ROI are benign or malignant, and if malignant, it can also determine the subtype of lung cancer.
With adenocarcinoma accounting for about 40% and squamous cell carcinoma contributing about 25%–30% of newly diagnosed lung cancer among patients, AI has been extensively trained so far to detect these common subtypes of lung cancer. Such an AI was trained by the researchers at New York University using more than 1600 histopathological slides from lung specimens made available by The Cancer Genome Atlas More Details and the results produced by the AI were compared with reports from experienced pathologists. AI not only told whether the ROI was normal or malignant, including the subtype, which was on par with pathologists' reports, but also produced these results within a few seconds.[34]
In non-small cell lung cancer, the identification of genetic mutations involving epidermal growth factor receptor (EGFR), gene fusions of anaplastic lymphoma kinase (ALK), and the overexpression of Programmed Death-Ligand 1 (PD-L1) predicts the prognosis and choice of treatment. The latest application of AI-DL in lung cancer is the ability to predict and also identify EGFR mutations (including subtypes), ALK, and PD-L1 expression.[35],[36],[37],[38],[39],[40]
| Lung Cancer-Management | | |
An AI can be used to map the surgical site in patients undergoing surgical resection of lung tumors and can also determine whether patients will need adjuvant chemotherapy.[41],[42],[43] In patients receiving radiotherapy, AI can also predict the occurrence of postradiation pneumonitis, albeit with limited accuracy for now.[41],[44]
In lung cancer, AI is also being used by researchers to predict treatment response and survival.[45],[46] This is achieved by analyzing the data from radiology and pathology images together with the clinical data of the patients.[47] A retrospective study by Dercle et al. using radiomics to predict the sensitivity of tumors to nivolumab, docetaxel, and gefitinib showed an area under curve (AUC) of 0.77 for nivolumab, an AUC of 0.67 for docetaxel, and an AUC of 0.82 for gefitinib.[46] Wen et al. used radiomics on the pretreatment CT scan images of 120 patients with non-small cell lung cancer (NSCLC) in combination with tumor mutation burden, clinical, and tumor morphological data. The AI was able to predict the programmed death-ligand 1 (PD-L1) expression level with an AUC = 0.839.[48]
Studies are being carried out to predict metastasis in patients diagnosed with lung cancer. Sun et al. applied CT-based radiomics to predict brain metastasis in curatively resected locally advanced NSCLC which indicates treatment failure, and concluded that AI integration provides better predictive performance.[49]
| Interstitial Lung Disease | | |
In at-risk populations such as those with familial idiopathic pulmonary fibrosis (IPF) and rheumatoid arthritis, the use of an automated computer algorithm to evaluate radiological anomalies in high-resolution CT scans has been demonstrated to objectively increase the diagnosis of interstitial lung disease (ILD).[50] Several researchers have applied AI for screening,[50],[51] diagnosis,[52],[53],[54] and to predict prognosis,[13],[55] mortality,[56] the development of lung cancer,[57] and therapeutic response in ILDs.[58],[59]
AI-DL has improved the diagnostic precision of chronic hypersensitivity pneumonitis, cryptogenic organizing pneumonia, nonspecific interstitial pneumonia, and common interstitial pneumonia patterns.[54] AI had demonstrated good performance in detecting chronic fibrosing ILD in chest X-rays.[60] A study comparing the performance of an AI-DL algorithm to that of experienced radiologists in classifying ILD using a cohort of 150 high-resolution CT images found that the AI had an accuracy of 73.3% compared to the median accuracy of all radiologists (70.7%), and outperformed 60 (66%) of the 91 radiologists.[61] A meta-analysis of 19 studies that used AI for ILD diagnosis using chest CT showed diagnostic accuracy ranging from 78% to 91%.[62]
Raghu et al. trained an ML algorithm on data from histopathology and RNA sequencing of transbronchial lung biopsies from 90 patients to create a molecular signature for usual interstitial pneumonia. The algorithm identified usual interstitial pneumonia in transbronchial lung biopsy samples with 88% specificity and 70% sensitivity when compared to standard diagnostic histopathology. In those patients in whom the CT scan showed possible or inconsistent usual interstitial pneumonia findings, the algorithm achieved a positive predictive value of 81% for underlying biopsy-proven usual interstitial pneumonia.[63]
Interstitial lung abnormalities (ILA) are high-attenuation subpleural findings of variable appearance on CT that, in some cases, progress to IPF. Researchers used AI-ML on the CT data from the chronic obstructive pulmonary disease genetic epidemiology (COPDGene) trial to predict the progression of ILA among 301 subjects and concluded that ILA progression can be accurately detected by AI systems.[51]
Although the application of AI in the diagnosis and morphological classification of ILD is promising, at the current level, the performance of AI is not satisfactory enough to rely on it entirely for the diagnosis.[62] The results produced by the AI algorithm should be interpreted in conjunction with the appropriate clinical setting and only after due consideration by a multidisciplinary team.[63]
| Asthma | | |
AI algorithms are being explored for use in asthma screening, diagnosis, identifying phenotypes, and determining asthma control and management.
To diagnose asthma, researchers have combined an AI algorithm with a number of techniques, including forced oscillation techniques to identify airway obstruction,[64] using a wearable sensor to identify wheezing sounds,[65] data from capnography,[66] clinical characteristics, and spirometry data from patient health records.[67],[68]
AI technology can analyze various transcriptomic markers to identify the genes that are crucial for the emergence of asthma and may also be used to identify different asthma phenotypes.[69],[70],[71] Researchers have used an AI algorithm to analyze hundreds of variables from 346 adult asthma patients and identified four clusters of subphenotypes with differential responses to corticosteroid treatment.[72] In another study, nuclear magnetic resonance spectroscopy analysis of exhaled breath condensate was used to differentiate asthma patients from healthy controls. ML was then used to identify three clusters of asthma patients with distinct endotypes.[73]
Researchers have combined ML and medical knowledge to assess the risk of exacerbation and asthma control among patients, which may alert patients and doctors via electronic health monitoring systems or mobile applications.[74] To determine the risk of asthma exacerbation, a team of researchers developed a wearable smartwatch for kids with asthma that collected physiological (heart rate and spirometry) and environmental data (dust, particle matter, temperature, humidity) through a variety of sensors. The data was sent to a cloud server to determine the level of the risk of exacerbation using AI. They were able to assess the asthma exacerbation risk with 80.10% ± 14.13% accuracy, and the alerts were sent to a mobile phone application for easy interpretation by children.[75] Another study used AI to predict exacerbations by analyzing the nocturnal recordings of physiological data (nocturnal heart rate, respiratory rate, relative stroke volume, and movement) from pediatric subjects with asthma through a contactless bed sensor. For identifying asthma exacerbations, their model achieved 87.4% accuracy, 47.2% sensitivity, and 96.3% specificity.[76] Researchers have even created a model using data from Twitter, Google search interests, and environmental sensor data to predict the number of emergency room visits for asthma exacerbations based on near-real-time environmental and social media data with approximately 70% precision. The researchers claim that it can be helpful for public health surveillance, emergency preparedness, and targeted patient interventions.[77]
AI technology was also used in a study where the auscultation recordings of asthma patients were analyzed using AI-ML to differentiate asthma patients with and without abnormal breath sounds.[78] AI-DL and ML have also been used to differentiate asthma, COPD, and ACO by analyzing the clinical data from patients' digital health records.[79]
AI has also been used to predict the response to corticosteroids in asthmatics. Qin et al. used AI to assess small airway thickness to predict the response to inhaled steroids in asthmatics with small airway obstruction.[80]
| Chronic Obstructive Pulmonary Disease | | |
AI models have been developed to assess the severity, prognosis, treatment response, risk of exacerbations, and mortality among COPD patients.
LDCT images obtained from lung cancer screening can be used to detect COPD changes using an AI.[81] A study demonstrated that by combining spirometry (Tiffeneau-Index) and CT data (emphysema quantification), AI can recognize patients with mild to moderate COPD even before there are any significant changes in lung function testing.[82] AI algorithm has also been trained to identify COPD patients by identifying the low serum levels of N-acetyl-glycoprotein, and lipoprotein among COPD patients compared to the controls with an accuracy of more than 80%.[83] Using data from the COPDGene study database, AI has also been used to discover specific COPD phenotypes as well as the genetic and molecular pathways causing disease development in various COPD subtypes.[84]
AI algorithms can help clinicians predict the factors responsible for mortality in COPD. Moll et al. created an AI algorithm that predicted all-cause mortality in moderate to severe COPD patients using 30 different clinical, spirometric, and imaging features as inputs. The AI identified the top predictors of mortality as 6-min walk distance, FEV1% predicted, higher BODE scores, and age. The AI identified the top imaging predictor as the pulmonary artery-to-aorta ratio.[85]
| Prediction of Acute Exacerbations in Chronic Obstructive Pulmonary Disease | | |
Various studies have been done to identify the severity and predict future exacerbations in patients with COPD using AI.[86],[87],[88],[89] This is achieved by feeding the data from patients' health records, self-reported symptoms by patients, monitoring certain parameters using sensor-enabled devices, using mobile applications, or computer software.
In a study, patients' breath sounds were recorded daily for a period of 6 months using sensor devices through telemonitoring. The data were examined using an ML algorithm, which had a detection accuracy of 78% and could predict acute exacerbation of COPD on average 4.4 days before the onset.[90] Another study employed the ML algorithm to predict COPD exacerbations among the participants in TelEPOC, a telemedicine program designed for COPD patients. The patients provided the TelEPOC database with daily reports of their heart rate, temperature, oxygen saturation, respiration rate, steps taken, and a questionnaire regarding their symptoms. Based on these findings, the researchers developed an alarm system with three levels of exacerbation (green, yellow, and red), and they were able to predict an exacerbation within the next 3 days with a receiver operating characteristic curve of 0.87.[91]
Researchers have used AI-ML for the early detection of acute respiratory failure, ventilator dependence, and mortality in patients with COPD after hospitalization with excellent predictive performance.[92]
All of these studies generally conclude that AI and ML can improve outcomes in patients at risk of developing an acute exacerbation of COPD by correlating patients' health records, symptoms, and certain clinical parameters.
| Pulmonary Function Testing | | |
Pulmonary function testing includes spirometry, measurement of lung volumes, and carbon monoxide diffusion capacity of the lungs. Currently, the interpretation of lung function tests is done by health-care workers based on published guidelines.
However, there is diversity in reporting because of inter-reader variability, which may be brought on by a lack of knowledge of lung physiology, ignorance of the recommendations, training deficiencies, or oversight.[93] In high-volume settings where thousands of reports could be created per month, this can be problematic.
The researchers used an AI-DL algorithm to automate the reporting process by analyzing 16,502 files in portable document format of spirometry reports with more than 92% accuracy.[94] Another study compared the performance of AI and pulmonologists in the interpretation of spirometry data as per the American Thoracic Society/European Respiratory Society guidelines. Shockingly, AI perfectly matched the spirometry pattern interpretations (100%) compared to 74.4% by pulmonologists and gave a correct diagnosis in 82% of cases compared to 44.6% by pulmonologists.[95]
| Pneumothorax | | |
The chest X-ray remains the most commonly used radiological investigation in the diagnosis of pneumothorax. However, a small pneumothorax may not be seen on chest X-rays. Up to 20% of them may be missed on chest X-rays but can be detected by CT scans (occult pneumothorax).[96]
Researchers have developed various AI models to identify pneumothorax in large sets of chest X-rays, mainly from open-source databases, retrained them using their own customized training dataset, and compared their performance to radiologist's readings of those X-rays with good results.[97],[98],[99],[100],[101],[102],[103],[104] AI was also used to quantify the pneumothorax by researchers, with promising results.[105]
Sometimes the performance of the AI algorithm can be limited by poor-quality radiographs, and the results can be biased due to the presence of chest tubes. A study compared the performance of the AI algorithm in the detection of pneumothorax in X-ray images with and without in-image pixel annotations of the dehiscent visceral pleura and concluded that the performance of the AI algorithm improved with annotations while also suppressing the confounding effect of chest tubes.[106]
The results of these studies show that a well-trained AI model with high-quality radiology images can detect pneumothorax with high accuracy and reliability, and the automated detection of pneumothorax in radiographs by an AI algorithm can improve patient care and outcomes.
| Pleural Effusion | | |
AI technology is being explored by researchers to automatically identify the presence of pleural effusion and also quantify the pleural effusion in radiological images.[107],[108] This can reduce the burden on health-care workers in high-volume settings and lead to improved diagnosis.
AI technology is also being explored to differentiate between benign and malignant pleural effusions.[109] Multiple parameters such as pleural fluid volume, laterality, pleural fluid tumor markers, pathology, biochemical, cytology, and histopathology reports, and various blood parameters were fed through various AI-ML models to differentiate benign and malignant pleural effusions with an AUC of >0.9.[110] AI-ML algorithms were used on the whole slide images (WSIs) of pleural effusion cytology to differentiate benign from malignancy with accuracy, sensitivity, and specificity of 91.67%, 87.50%, and 94.44%, respectively, with an AUC of 0.9526.[111]
Researchers have explored the performance of AI in detecting tuberculous pleural effusion by training four AI algorithms using a set of 28 features obtained from statistical analysis and comparing their performances with pleural fluid ADA. The most effective algorithm had a sensitivity and specificity of 90.6 and 92.3% for identifying tuberculous pleural effusion, respectively.[112]
By analyzing the clinical, radiological, and pleural fluid analysis reports using an AI, a study attempted to identify the variables associated with treatment failure among patients treated for complicated parapneumonic effusion and empyema with intrapleural tissue plasminogen activator/deoxyribonuclease (tPA/DNase). The results revealed that loculated pleural effusion, protein level, abscess or necrotizing pneumonia, and pleural thickening were the key predictors of tPA/DNase therapy failure.[113]
| SARS-CoV-2/COVID-19 Pandemic | | |
During the COVID-19 pandemic, researchers used AI for a variety of tasks, including diagnosis, triaging, management, tracing contacts, and forecasting the next outbreak.
In the very early stages of the COVID-19 pandemic, an AI epidemiology algorithm (BlueDot) was used to predict the spread of this novel infection by analyzing the data generated from the International Air Transport Association.[114] During the pandemic, various countries created digital apps to collect data such as location and symptoms which were then analyzed by AI to help in forecasting and contact tracing.[115]
Although reverse transcription polymerase chain reaction (RT-PCR) is regarded as the gold standard for the diagnosis of COVID-19, a CT thorax can be rapidly and reliably used to screen for COVID-19 pneumonia and determine the extent of lung involvement.[116],[117] Pure ground glass opacities (GGOs), usually bilateral and peripheral, were the most common CT finding in COVID-19 infection. The combination of GGOs with consolidation, reticular opacities, and other atypical features (pleural effusion and lymphadenopathy) may also be seen.[118] CT scan sensitivity and specificity for the diagnosis of COVID-19 pneumonia vary greatly between studies. In general, CT scans have good sensitivity with relatively low specificity.[119]
Several AI models (e.g., COVNet) were developed by researchers that used ML to automatically detect COVID-19 pneumonia in the radiological images by training the AI on the datasets obtained from patients diagnosed with COVID-19 pneumonia, atypical pneumonia, community-acquired pneumonia (CAP) and other nonpneumonia abnormalities. The AI models were able to achieve a sensitivity >87%, a specificity >88%, and an AUC >0.9 for the diagnosis of COVID-19 pneumonia and were also able to distinguish COVID-19 pneumonia from CAP and other lung abnormalities.[120],[121] In one study, the AI algorithm was even able to predict COVID-19 pneumonia in patients with an initial RT-PCR negative report with an accuracy of 85%.[121] Another team of researchers was able to train an AI model that identified GGOs and had an AUC of 0.95 for distinguishing between COVID-19 and bacterial pneumonia.[122] One algorithm took only 1.93 s to process a patient's CT using a dedicated graphic processor unit (GPU) with an AUC of 0.959 for the diagnosis of COVID-19 pneumonia in CT.[123]
Several AI-DL models have also been trained with chest X-rays of patients with COVID-19 pneumonia, with detection accuracy ranging from 82.9% to 98.08%.[124],[125],[126],[127],[128],[129]
| Other Pneumonias | | |
AI models were successfully trained with X-ray images to automatically detect consolidation.[130] One AI model (InceptionV3) was able to achieve an accuracy of 92.8% with a sensitivity of 93.2%, a specificity of 90.1%, and an AUC of 0.968 for detecting viral pneumonia when compared to normal chest X-rays. The AI model also achieved an accuracy of 90.7% with a sensitivity of 88.6%, a specificity of 90.9%, and an AUC of 0.94 for distinguishing bacterial from viral pneumonia.[131] Another AI model (customized VGG16) achieved 96.2% accuracy for detecting pneumonia and 93.6% accuracy for distinguishing bacterial from viral pneumonia in pediatric chest X-rays.[132]
CAD4Kids is a CAD system developed by researchers to automatically detect pneumonia in chest radiography obtained from children and also to evaluate its accuracy compared to readings by independent radiologists. CAD4Kids had an AUC of 0.850, a sensitivity of 76%, and a specificity of 80% for identifying primary-endpoint pneumonia on chest radiographs in comparison to the reference radiologist consensus reading.[133]
| Tuberculosis | | |
CAD systems were first developed to detect lesions in the chest X-rays of people suspected to have tuberculosis (TB). CAD systems detect the preset morphological features of lesions in chest X-rays to screen for TB. The only commercially available CAD system is CAD4TB, proprietary software owned by Delft Imaging Systems (Veenendaal, Netherlands). In 2018, version 6 of CAD4TB was released, which was the first version to use deep-learning technology to automatically detect the abnormalities in a chest X-ray in <15 s.[134] In another study, the researchers evaluated the efficacy of CNNs for detecting TB in chest radiographs by comparing the performances of two different CNNs: AlexNet and GoogLeNet. The authors concluded that the combination of AlexNet and GoogLeNet DCNNs was the best-performing classifier with an AUC of 0.99, which, when combined with the interpretation of a radiologist, was able to achieve an impressive 97.3% sensitivity and 100% specificity.[135] A CAD system was even used to evaluate the treatment response in TB by comparing the chest X-rays taken before starting the treatment, during the treatment, and at the end of the treatment using a subtraction image analysis algorithm, which will determine whether the patient is showing improvement from the treatment or not.[136]
AI-DL was also used in a study to identify drug-sensitive and drug-resistant TB from the radiological appearance based on a dataset of 135 images (45% sensitive TB cases, 55% resistant TB cases).[137] Although the DL algorithm achieved an underwhelming AUC of 0.66, it does leave open the possibility that in the near future, DL algorithms will become so advanced that it will be possible to achieve it.
AI technology is already being used to discover new drugs for the treatment of TB.[138],[139] Another possibility for the future is that the AI-DL can be used to analyze patients' clinical, radiological, microbiological, and treatment data to identify risk factors for the development of TB, including any novel gene associated with TB disease, predict adverse drug events, and predict the risk of relapse, drug resistance, and mortality.[140],[141],[142] AI technology may also be utilized for TB surveillance in the future.
| Bronchoscopy | | |
AI technology had been used to identify endobronchial anatomy and also to differentiate subtypes of lung cancer based on their appearance and texture in the bronchoscopic images.[143],[144]
Robotic-assisted bronchoscopy (RAB) is the latest weapon in the arsenal of interventional pulmonologists. Currently, two Food and Drug Administration-approved robotic platforms are available: Monarch™ (Auris Health) and Ion™ Endoluminal System (Intuitive Surgical).[145],[146] Using virtual 3D reconstruction using CT data, precise controls, sensors, and AI technology, RAB can reach peripheral airways further than traditional bronchoscopes.[147] Studies have reported ease of use and improved sample yield compared to traditional bronchoscope techniques.[147],[148],[149]
Rapid on-site evaluation (ROSE) is a technique where the sample obtained through bronchoscopic procedures is immediately evaluated by microscopy to ascertain the sample's adequacy, thereby improving the diagnostic accuracy. Researchers have applied DL models to the WSI of samples obtained from bronchoscopy to determine whether they are benign or malignant, with an AUC of 0.9846. The AI-based ROSE model achieved an accuracy of 84.57% when compared to the accuracy achieved by one senior cytopathologist (96.90%) and two junior cytopathologists (83.30%).[150] The future is not far off, where real-time AI-guided bronchoscopy and sampling are possibilities.
| Polysomnography | | |
The American Academy of Sleep Medicine, in its statement regarding the use of AI in sleep medicine, has opined that AI can analyze a large volume of patients' sleep physiological data and may provide novel insights into the disorders related to sleep.[151] It is also possible to program the AI to generate fast, reliable, and accurate sleep scores and associated events during polysomnography (PSG). Studies are already exploring the reliability and accuracy of AI-generated reports from the given PSG data.[152] Currently, more research and data are required for the application of AI in sleep medicine.
In the future, AI's role may expand beyond the rapid scoring of sleep. By combining the patient's clinical data, genomic data, behavioral pattern, and sleep data, it is possible that in the future, an AI may play a significant role in identifying various endotypes in sleep disorders and assisting the doctor in formulating the treatment plan.[151]
| Pulmonary Artery Hypertension | | |
AI technology is being utilized to further our understanding of pulmonary artery hypertension (PAH). For example, a study used ML on the data (levels of circulating cytokines, chemokines, and other factors) obtained from PAH patients to classify them into clusters based on their distinct proteomic immune profiles and also to predict the level of risk and 5-year transplant-free survival rates among these clusters.[153] Another study used ML on a database of 90 patients containing data on their pulmonary artery pressure (PAP), which was determined invasively, and echocardiographic estimates of PAP, which were obtained within 24 h. They found that ML was able to predict pulmonary hypertension based on broader ECHO findings, with little reliance on estimated right atrial pressure.[154]
| Lung Sound Analysis | | |
Several studies have explored the use of AI technology to automatically detect and classify abnormal breath sounds (wheezes, rhonchi, and coarse and fine crackles) from normal breath sounds among children and adult patients. This is done by analyzing the recorded breath sounds, usually obtained via electronic stethoscopes. The results are promising, as the AI can detect adventitious breath sounds with a reasonably high degree of accuracy.[155],[156]
In a study, the researchers evaluated the performances of AI and doctors for detecting wheezes and crackles in the sound files obtained from 4033 adults. The authors found that the AI had better agreement than its human counterparts when evaluating the presence of wheezes and crackles from the same audio files and concluded that the lung sound classification done by the AI is more reliable than when done by an average physician.[157]
Another study evaluated the performance of their ML model to automatically detect “croup” from cough sounds obtained from patients. Their AI model was able to achieve a sensitivity and specificity of 92.31% and 85.29% for the automatic detection of croup from recorded cough sounds.[158]
| Challenges | | |
The application of AI to digital images such as radiological and histopathology images require high-resolution images with minimal blurring and artifacts. The AI algorithm also needs to be trained with thousands of images that contain normal and abnormal findings to decrease the validation error.[159] Training an AI with a limited set of images will lead to poor sensitivity and a high number of false positives. On the other hand, researchers should keep adapting and innovating to create or improve existing AI to perform better while minimizing the risk of errors.
Hardware limitations
Creating and storing several thousand high-resolution digital images will require a large storage (local or cloud) space due to the large file size of individual images. Compiling and processing these thousands of images requires high processing power (central processing unit and GPU).
Ethics
The application of AI technology requires a large quantity of patients' personal and clinical data to be accessed and transferred between various places such as wards, laboratories, and consultation rooms which will pose a security risk of the patient's data being illegally accessed, stolen, or hacked. Apart from doctors, patients' data may be accessed by bioengineers, statisticians, etc., which may raise patient-doctor confidentiality and privacy concerns.[160] New laws may be required to address some of these issues in the near future.
Errors or bias
Data generated by AI from small cohorts may lead to erroneous recommendations or bias when applied to a larger group of patients. This can probably be eliminated by training the AI with a larger dataset so that the results generated represent a wider variety of the general population. Another important step to reduce errors and improve the results is the use of external validation of the AI algorithm using an independent dataset to determine whether the algorithm produces reliable results under various circumstances.
| Conclusion | | |
Overall, AI has the potential to transform the field of medicine by providing faster and more accurate diagnoses, improving treatment planning, and reducing health-care costs. Although AI performs at par with or better than doctors in most areas of application, it is the integration of AI and humans that has produced very good results compared to either of them alone. Thus, it is important to note that AI should be used in conjunction with human expertise and clinical judgment to ensure the best outcomes for patients. As far as medical specialties are concerned, the performance of AI technology is constrained by limited processing power, the unavailability of high-quality medical images, and/or inadequate tuning. In the future, when further advances are made in computing technology in the form of widespread use of machines with higher processing power, widespread availability of commercial high-speed internet connections, and the digitization of patients' medical records and reports, the AI technology can be fully trained on very large datasets collected from multiple countries or continents, which may unlock new dimensions with respect to how we approach the diseases of humanity.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
| References | | |
| 1. | |
| 2. | Briganti G, Le Moine O. Artificial intelligence in medicine: Today and tomorrow. Front Med (Lausanne) 2020;7:27.  |
| 3. | |
| 4. | Rajkomar A, Oren E, Chen K, Dai AM, Hajaj N, Hardt M, et al. Scalable and accurate deep learning with electronic health records. NPJ Digit Med 2018;1:18. |
| 5. | Gupta R, Srivastava D, Sahu M, Tiwari S, Ambasta RK, Kumar P. Artificial intelligence to deep learning: Machine intelligence approach for drug discovery. Mol Divers 2021;25:1315-60. |
| 6. | National Lung Screening Trial Research Team, Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395-409. |
| 7. | |
| 8. | de Koning HJ, Meza R, Plevritis SK, ten Haaf K, Munshi VN, Jeon J, et al. Benefits and harms of computed tomography lung cancer screening strategies: A comparative modeling study for the U.S. Preventive Services Task Force. Ann Intern Med 2014;160:311-20. |
| 9. | US Preventive Services Task Force, Krist AH, Davidson KW, Mangione CM, Barry MJ, Cabana M, et al. Screening for lung cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2021;325:962-70. |
| 10. | Jonas DE, Reuland DS, Reddy SM, Nagle M, Clark SD, Weber RP, et al. Screening for lung cancer with low-dose computed tomography: Updated evidence report and systematic review for the US Preventive Services Task Force. JAMA 2021;325:971-87. |
| 11. | Svoboda E. Artificial intelligence is improving the detection of lung cancer. Nature 2020;587:S20-2. |
| 12. | Lodwick GS, Keats TE, Dorst JP. The coding of roentgen images for computer analysis as applied to lung cancer. Radiology 1963;81:185-200. |
| 13. | Chassagnon G, De Margerie-Mellon C, Vakalopoulou M, Marini R, Hoang-Thi TN, Revel MP, et al. Artificial intelligence in lung cancer: Current applications and perspectives. Jpn J Radiol 2023;41:235-44. |
| 14. | Sim Y, Chung MJ, Kotter E, Yune S, Kim M, Do S, et al. Deep convolutional neural network-based software improves radiologist detection of malignant lung nodules on chest radiographs. Radiology 2020;294:199-209. |
| 15. | Yoo H, Kim KH, Singh R, Digumarthy SR, Kalra MK. Validation of a deep learning algorithm for the detection of malignant pulmonary nodules in chest radiographs. JAMA Netw Open 2020;3:e2017135. |
| 16. | Li L, Liu Z, Huang H, Lin M, Luo D. Evaluating the performance of a deep learning-based computer-aided diagnosis (DL-CAD) system for detecting and characterizing lung nodules: Comparison with the performance of double reading by radiologists. Thorac Cancer 2019;10:183-92. |
| 17. | |
| 18. | Shan H, Padole A, Homayounieh F, Kruger U, Khera RD, Nitiwarangkul C, et al. Competitive performance of a modularized deep neural network compared to commercial algorithms for low-dose CT image reconstruction. Nat Mach Intell 2019;1:269-76. |
| 19. | Chiu HY, Chao HS, Chen YM. Application of artificial intelligence in lung cancer. Cancers (Basel) 2022;14:1370. |
| 20. | Ardila D, Kiraly AP, Bharadwaj S, Choi B, Reicher JJ, Peng L, et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nat Med 2019;25:954-61. |
| 21. | |
| 22. | Hawkins S, Wang H, Liu Y, Garcia A, Stringfield O, Krewer H, et al. Predicting malignant nodules from screening CT scans. J Thorac Oncol 2016;11:2120-8. |
| 23. | Lee SH, Lee SM, Goo JM, Kim KG, Kim YJ, Park CM. Usefulness of texture analysis in differentiating transient from persistent part-solid nodules (PSNs): A retrospective study. PLoS One 2014;9:e85167. |
| 24. | Chae HD, Park CM, Park SJ, Lee SM, Kim KG, Goo JM. Computerized texture analysis of persistent part-solid ground-glass nodules: Differentiation of preinvasive lesions from invasive pulmonary adenocarcinomas. Radiology 2014;273:285-93. |
| 25. | Wilson R, Devaraj A. Radiomics of pulmonary nodules and lung cancer. Transl Lung Cancer Res 2017;6:86-91. |
| 26. | Shmatko A, Ghaffari Laleh N, Gerstung M, Kather JN. Artificial intelligence in histopathology: Enhancing cancer research and clinical oncology. Nat Cancer 2022;3:1026-38. |
| 27. | Fernandez G, Prastawa M, Madduri AS, Scott R, Marami B, Shpalensky N, et al. Development and validation of an AI-enabled digital breast cancer assay to predict early-stage breast cancer recurrence within 6 years. Breast Cancer Res 2022;24:93. |
| 28. | Saldanha OL, Muti HS, Grabsch HI, Langer R, Dislich B, Kohlruss M, et al. Direct prediction of genetic aberrations from pathology images in gastric cancer with swarm learning. Gastric Cancer 2023;26:264-74. |
| 29. | Litjens G, Sánchez CI, Timofeeva N, Hermsen M, Nagtegaal I, Kovacs I, et al. Deep learning as a tool for increased accuracy and efficiency of histopathological diagnosis. Sci Rep 2016;6:26286. |
| 30. | Nagpal K, Foote D, Liu Y, Chen PH, Wulczyn E, Tan F, et al. Development and validation of a deep learning algorithm for improving Gleason scoring of prostate cancer. Npj Digit Med 2019;2:1-10. |
| 31. | Wu M, Yan C, Liu H, Liu Q. Automatic classification of ovarian cancer types from cytological images using deep convolutional neural networks. Biosci Rep 2018;38:BSR20180289. |
| 32. | Zhang L, Le Lu, Nogues I, Summers RM, Liu S, Yao J. DeepPap: Deep convolutional networks for cervical cell classification. IEEE J Biomed Health Inform 2017;21:1633-43. |
| 33. | Chang HY, Jung CK, Woo JI, Lee S, Cho J, Kim SW, et al. Artificial intelligence in pathology. J Pathol Transl Med 2019;53:1-12. |
| 34. | |
| 35. | Wang S, Shi J, Ye Z, Dong D, Yu D, Zhou M, et al. Predicting EGFR mutation status in lung adenocarcinoma on computed tomography image using deep learning. Eur Respir J 2019;53:1800986. |
| 36. | Li S, Ding C, Zhang H, Song J, Wu L. Radiomics for the prediction of EGFR mutation subtypes in non-small cell lung cancer. Med Phys 2019;46:4545-52. |
| 37. | Coudray N, Ocampo PS, Sakellaropoulos T, Narula N, Snuderl M, Fenyö D, et al. Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning. Nat Med 2018;24:1559-67. |
| 38. | Song L, Zhu Z, Mao L, Li X, Han W, Du H, et al. Clinical, conventional CT and radiomic feature-based machine learning models for predicting ALK rearrangement status in lung adenocarcinoma patients. Front Oncol 2020;10:369. |
| 39. | Zhu Y, Liu YL, Feng Y, Yang XY, Zhang J, Chang DD, et al. A CT-derived deep neural network predicts for programmed death ligand-1 expression status in advanced lung adenocarcinomas. Ann Transl Med 2020;8:930. |
| 40. | Jiang M, Sun D, Guo Y, Guo Y, Xiao J, Wang L, et al. Assessing PD-L1 expression level by radiomic features from PET/CT in nonsmall cell lung cancer patients: An initial result. Acad Radiol 2020;27:171-9. |
| 41. | Bourbonne V, Da-Ano R, Jaouen V, Lucia F, Dissaux G, Bert J, et al. Radiomics analysis of 3D dose distributions to predict toxicity of radiotherapy for lung cancer. Radiother Oncol 2021;155:144-50. |
| 42. | Yoshiyasu N, Kojima F, Hayashi K, Bando T. Radiomics technology for identifying early-stage lung adenocarcinomas suitable for sublobar resection. J Thorac Cardiovasc Surg 2021;162:477-85.e1. |
| 43. | Jones GD, Brandt WS, Shen R, Sanchez-Vega F, Tan KS, Martin A, et al. A genomic-pathologic annotated risk model to predict recurrence in early-stage lung adenocarcinoma. JAMA Surg 2021;156:e205601. |
| 44. | Krafft SP, Rao A, Stingo F, Briere TM, Court LE, Liao Z, et al. The utility of quantitative CT radiomics features for improved prediction of radiation pneumonitis. Med Phys 2018;45:5317-24. |
| 45. | Le VH, Kha QH, Hung TN, Le NQ. Risk score generated from CT-based radiomics signatures for overall survival prediction in non-small cell lung cancer. Cancers (Basel) 2021;13:3616. |
| 46. | Dercle L, Fronheiser M, Lu L, Du S, Hayes W, Leung DK, et al. Identification of non-small cell lung cancer sensitive to systemic cancer therapies using radiomics. Clin Cancer Res 2020;26:2151-62. |
| 47. | Echle A, Rindtorff NT, Brinker TJ, Luedde T, Pearson AT, Kather JN. Deep learning in cancer pathology: A new generation of clinical biomarkers. Br J Cancer 2021;124:686-96. |
| 48. | Wen Q, Yang Z, Dai H, Feng A, Li Q. Radiomics study for predicting the expression of PD-L1 and tumor mutation burden in non-small cell lung cancer based on CT images and clinicopathological features. Front Oncol 2021;11:620246. |
| 49. | Sun F, Chen Y, Chen X, Sun X, Xing L. CT-based radiomics for predicting brain metastases as the first failure in patients with curatively resected locally advanced non-small cell lung cancer. Eur J Radiol 2021;134:109411. |
| 50. | Rosas IO, Yao J, Avila NA, Chow CK, Gahl WA, Gochuico BR. Automated quantification of high-resolution CT scan findings in individuals at risk for pulmonary fibrosis. Chest 2011;140:1590-7. |
| 51. | |
| 52. | Huang S, Lee F, Miao R, Si Q, Lu C, Chen Q. A deep convolutional neural network architecture for interstitial lung disease pattern classification. Med Biol Eng Comput 2020;58:725-37. |
| 53. | Agarwala S, Kale M, Kumar D, Swaroop R, Kumar A, Kumar Dhara A, et al. Deep learning for screening of interstitial lung disease patterns in high-resolution CT images. Clin Radiol 2020;75:481.e1-8. |
| 54. | Choe J, Hwang HJ, Seo JB, Lee SM, Yun J, Kim MJ, et al. Content-based image retrieval by using deep learning for interstitial lung disease diagnosis with chest CT. Radiology 2022;302:187-97. |
| 55. | Aliboni L, Dias OM, Baldi BG, Sawamura MV, Chate RC, Carvalho CR, et al. A convolutional neural network approach to quantify lung disease progression in patients with fibrotic hypersensitivity pneumonitis (HP). Acad Radiol 2022;29:e149-56. |
| 56. | Xu W, Wu W, Zheng Y, Chen Z, Tao X, Zhang D, et al. A computed tomography radiomics-based prediction model on interstitial lung disease in anti-MDA5-positive dermatomyositis. Front Med (Lausanne) 2021;8:768052. |
| 57. | Liang CH, Liu YC, Wan YL, Yun CH, Wu WJ, López-González R, et al. Quantification of cancer-developing idiopathic pulmonary fibrosis using whole-lung texture analysis of HRCT images. Cancers (Basel) 2021;13:5600. |
| 58. | Lancaster L, Goldin J, Trampisch M, Kim GH, Ilowite J, Homik L, et al. Effects of nintedanib on quantitative lung fibrosis score in idiopathic pulmonary fibrosis. Open Respir Med J 2020;14:22-31. |
| 59. | Kim GH, Goldin JG, Hayes W, Oh A, Soule B, Du S. The value of imaging and clinical outcomes in a phase II clinical trial of a lysophosphatidic acid receptor antagonist in idiopathic pulmonary fibrosis. Ther Adv Respir Dis 2021;15:17534666211004238. |
| 60. | Nishikiori H, Hirota K, Suzuki T, Takagi Y, Honda S, Ikeda K, et al. Validation of the artificial intelligence software to detect chronic fibrosing interstitial lung diseases in chest X-ray. Eur Respir J 2021;58(Suppl 65). Available from: https://erj. ersjournals.com/content/58/suppl_65/oa1211. [Last accessed on 2023 Jan 22]. |
| 61. | Walsh SL, Calandriello L, Silva M, Sverzellati N. Deep learning for classifying fibrotic lung disease on high-resolution computed tomography: A case-cohort study. Lancet Respir Med 2018;6:837-45. |
| 62. | Soffer S, Morgenthau AS, Shimon O, Barash Y, Konen E, Glicksberg BS, et al. Artificial intelligence for interstitial lung disease analysis on chest computed tomography: A systematic review. Acad Radiol 2022;29 Suppl 2:S226-35. |
| 63. | Raghu G, Flaherty KR, Lederer DJ, Lynch DA, Colby TV, Myers JL, et al. Use of a molecular classifier to identify usual interstitial pneumonia in conventional transbronchial lung biopsy samples: A prospective validation study. Lancet Respir Med 2019;7:487-96. |
| 64. | Amaral JL, Lopes AJ, Veiga J, Faria AC, Melo PL. High-accuracy detection of airway obstruction in asthma using machine learning algorithms and forced oscillation measurements. Comput Methods Programs Biomed 2017;144:113-25. |
| 65. | Oletic D, Bilas V. Asthmatic wheeze detection from compressively sensed respiratory sound spectra. IEEE J Biomed Health Inform 2018;22:1406-14. |
| 66. | Singh OP, Palaniappan R, Malarvilli MB. Automatic Quantitative Analysis of Human Respired Carbon Dioxide Waveform for Asthma and Non-Asthma Classification Using Support Vector Machine. IEEE Access; 2018. Available from: http://dx.doi.org/10.1109/ACCESS.20180.2871091. [Last accessed on 2023 Feb 09]. |
| 67. | Kaur H, Sohn S, Wi CI, Ryu E, Park MA, Bachman K, et al. Automated chart review utilizing natural language processing algorithm for asthma predictive index. BMC Pulm Med 2018;18:34. |
| 68. | Spathis D, Vlamos P. Diagnosing asthma and chronic obstructive pulmonary disease with machine learning. Health Informatics J 2019;25:811-27. |
| 69. | Fang F, Pan J, Li Y, Li Y, Feng X, Wang J. Identification of potential transcriptomic markers in developing asthma: An integrative analysis of gene expression profiles. Mol Immunol 2017;92:38-44. |
| 70. | Pandey G, Pandey OP, Rogers AJ, Ahsen ME, Hoffman GE, Raby BA, et al. A nasal brush-based classifier of asthma identified by machine learning analysis of nasal RNA Sequence Data. Sci Rep 2018;8:8826. |
| 71. | Ross MK, Yoon J, van der Schaar A, van der Schaar M. Discovering pediatric asthma phenotypes on the basis of response to controller medication using machine learning. Ann Am Thorac Soc 2018;15:49-58. |
| 72. | Wu W, Bang S, Bleecker ER, Castro M, Denlinger L, Erzurum SC, et al. Multiview cluster analysis identifies variable corticosteroid response phenotypes in severe asthma. Am J Respir Crit Care Med 2019;199:1358-67. |
| 73. | Sinha A, Desiraju K, Aggarwal K, Kutum R, Roy S, Lodha R, et al. Exhaled breath condensate metabolome clusters for endotype discovery in asthma. J Transl Med 2017;15:262. |
| 74. | Khasha R, Sepehri MM, Mahdaviani SA. An ensemble learning method for asthma control level detection with leveraging medical knowledge-based classifier and supervised learning. J Med Syst 2019;43:158. |
| 75. | Hosseini A, Buonocore CM, Hashemzadeh S, Hojaiji H, Kalantarian H, Sideris C, et al. Feasibility of a secure wireless sensing smartwatch application for the self-management of pediatric asthma. Sensors (Basel) 2017;17:1780. |
| 76. | Huffaker MF, Carchia M, Harris BU, Kethman WC, Murphy TE, Sakarovitch CC, et al. Passive nocturnal physiologic monitoring enables early detection of exacerbations in children with asthma. A proof-of-concept study. Am J Respir Crit Care Med 2018;198:320-8. |
| 77. | Ram S, Zhang W, Williams M, Pengetnze Y. Predicting asthma-related emergency department visits using big data. IEEE J Biomed Health Inform 2015;19:1216-23. |
| 78. | Hafke-Dys H, Kuźnar-Kamińska B, Grzywalski T, Maciaszek A, Szarzyński K, Kociński J. Artificial intelligence approach to the monitoring of respiratory sounds in asthmatic patients. Front Physiol 2021;12:745635. |
| 79. | Joumaa H, Sigogne R, Maravic M, Perray L, Bourdin A, Roche N. Artificial intelligence to differentiate asthma from COPD in medico-administrative databases. BMC Pulm Med 2022;22:357. |
| 80. | Qin Y, Wang J, Han Y, Lu L. Deep learning algorithms-based CT images in glucocorticoid therapy in asthma children with small airway obstruction. J Healthc Eng 2021;2021:5317403. |
| 81. | Tang LY, Coxson HO, Lam S, Leipsic J, Tam RC, Sin DD. Towards large-scale case-finding: Training and validation of residual networks for detection of chronic obstructive pulmonary disease using low-dose CT. Lancet Digit Health 2020;2:e259-67. |
| 82. | Fischer AM, Varga-Szemes A, Martin SS, Sperl JI, Sahbaee P, Neumann D, et al. Artificial intelligence-based fully automated per lobe segmentation and emphysema-quantification based on chest computed tomography compared with global initiative for chronic obstructive lung disease severity of smokers. J Thorac Imaging 2020;35 Suppl 1:S28-34. |
| 83. | Zheng H, Hu Y, Dong L, Shu Q, Zhu M, Li Y, et al. Predictive diagnosis of chronic obstructive pulmonary disease using serum metabolic biomarkers and least-squares support vector machine. J Clin Lab Anal 2021;35:e23641. |
| 84. | Castaldi PJ, Boueiz A, Yun J, Estepar RS, Ross JC, Washko G, et al. Machine learning characterization of COPD subtypes: Insights from the COPD gene study. Chest 2020;157:1147-57. |
| 85. | Moll M, Qiao D, Regan EA, Hunninghake GM, Make BJ, Tal-Singer R, et al. Machine learning and prediction of all-cause mortality in COPD. Chest 2020;158:952-64. |
| 86. | Hussain A, Choi HE, Kim HJ, Aich S, Saqlain M, Kim HC. Forecast the exacerbation in patients of chronic obstructive pulmonary disease with clinical indicators using machine learning techniques. Diagnostics (Basel) 2021;11:829. |
| 87. | Chmiel FP, Burns DK, Pickering JB, Blythin A, Wilkinson TM, Boniface MJ. Prediction of chronic obstructive pulmonary disease exacerbation events by using patient self-reported data in a digital health app: Statistical evaluation and machine learning approach. JMIR Med Inform 2022;10:e26499. |
| 88. | Zeng S, Arjomandi M, Tong Y, Liao ZC, Luo G. Developing a machine learning model to predict severe chronic obstructive pulmonary disease exacerbations: Retrospective cohort study. J Med Internet Res 2022;24:e28953. |
| 89. | Kor CT, Li YR, Lin PR, Lin SH, Wang BY, Lin CH. Explainable machine learning model for predicting first-time acute exacerbation in patients with chronic obstructive pulmonary disease. J Pers Med 2022;12:228. |
| 90. | Fernandez-Granero MA, Sanchez-Morillo D, Leon-Jimenez A. An artificial intelligence approach to early predict symptom-based exacerbations of COPD. Biotechnol Biotechnol Equip 2018;32:778-84. |
| 91. | |
| 92. | Liao KM, Liu CF, Chen CJ, Shen YT. Machine learning approaches for predicting acute respiratory failure, ventilator dependence, and mortality in chronic obstructive pulmonary disease. Diagnostics (Basel) 2021;11:2396. |
| 93. | Giri PC, Chowdhury AM, Bedoya A, Chen H, Lee HS, Lee P, et al. Application of machine learning in pulmonary function assessment where are we now and where are we going? Front Physiol 2021;12:678540. |
| 94. | Wang Y, Li Y, Chen W, Zhang C, Liang L, Huang R, et al. Deep learning for spirometry quality assurance with spirometric indices and curves. Respir Res 2022;23:98. |
| 95. | Topalovic M, Das N, Burgel PR, Daenen M, Derom E, Haenebalcke C, et al. Artificial intelligence outperforms pulmonologists in the interpretation of pulmonary function tests. Eur Respir J 2019;53:1801660. |
| 96. | Brar MS, Bains I, Brunet G, Nicolaou S, Ball CG, Kirkpatrick AW. Occult pneumothoraces truly occult or simply missed: Redux. J Trauma 2010;69:1335-7. |
| 97. | Taylor AG, Mielke C, Mongan J. Automated detection of moderate and large pneumothorax on frontal chest X-rays using deep convolutional neural networks: A retrospective study. PLoS Med 2018;15:e1002697. |
| 98. | Park S, Lee SM, Kim N, Choe J, Cho Y, Do KH, et al. Application of deep learning-based computer-aided detection system: Detecting pneumothorax on chest radiograph after biopsy. Eur Radiol 2019;29:5341-8. |
| 99. | Kitamura G, Deible C. Retraining an open-source pneumothorax detecting machine learning algorithm for improved performance to medical images. Clin Imaging 2020;61:15-9. |
| 100. | Wang H, Gu H, Qin P, Wang J. CheXLocNet: Automatic localization of pneumothorax in chest radiographs using deep convolutional neural networks. PLoS One 2020;15:e0242013. |
| 101. | Cho Y, Kim JS, Lim TH, Lee I, Choi J. Detection of the location of pneumothorax in chest X-rays using small artificial neural networks and a simple training process. Sci Rep 2021;11:13054. |
| 102. | Thian YL, Ng D, Hallinan JT, Jagmohan P, Sia SY, Tan CH, et al. Deep learning systems for pneumothorax detection on chest radiographs: A multicenter external validation study. Radiol Artif Intell 2021;3:e200190. |
| 103. | Hillis JM, Bizzo BC, Mercaldo S, Chin JK, Newbury-Chaet I, Digumarthy SR, et al. Evaluation of an artificial intelligence model for detection of pneumothorax and tension pneumothorax in chest radiographs. JAMA Netw Open 2022;5:e2247172. |
| 104. | Feng S, Liu Q, Patel A, Bazai SU, Jin CK, Kim JS, et al. Automated pneumothorax triaging in chest X-rays in the New Zealand population using deep-learning algorithms. J Med Imaging Radiat Oncol 2022;66:1035-43. |
| 105. | Kim D, Lee JH, Kim SW, Hong JM, Kim SJ, Song M, et al. Quantitative measurement of pneumothorax using artificial intelligence management model and clinical application. Diagnostics (Basel) 2022;12:1823. |
| 106. | Rueckel J, Huemmer C, Fieselmann A, Ghesu FC, Mansoor A, Schachtner B, et al. Pneumothorax detection in chest radiographs: Optimizing artificial intelligence system for accuracy and confounding bias reduction using in-image annotations in algorithm training. Eur Radiol 2021;31:7888-900. |
| 107. | Ibrahim RF, Yhiea NM, Mohammed AM, Mohamed AM. Pleural effusion detection using machine learning and deep learning based on computer vision. In: Hassanien AE, Snášel V, Tang M, Sung TW, Chang KC, editors. Proceedings of the 8 th International Conference on Advanced Intelligent Systems and Informatics 2022. Cham: Springer International Publishing; 2023. p. 199-210. |
| 108. | Huang T, Yang R, Shen L, Feng A, Li L, He N, et al. Deep transfer learning to quantify pleural effusion severity in chest X-rays. BMC Med Imaging 2022;22:100. |
| 109. | |
| 110. | Li Y, Tian S, Huang Y, Dong W. Driverless artificial intelligence framework for the identification of malignant pleural effusion. Transl Oncol 2021;14:100896. |
| 111. | Xie X, Fu CC, Lv L, Ye Q, Yu Y, Fang Q, et al. Deep convolutional neural network-based classification of cancer cells on cytological pleural effusion images. Mod Pathol 2022;35:609-14. |
| 112. | Ren Z, Hu Y, Xu L. Identifying tuberculous pleural effusion using artificial intelligence machine learning algorithms. Respir Res 2019;20:220. |
| 113. | Khemasuwan D, Sorensen J, Griffin DC. Predictive variables for failure in administration of intrapleural tissue plasminogen activator/deoxyribonuclease in patients with complicated parapneumonic effusions/empyema. Chest 2018;154:550-6. |
| 114. | Bogoch II, Watts A, Thomas-Bachli A, Huber C, Kraemer MU, Khan K. Pneumonia of unknown aetiology in Wuhan, China: Potential for international spread via commercial air travel. J Travel Med 2020;27:taaa008. |
| 115. | Lalmuanawma S, Hussain J, Chhakchhuak L. Applications of machine learning and artificial intelligence for COVID-19 (SARS-CoV-2) pandemic: A review. Chaos Solitons Fractals 2020;139:110059. |
| 116. | Fang Y, Zhang H, Xie J, Lin M, Ying L, Pang P, et al. Sensitivity of chest CT for COVID-19: Comparison to RT-PCR. Radiology 2020;296:E115-7. |
| 117. | Li K, Wu J, Wu F, Guo D, Chen L, Fang Z, et al. The clinical and chest CT features associated with severe and critical COVID-19 pneumonia. Invest Radiol 2020;55:327-31. |
| 118. | Majidi H, Niksolat F. Chest CT in patients suspected of COVID-19 infection: A reliable alternative for RT-PCR. Am J Emerg Med 2020;38:2730-2. |
| 119. | Kovács A, Palásti P, Veréb D, Bozsik B, Palkó A, Kincses ZT. The sensitivity and specificity of chest CT in the diagnosis of COVID-19. Eur Radiol 2021;31:2819-24. |
| 120. | Li L, Qin L, Xu Z, Yin Y, Wang X, Kong B, et al. Using artificial intelligence to detect COVID-19 and community-acquired pneumonia based on pulmonary CT: Evaluation of the diagnostic accuracy. Radiology 2020;296:E65-71. |
| 121. | Wang S, Kang B, Ma J, Zeng X, Xiao M, Guo J, et al. A deep learning algorithm using CT images to screen for corona virus disease (COVID-19). Eur Radiol 2021;31:6096-104. |
| 122. | Song Y, Zheng S, Li L, Zhang X, Zhang X, Huang Z, et al. Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images. IEEE/ACM Trans Comput Biol Bioinform 2021;18:2775-80. |
| 123. | |
| 124. | Apostolopoulos ID, Mpesiana TA. COVID-19: Automatic detection from X-ray images utilizing transfer learning with convolutional neural networks. Phys Eng Sci Med 2020;43:635-40. |
| 125. | Wang L, Lin ZQ, Wong A. COVID-Net: A tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-ray images. Sci Rep 2020;10:19549. |
| 126. | Sethy PK, Behera SK, Ratha PK, Biswas P. Detection of coronavirus disease (COVID-19) based on deep features and support vector machine. Int J Math Eng Manag Sci 2020;5:643-51. |
| 127. | Hemdan EE, Shouman MA, Karar ME. COVIDX-Net: A Framework of Deep Learning Classifiers to Diagnose COVID-19 in X-Ray Images arXiv; 2020. Available from: http://arxiv.org/abs/2003.11055. [Last accessed on 2023 Feb 02]. |
| 128. | Ozturk T, Talo M, Yildirim EA, Baloglu UB, Yildirim O, Rajendra Acharya U. Automated detection of COVID-19 cases using deep neural networks with X-ray images. Comput Biol Med 2020;121:103792. |
| 129. | Narin A, Kaya C, Pamuk Z. Automatic detection of coronavirus disease (COVID-19) using X-ray images and deep convolutional neural networks. Pattern Anal Appl 2021;24:1207-20. |
| 130. | Behzadi-Khormouji H, Rostami H, Salehi S, Derakhshande-Rishehri T, Masoumi M, Salemi S, et al. Deep learning, reusable and problem-based architectures for detection of consolidation on chest X-ray images. Comput Methods Programs Biomed 2020;185:105162. |
| 131. | Kermany DS, Goldbaum M, Cai W, Valentim CC, Liang H, Baxter SL, et al. Identifying medical diagnoses and treatable diseases by image-based deep learning. Cell 2018;172:1122-31.e9. |
| 132. | Rajaraman S, Candemir S, Kim I, Thoma G, Antani S. Visualization and interpretation of convolutional neural network predictions in detecting pneumonia in pediatric chest radiographs. Appl Sci (Basel) 2018;8:1715. |
| 133. | Mahomed N, van Ginneken B, Philipsen RH, Melendez J, Moore DP, Moodley H, et al. Computer-aided diagnosis for World Health Organization-defined chest radiograph primary-endpoint pneumonia in children. Pediatr Radiol 2020;50:482-91. |
| 134. | Murphy K, Habib SS, Zaidi SM, Khowaja S, Khan A, Melendez J, et al. Computer aided detection of tuberculosis on chest radiographs: An evaluation of the CAD4TB v6 system. Sci Rep 2020;10:5492. |
| 135. | Lakhani P, Sundaram B. Deep learning at chest radiography: Automated classification of pulmonary tuberculosis by using convolutional neural networks. Radiology 2017;284:574-82. |
| 136. | Lieberman R, Kwong H, Liu B, Huang H. Computer-assisted detection (CAD) methodology for early detection of response to pharmaceutical therapy in tuberculosis patients. Proc SPIE Int Soc Opt Eng 2009;7260:726030. |
| 137. | Jaeger S, Juarez-Espinosa OH, Candemir S, Poostchi M, Yang F, Kim L, et al. Detecting drug-resistant tuberculosis in chest radiographs. Int J Comput Assist Radiol Surg 2018;13:1915-25. |
| 138. | Usmani SS, Bhalla S, Raghava GP. Prediction of antitubercular peptides from sequence information using ensemble classifier and hybrid features. Front Pharmacol 2018;9:954. |
| 139. | Lane T, Russo DP, Zorn KM, Clark AM, Korotcov A, Tkachenko V, et al. Comparing and validating machine learning models for mycobacterium tuberculosis drug discovery. Mol Pharm 2018;15:4346-60. |
| 140. | Doshi R, Falzon D, Thomas BV, Temesgen Z, Sadasivan L, Migliori GB, et al. Tuberculosis control, and the where and why of artificial intelligence. ERJ Open Res 2017;3:00056-2017. |
| 141. | Jamal S, Khubaib M, Gangwar R, Grover S, Grover A, Hasnain SE. Artificial intelligence and machine learning based prediction of resistant and susceptible mutations in Mycobacterium tuberculosis. Sci Rep 2020;10:5487. |
| 142. | Liang S, Ma J, Wang G, Shao J, Li J, Deng H, et al. The application of artificial intelligence in the diagnosis and drug resistance prediction of pulmonary tuberculosis. Front Med (Lausanne) 2022;9:935080. |
| 143. | Feng PH, Lin YT, Lo CM. A machine learning texture model for classifying lung cancer subtypes using preliminary bronchoscopic findings. Med Phys 2018;45:5509-14. |
| 144. | Yoo JY, Kang SY, Park JS, Cho YJ, Park SY, Yoon HI, et al. Deep learning for anatomical interpretation of video bronchoscopy images. Sci Rep 2021;11:23765. |
| 145. | Agrawal A, Hogarth DK, Murgu S. Robotic bronchoscopy for pulmonary lesions: A review of existing technologies and clinical data. J Thorac Dis 2020;12:3279-86. |
| 146. | Lu M, Nath S, Semaan RW. A review of robotic-assisted bronchoscopy platforms in the sampling of peripheral pulmonary lesions. J Clin Med 2021;10:5678. |
| 147. | Chen AC, Gillespie CT. Robotic endoscopic airway challenge: REACH assessment. Ann Thorac Surg 2018;106:293-7. |
| 148. | Chaddha U, Kovacs SP, Manley C, Hogarth DK, Cumbo-Nacheli G, Bhavani SV, et al. Robot-assisted bronchoscopy for pulmonary lesion diagnosis: Results from the initial multicenter experience. BMC Pulm Med 2019;19:243. |
| 149. | Chen AC, Pastis NJ Jr., Mahajan AK, Khandhar SJ, Simoff MJ, Machuzak MS, et al. Robotic bronchoscopy for peripheral pulmonary lesions: A multicenter pilot and feasibility study (BENEFIT). Chest 2021;159:845-52. |
| 150. | Ai D, Hu Q, Chao YC, Fu CC, Yuan W, Lv L, et al. Artificial intelligence-based rapid on-site cytopathological evaluation for bronchoscopy examinations. Intell Based Med 2022;6:100069. |
| 151. | Goldstein CA, Berry RB, Kent DT, Kristo DA, Seixas AA, Redline S, et al. Artificial intelligence in sleep medicine: An American Academy of Sleep Medicine position statement. J Clin Sleep Med 2020;16:605-7. |
| 152. | Jansri U, Chirakalwasan N, Chaitusaney B, Busayakanon S, Khongjui T, Tretriluxana S. Automatic Sleep Data Scoring by Artificial Intelligence: A Pilot Study in Thai Population. In: 2021 7 th International Conference on Engineering, Applied Sciences and Technology (ICEAST); 2021. p. 29-33. |
| 153. | Sweatt AJ, Hedlin HK, Balasubramanian V, Hsi A, Blum LK, Robinson WH, et al. Discovery of distinct immune phenotypes using machine learning in pulmonary arterial hypertension. Circ Res 2019;124:904-19. |
| 154. | Leha A, Hellenkamp K, Unsöld B, Mushemi-Blake S, Shah AM, Hasenfuß G, et al. A machine learning approach for the prediction of pulmonary hypertension. PLoS One 2019;14:e0224453. |
| 155. | Kevat A, Kalirajah A, Roseby R. Artificial intelligence accuracy in detecting pathological breath sounds in children using digital stethoscopes. Respir Res 2020;21:253. |
| 156. | Grzywalski T, Piecuch M, Szajek M, Bręborowicz A, Hafke-Dys H, Kociński J, et al. Practical implementation of artificial intelligence algorithms in pulmonary auscultation examination. Eur J Pediatr 2019;178:883-90. |
| 157. | Melbye H, Ravn J, Pabiszczak M, Bongo LA, Solis JCA. Validity of deep learning algorithms for detecting wheezes and crackles from lung sound recordings in adults [Internet]. medRxiv; 2022. p. 2022.11.18.22282442. Available from: https://www.medrxiv.org/content/10.1101/2022.11.18.22282442v1. [Last cited on 2023 Feb 11]. |
| 158. | Sharan RV, Abeyratne UR, Swarnkar VR, Porter P. Automatic croup diagnosis using cough sound recognition. IEEE Trans Biomed Eng 2019;66:485-95. |
| 159. | Campanella G, Hanna MG, Geneslaw L, Miraflor A, Werneck Krauss Silva V, Busam KJ, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nat Med 2019;25:1301-9. |
| 160. | Stoeklé HC, Mamzer-Bruneel MF, Frouart CH, Le Tourneau C, Laurent-Puig P, Vogt G, et al. Molecular tumor boards: Ethical issues in the new era of data medicine. Sci Eng Ethics 2018;24:307-22. |
[Figure 1]
[Table 1], [Table 2]
|
Search Pubmed for