ABSTRACT
Background and study aims Low-quality colonoscopy increases cancer risk but measuring quality remains challenging. We developed an automated, interactive assessment of colonoscopy quality (AI-CQ) using machine learning (ML). Methods Based on quality guidelines, metrics selected for AI development included insertion time (IT), withdrawal time (WT), polyp detection rate (PDR), and polyps per colonoscopy (PPC). Two novel metrics were also developed: HQ-WT (time during withdrawal with clear image) and WT-PT (withdrawal time subtracting polypectomy time). The model was pre-trained using a self-supervised vision transformer on unlabeled colonoscopy images and then finetuned for multi-label classification on another mutually exclusive colonoscopy image dataset. A timeline of video predictions and metric calculations were presented to clinicians in addition to the raw video using a web-based application. The model was externally validated using 50 colonoscopies at a second hospital. Results The AI-CQ accuracy to identify cecal intubation was 88%. IT ( P = 0.99) and WT ( P = 0.99) were highly correlated between manual and AI-CQ measurements with a median difference of 1.5 seconds and 4.5 seconds, respectively. AI-CQ PDR did not significantly differ from manual PDR (47.6% versus 45.5%, P = 0.66). Retroflexion was correctly identified in 95.2% and number of right colon evaluations in 100% of colonoscopies. HQ-WT was 45.9% of, and significantly correlated with ( P = 0.85) WT time. Conclusions An interactive AI assessment of colonoscopy skill can automatically assess quality. We propose that this tool can be utilized to rapidly identify and train providers in need of remediation.
ABSTRACT
Purpose To evaluate the impact of an artificial intelligence (AI) assistant for lung cancer screening on multinational clinical workflows. Materials and Methods An AI assistant for lung cancer screening was evaluated on two retrospective randomized multireader multicase studies where 627 (141 cancer-positive cases) low-dose chest CT cases were each read twice (with and without AI assistance) by experienced thoracic radiologists (six U.S.-based or six Japan-based radiologists), resulting in a total of 7524 interpretations. Positive cases were defined as those within 2 years before a pathology-confirmed lung cancer diagnosis. Negative cases were defined as those without any subsequent cancer diagnosis for at least 2 years and were enriched for a spectrum of diverse nodules. The studies measured the readers' level of suspicion (on a 0-100 scale), country-specific screening system scoring categories, and management recommendations. Evaluation metrics included the area under the receiver operating characteristic curve (AUC) for level of suspicion and sensitivity and specificity of recall recommendations. Results With AI assistance, the radiologists' AUC increased by 0.023 (0.70 to 0.72; P = .02) for the U.S. study and by 0.023 (0.93 to 0.96; P = .18) for the Japan study. Scoring system specificity for actionable findings increased 5.5% (57% to 63%; P < .001) for the U.S. study and 6.7% (23% to 30%; P < .001) for the Japan study. There was no evidence of a difference in corresponding sensitivity between unassisted and AI-assisted reads for the U.S. (67.3% to 67.5%; P = .88) and Japan (98% to 100%; P > .99) studies. Corresponding stand-alone AI AUC system performance was 0.75 (95% CI: 0.70, 0.81) and 0.88 (95% CI: 0.78, 0.97) for the U.S.- and Japan-based datasets, respectively. Conclusion The concurrent AI interface improved lung cancer screening specificity in both U.S.- and Japan-based reader studies, meriting further study in additional international screening environments. Keywords: Assistive Artificial Intelligence, Lung Cancer Screening, CT Supplemental material is available for this article. Published under a CC BY 4.0 license.
Subject(s)
Artificial Intelligence , Early Detection of Cancer , Lung Neoplasms , Tomography, X-Ray Computed , Humans , Lung Neoplasms/diagnosis , Lung Neoplasms/epidemiology , Japan , United States/epidemiology , Retrospective Studies , Early Detection of Cancer/methods , Female , Male , Middle Aged , Aged , Sensitivity and Specificity , Radiographic Image Interpretation, Computer-Assisted/methodsABSTRACT
PRCIS: A deep learning model trained on macular OCT imaging studies detected clinically significant functional glaucoma progression and was also able to predict future progression. OBJECTIVE: To use macular optical coherence tomography (OCT) imaging to predict the future and detect concurrent visual field progression, respectively, using deep learning. DESIGN: A retrospective cohort study. SUBJECTS: A pretraining data set was comprised of 7,702,201 B-scan images from 151,389 macular OCT studies. The progression detection task included 3902 macular OCT imaging studies from 1534 eyes of 828 patients with glaucoma, and the progression prediction task included 1346 macular OCT studies from 1205 eyes of 784. METHODS: A novel deep learning method was developed to detect glaucoma progression and predict future progression using macular OCT, based on self-supervised pretraining of a vision transformer (ViT) model on a large, unlabeled data set of OCT images. Glaucoma progression was defined as a mean deviation (MD) rate of change of ≤ -0.5 dB/year over 5 consecutive Humphrey visual field tests, and rapid progression was defined as MD change ≤ -1 dB/year. MAIN OUTCOME MEASURES: Diagnostic performance of the ViT model for prediction of future visual field progression and detection of concurrent visual field progression using area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. RESULTS: The model distinguished stable eyes from progressing eyes, achieving an AUC of 0.90 (95% CI, 0.88-0.91). Rapid progression was detected with an AUC of 0.92 (95% CI, 0.91-0.93). The model also demonstrated high predictive ability for forecasting future glaucoma progression, with an AUC of 0.85 (95% CI 0.83-0.87). Rapid progression was predicted with an AUC of 0.84 (95% CI 0.81-0.86). CONCLUSIONS: A deep learning model detected clinically significant functional glaucoma progression using macular OCT imaging studies and was also able to predict future progression. Early identification of patients undergoing glaucoma progression or at high risk for future progression may aid in clinical decision-making.
Subject(s)
Deep Learning , Glaucoma , Humans , Visual Fields , Tomography, Optical Coherence/methods , Retrospective Studies , Intraocular Pressure , Retinal Ganglion Cells , Glaucoma/diagnosis , Visual Field Tests/methodsABSTRACT
Importance: Multimodal generative artificial intelligence (AI) methodologies have the potential to optimize emergency department care by producing draft radiology reports from input images. Objective: To evaluate the accuracy and quality of AI-generated chest radiograph interpretations in the emergency department setting. Design, Setting, and Participants: This was a retrospective diagnostic study of 500 randomly sampled emergency department encounters at a tertiary care institution including chest radiographs interpreted by both a teleradiology service and on-site attending radiologist from January 2022 to January 2023. An AI interpretation was generated for each radiograph. The 3 radiograph interpretations were each rated in duplicate by 6 emergency department physicians using a 5-point Likert scale. Main Outcomes and Measures: The primary outcome was any difference in Likert scores between radiologist, AI, and teleradiology reports, using a cumulative link mixed model. Secondary analyses compared the probability of each report type containing no clinically significant discrepancy with further stratification by finding presence, using a logistic mixed-effects model. Physician comments on discrepancies were recorded. Results: A total of 500 ED studies were included from 500 unique patients with a mean (SD) age of 53.3 (21.6) years; 282 patients (56.4%) were female. There was a significant association of report type with ratings, with post hoc tests revealing significantly greater scores for AI (mean [SE] score, 3.22 [0.34]; P < .001) and radiologist (mean [SE] score, 3.34 [0.34]; P < .001) reports compared with teleradiology (mean [SE] score, 2.74 [0.34]) reports. AI and radiologist reports were not significantly different. On secondary analysis, there was no difference in the probability of no clinically significant discrepancy between the 3 report types. Further stratification of reports by presence of cardiomegaly, pulmonary edema, pleural effusion, infiltrate, pneumothorax, and support devices also yielded no difference in the probability of containing no clinically significant discrepancy between the report types. Conclusions and Relevance: In a representative sample of emergency department chest radiographs, results suggest that the generative AI model produced reports of similar clinical accuracy and textual quality to radiologist reports while providing higher textual quality than teleradiologist reports. Implementation of the model in the clinical workflow could enable timely alerts to life-threatening pathology while aiding imaging interpretation and documentation.
Subject(s)
Artificial Intelligence , Emergency Medical Services , Humans , Female , Middle Aged , Male , Retrospective Studies , Emergency Service, Hospital , RadiologistsABSTRACT
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia in clinical practice and has a well-established association with coronary artery bypass graft (CABG) surgery. Being able to predict post-operative AF (POAF) may improve surgical outcomes. This study retrospectively assembled a large cohort of 3,807 first-time CABG patients with no prior AF to study factors that contribute to occurrence of POAF, in addition to testing models that may predict its incidence. Several clinical features with established relevance to POAF were extracted from the EHR, along with a record of medications administered intra-operatively. Tests of performance with logistic regression, decision tree, and neural network predictive models showed slight improvements when incorporating medication information. Analysis of the clinical and medications data indicate that there may be effects contributing to POAF incidence captured in the medication administration records. Our results show that improved predictive performance is achievable by incorporating a record of medications administered intra-operatively.
ABSTRACT
Machine-learning models for medical tasks can match or surpass the performance of clinical experts. However, in settings differing from those of the training dataset, the performance of a model can deteriorate substantially. Here we report a representation-learning strategy for machine-learning models applied to medical-imaging tasks that mitigates such 'out of distribution' performance problem and that improves model robustness and training efficiency. The strategy, which we named REMEDIS (for 'Robust and Efficient Medical Imaging with Self-supervision'), combines large-scale supervised transfer learning on natural images and intermediate contrastive self-supervised learning on medical images and requires minimal task-specific customization. We show the utility of REMEDIS in a range of diagnostic-imaging tasks covering six imaging domains and 15 test datasets, and by simulating three realistic out-of-distribution scenarios. REMEDIS improved in-distribution diagnostic accuracies up to 11.5% with respect to strong supervised baseline models, and in out-of-distribution settings required only 1-33% of the data for retraining to match the performance of supervised models retrained using all available data. REMEDIS may accelerate the development lifecycle of machine-learning models for medical imaging.
Subject(s)
Machine Learning , Supervised Machine Learning , Diagnostic ImagingABSTRACT
BACKGROUND: Functional lumen imaging probe (FLIP) Panometry is performed at the time of sedated endoscopy and evaluates esophageal motility in response to distension. This study aimed to develop and test an automated artificial intelligence (AI) platform that could interpret FLIP Panometry studies. METHODS: The study cohort included 678 consecutive patients and 35 asymptomatic controls that completed FLIP Panometry during endoscopy and high-resolution manometry (HRM). "True" study labels for model training and testing were assigned by experienced esophagologists per a hierarchical classification scheme. The supervised, deep learning, AI model generated FLIP Panometry heatmaps from raw FLIP data and based on convolutional neural networks assigned esophageal motility labels using a two-stage prediction model. Model performance was tested on a 15% held-out test set (n = 103); the remainder of the studies were utilized for model training (n = 610). KEY RESULTS: "True" FLIP labels across the entire cohort included 190 (27%) "normal," 265 (37%) "not normal/not achalasia," and 258 (36%) "achalasia." On the test set, both the Normal/Not normal and the achalasia/not achalasia models achieved an accuracy of 89% (with 89%/88% recall, 90%/89% precision, respectively). Of 28 patients with achalasia (per HRM) in the test set, 0 were predicted as "normal" and 93% as "achalasia" by the AI model. CONCLUSIONS: An AI platform provided accurate interpretation of FLIP Panometry esophageal motility studies from a single center compared with the impression of experienced FLIP Panometry interpreters. This platform may provide useful clinical decision support for esophageal motility diagnosis from FLIP Panometry studies performed at the time of endoscopy.
Subject(s)
Esophageal Achalasia , Esophageal Motility Disorders , Humans , Artificial Intelligence , Esophageal Achalasia/diagnosis , Endoscopy, Gastrointestinal , Manometry/methods , Gastrointestinal Transit , Esophagogastric JunctionABSTRACT
Background Patients with congenital heart disease (CHD) are at risk for the development of low cardiac output and other physiologic derangements, which could be detected early through continuous stroke volume (SV) measurement. Unfortunately, existing SV measurement methods are limited in the clinic because of their invasiveness (eg, thermodilution), location (eg, cardiac magnetic resonance imaging), or unreliability (eg, bioimpedance). Multimodal wearable sensing, leveraging the seismocardiogram, a sternal vibration signal associated with cardiomechanical activity, offers a means to monitoring SV conveniently, affordably, and continuously. However, it has not been evaluated in a population with significant anatomical and physiological differences (ie, children with CHD) or compared against a true gold standard (ie, cardiac magnetic resonance). Here, we present the feasibility of wearable estimation of SV in a diverse CHD population (N=45 patients). Methods and Results We used our chest-worn wearable biosensor to measure baseline ECG and seismocardiogram signals from patients with CHD before and after their routine cardiovascular magnetic resonance imaging, and derived features from the measured signals, predominantly systolic time intervals, to estimate SV using ridge regression. Wearable signal features achieved acceptable SV estimation (28% error with respect to cardiovascular magnetic resonance imaging) in a held-out test set, per cardiac output measurement guidelines, with a root-mean-square error of 11.48 mL and R2 of 0.76. Additionally, we observed that using a combination of electrical and cardiomechanical features surpassed the performance of either modality alone. Conclusions A convenient wearable biosensor that estimates SV enables remote monitoring of cardiac function and may potentially help identify decompensation in patients with CHD.
Subject(s)
Heart Defects, Congenital , Wearable Electronic Devices , Child , Heart , Heart Defects, Congenital/complications , Heart Defects, Congenital/diagnosis , Humans , Stroke Volume/physiology , ThermodilutionABSTRACT
Background Developing deep learning models for radiology requires large data sets and substantial computational resources. Data set size limitations can be further exacerbated by distribution shifts, such as rapid changes in patient populations and standard of care during the COVID-19 pandemic. A common partial mitigation is transfer learning by pretraining a "generic network" on a large nonmedical data set and then fine-tuning on a task-specific radiology data set. Purpose To reduce data set size requirements for chest radiography deep learning models by using an advanced machine learning approach (supervised contrastive [SupCon] learning) to generate chest radiography networks. Materials and Methods SupCon helped generate chest radiography networks from 821 544 chest radiographs from India and the United States. The chest radiography networks were used as a starting point for further machine learning model development for 10 prediction tasks (eg, airspace opacity, fracture, tuberculosis, and COVID-19 outcomes) by using five data sets comprising 684 955 chest radiographs from India, the United States, and China. Three model development setups were tested (linear classifier, nonlinear classifier, and fine-tuning the full network) with different data set sizes from eight to 85. Results Across a majority of tasks, compared with transfer learning from a nonmedical data set, SupCon reduced label requirements up to 688-fold and improved the area under the receiver operating characteristic curve (AUC) at matching data set sizes. At the extreme low-data regimen, training small nonlinear models by using only 45 chest radiographs yielded an AUC of 0.95 (noninferior to radiologist performance) in classifying microbiology-confirmed tuberculosis in external validation. At a more moderate data regimen, training small nonlinear models by using only 528 chest radiographs yielded an AUC of 0.75 in predicting severe COVID-19 outcomes. Conclusion Supervised contrastive learning enabled performance comparable to state-of-the-art deep learning models in multiple clinical tasks by using as few as 45 images and is a promising method for predictive modeling with use of small data sets and for predicting outcomes in shifting patient populations. © RSNA, 2022 Online supplemental material is available for this article.
Subject(s)
COVID-19 , Deep Learning , Humans , Radiography, Thoracic/methods , Radiographic Image Interpretation, Computer-Assisted/methods , Pandemics , COVID-19/diagnostic imaging , Retrospective Studies , Radiography , Machine LearningABSTRACT
BACKGROUND: Racial bias is a key concern regarding the development, validation, and implementation of machine learning (ML) models in clinical settings. Despite the potential of bias to propagate health disparities, racial bias in clinical ML has yet to be thoroughly examined and best practices for bias mitigation remain unclear. OBJECTIVE: Our objective was to perform a scoping review to characterize the methods by which the racial bias of ML has been assessed and describe strategies that may be used to enhance algorithmic fairness in clinical ML. METHODS: A scoping review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) Extension for Scoping Reviews. A literature search using PubMed, Scopus, and Embase databases, as well as Google Scholar, identified 635 records, of which 12 studies were included. RESULTS: Applications of ML were varied and involved diagnosis, outcome prediction, and clinical score prediction performed on data sets including images, diagnostic studies, clinical text, and clinical variables. Of the 12 studies, 1 (8%) described a model in routine clinical use, 2 (17%) examined prospectively validated clinical models, and the remaining 9 (75%) described internally validated models. In addition, 8 (67%) studies concluded that racial bias was present, 2 (17%) concluded that it was not, and 2 (17%) assessed the implementation of bias mitigation strategies without comparison to a baseline model. Fairness metrics used to assess algorithmic racial bias were inconsistent. The most commonly observed metrics were equal opportunity difference (5/12, 42%), accuracy (4/12, 25%), and disparate impact (2/12, 17%). All 8 (67%) studies that implemented methods for mitigation of racial bias successfully increased fairness, as measured by the authors' chosen metrics. Preprocessing methods of bias mitigation were most commonly used across all studies that implemented them. CONCLUSIONS: The broad scope of medical ML applications and potential patient harms demand an increased emphasis on evaluation and mitigation of racial bias in clinical ML. However, the adoption of algorithmic fairness principles in medicine remains inconsistent and is limited by poor data availability and ML model reporting. We recommend that researchers and journal editors emphasize standardized reporting and data availability in medical ML studies to improve transparency and facilitate evaluation for racial bias.
ABSTRACT
OBJECTIVE: The objective of our study was to determine if the waveform from a simple pulse oximeter-like device could be used to accurately assess intravascular volume status in cirrhosis. METHODS: Patients with cirrhosis underwent waveform recording as well as serum brain natriuretic peptide (BNP) on the day of their cardiac catheterization where invasive cardiac pressures were measured. Waveforms were processed to generate features for machine learning models in order to predict the filling pressures (regression) or to classify the patients as volume overloaded or not (defined as an LVEDP>15). RESULTS: Nine of 26 patients (35%) had intravascular volume overload. Regression analysis using PPG features (R2 = 0.66) was superior to BNP (R2 = 0.22). Linear discriminant analysis correctly classified patients with an accuracy of 78%, sensitivity of 60%, positive predictive value of 90%, and an AUROC of 0.87. CONCLUSIONS: Machine learning-enhanced analysis of pulse ox waveforms can estimate intravascular volume overload with a higher accuracy than conventionally measured BNP.
Subject(s)
Natriuretic Peptide, Brain , Oximetry , Humans , Liver Cirrhosis/diagnosis , Machine Learning , OxygenABSTRACT
High-resolution manometry (HRM) is the primary procedure used to diagnose esophageal motility disorders. Its manual interpretation and classification, including evaluation of swallow-level outcomes and then derivation of a study-level diagnosis based on Chicago Classification (CC), may be limited by inter-rater variability and inaccuracy of an individual interpreter. We hypothesized that an automatic diagnosis platform using machine learning and artificial intelligence approaches could be developed to accurately identify esophageal motility diagnoses. Further, a multi-stage modeling framework, akin to the step-wise approach of the CC, was utilized to leverage advantages of a combination of machine learning approaches including deep-learning models and feature-based models. Models were trained and tested using a dataset comprised of 1741 patients' HRM studies with CC diagnoses assigned by expert physician raters. In the swallow-level stage, three models based on convolutional neural networks (CNNs) were developed to predict swallow type and swallow pressurization (test accuracies of 0.88 and 0.93, respectively), and integrated relaxation pressure (IRP)(regression model with test error of 4.49 mmHg). At the study-level stage, model selection from families of the expert-knowledge-based rule models, xgboost models and artificial neural network(ANN) models were conducted. A simple model-agnostic strategy of model balancing motivated by Bayesian principles was utilized, which gave rise to model averaging weighted by precision scores. The averaged (blended) models and individual models were compared and evaluated, of which the best performance on test dataset is 0.81 in top-1 prediction, 0.92 in top-2 predictions. This is the first artificial-intelligence style model to automatically predict esophageal motility (CC) diagnoses from HRM studies using raw multi-swallow data and it achieved high accuracy. Thus, this proposed modeling framework could be broadly applied to assist with HRM interpretation in a clinical setting.
Subject(s)
Artificial Intelligence , Esophageal Motility Disorders , Bayes Theorem , Esophageal Motility Disorders/diagnosis , Humans , Machine Learning , Manometry/methodsABSTRACT
OBJECTIVE: Tracking changes in hemodynamic congestion and the consequent proactive readjustment of treatment has shown efficacy in reducing hospitalizations for patients with heart failure (HF). However, the cost-prohibitive nature of these invasive sensing systems precludes their usage in the large patient population affected by HF. The objective of this research is to estimate the changes in pulmonary artery mean pressure (PAM) and pulmonary capillary wedge pressure (PCWP) following vasodilator infusion during right heart catheterization (RHC), using changes in simultaneously recorded wearable seismocardiogram (SCG) signals captured with a small wearable patch. METHODS: A total of 20 patients with HF (20% women, median age 55 (interquartile range (IQR), 44-64) years, ejection fraction 24 (IQR, 16-43)) were fitted with a wearable sensing patch and underwent RHC with vasodilator challenge. We divided the dataset randomly into a training-testing set (n = 15) and a separate validation set (n = 5). We developed globalized (population) regression models to estimate changes in PAM and PCWP from the changes in simultaneously recorded SCG. RESULTS: The regression model estimated both pressures with good accuracies: root-mean-square-error (RMSE) of 2.5 mmHg and R2 of 0.83 for estimating changes in PAM, and RMSE of 1.9 mmHg and R2 of 0.93 for estimating changes in PCWP for the training-testing set, and RMSE of 2.7 mmHg and R2 of 0.81 for estimating changes in PAM, and RMSE of 2.9 mmHg and R2 of 0.95 for estimating changes in PCWP for the validation set respectively. CONCLUSION: Changes in wearable SCG signals may be used to track acute changes in intracardiac hemodynamics in patients with HF. SIGNIFICANCE: This method holds promise in tracking longitudinal changes in hemodynamic congestion in hemodynamically-guided remote home monitoring and treatment for patients with HF.
Subject(s)
Heart Failure , Wearable Electronic Devices , Feasibility Studies , Female , Heart Failure/diagnosis , Heart Failure/therapy , Hemodynamics , Humans , Machine Learning , Male , Middle Aged , Vasodilator AgentsABSTRACT
BACKGROUND: This study aimed to build and evaluate a deep learning, artificial intelligence (AI) model to automatically classify swallow types based on raw data from esophageal high-resolution manometry (HRM). METHODS: HRM studies on patients with no history of esophageal surgery were collected including 1,741 studies with 26,115 swallows labeled by swallow type (normal, hypercontractile, weak-fragmented, failed, and premature) by an expert interpreter per the Chicago Classification. The dataset was stratified and split into train/validation/test datasets for model development. Long short-term memory (LSTM), a type of deep-learning AI model, was trained and evaluated. The overall performance and detailed per-swallow type performance were analyzed. The interpretations of the supine swallows in a single study were further used to generate an overall classification of peristalsis. KEY RESULTS: The LSTM model for swallow type yielded accuracies from the train/validation/test datasets of 0.86/0.81/0.83. The model's interpretation for study-level classification of peristalsis yielded accuracy of 0.88 in the test dataset. Among model misclassification, 535/698 (77%) swallows and 25/35 (71%) studies were to adjacent categories, for example, normal to weak or normal to ineffective, respectively. CONCLUSIONS AND INFERENCES: A deep-learning AI model can automatically and accurately identify the Chicago Classification swallow types and peristalsis classification from raw HRM data. While future work to refine this model and incorporate overall manometric diagnoses are needed, this study demonstrates the role that AI will serve in the interpretation and classification of esophageal HRM studies.
Subject(s)
Deep Learning , Esophageal Motility Disorders , Artificial Intelligence , Deglutition , Esophageal Motility Disorders/diagnosis , Humans , Manometry , PeristalsisABSTRACT
In light of the recent Coronavirus disease (COVID-19) pandemic, peripheral oxygen saturation (SpO2) has shown to be amongst the vital signs most indicative of deterioration in persons with COVID-19. To allow for the continuous monitoring of SpO2, we attempted to demonstrate accurate SpO2 estimation using our custom chest-based wearable patch biosensor, capable of measuring electrocardiogram (ECG) and photoplethysmogram (PPG) signals with high fidelity. Through a breath-hold protocol, we collected physiological data with a wide dynamic range of SpO2 from 20 subjects. The ratio of ratios (R) used in pulse oximetry to estimate SpO2 was robustly extracted from the red and infrared PPG signals during the breath-hold segments using novel feature extraction and PPGgreen-based outlier rejection algorithms. Through subject independent training, we achieved a low root-mean-square error (RMSE) of 2.64 ± 1.14% and a Pearson correlation coefficient (PCC) of 0.89. With subject-specific calibration, we further reduced the RMSE to 2.27 ± 0.76% and increased the PCC to 0.91. In addition, we showed that calibration is more efficiently accomplished by standardizing and focusing on the duration of breath-hold rather than the resulting range in SpO2. The accurate SpO2 estimation provided by our custom biosensor and the algorithms provide research opportunities for a wide range of disease and wellness monitoring applications.
Subject(s)
COVID-19 , Monitoring, Physiologic/instrumentation , Wearable Electronic Devices , Biosensing Techniques , COVID-19/diagnosis , Electrocardiography , Humans , Oximetry , Oxygen , Oxygen Saturation , Photoplethysmography , SternumABSTRACT
Chest radiography (CXR) is the most widely-used thoracic clinical imaging modality and is crucial for guiding the management of cardiothoracic conditions. The detection of specific CXR findings has been the main focus of several artificial intelligence (AI) systems. However, the wide range of possible CXR abnormalities makes it impractical to detect every possible condition by building multiple separate systems, each of which detects one or more pre-specified conditions. In this work, we developed and evaluated an AI system to classify CXRs as normal or abnormal. For training and tuning the system, we used a de-identified dataset of 248,445 patients from a multi-city hospital network in India. To assess generalizability, we evaluated our system using 6 international datasets from India, China, and the United States. Of these datasets, 4 focused on diseases that the AI was not trained to detect: 2 datasets with tuberculosis and 2 datasets with coronavirus disease 2019. Our results suggest that the AI system trained using a large dataset containing a diverse array of CXR abnormalities generalizes to new patient populations and unseen diseases. In a simulated workflow where the AI system prioritized abnormal cases, the turnaround time for abnormal cases reduced by 7-28%. These results represent an important step towards evaluating whether AI can be safely used to flag cases in a general setting where previously unseen abnormalities exist. Lastly, to facilitate the continued development of AI models for CXR, we release our collected labels for the publicly available dataset.
Subject(s)
COVID-19/diagnostic imaging , Radiographic Image Interpretation, Computer-Assisted/methods , Tuberculosis/diagnostic imaging , Adult , Aged , Algorithms , Case-Control Studies , China , Deep Learning , Female , Humans , India , Male , Middle Aged , Radiography, Thoracic , United StatesABSTRACT
Objective: Pregnancy requires a complex physiological adaptation of the maternal cardiovascular system, which is disrupted in women with pregnancies complicated by preeclampsia, putting them at higher risk of future cardiovascular events. The measurement of body movements in response to cardiac ejection via ballistocardiogram (BCG) can be used to assess cardiovascular hemodynamics noninvasively in women with preeclampsia. Methods: Using a previously validated, modified weighing scale for assessment of cardiovascular hemodynamics through measurement of BCG and electrocardiogram (ECG) signals, we collected serial measurements throughout pregnancy and postpartum and analyzed data in 30 women with preeclampsia and 23 normotensive controls. Using BCG and ECG signals, we extracted measures of cardiac output, J-wave amplitude × heart rate (J-amp × HR). Mixed-effect models with repeated measures were used to compare J-amp × HRs between groups at different time points in pregnancy and postpartum. Results: In normotensive controls, the J-amp × HR was significantly lower early postpartum (E-PP) compared with the second trimester (T2; p = 0.016) and third trimester (T3; p = 0.001). Women with preeclampsia had a significantly lower J-amp × HR compared with normotensive controls during the first trimester (T1; p = 0.026). In the preeclampsia group, there was a trend toward an increase in J-amp × HR from T1 to T2 and then a drop in J-amp × HR at T3 and further drop at E-PP. Conclusions: We observe cardiac hemodynamic changes consistent with those reported using well-validated tools. In pregnancies complicated by preeclampsia, the maximal force of contraction is lower, suggesting lower cardiac output and a trend in hemodynamics consistent with the hyperdynamic disease model of preeclampsia.
ABSTRACT
Background and study aims Storage of full-length endoscopic procedures is becoming increasingly popular. To facilitate large-scale machine learning (ML) focused on clinical outcomes, these videos must be merged with the patient-level data in the electronic health record (EHR). Our aim was to present a method of accurately linking patient-level EHR data with cloud stored colonoscopy videos. Methods This study was conducted at a single academic medical center. Most procedure videos are automatically uploaded to the cloud server but are identified only by procedure time and procedure room. We developed and then tested an algorithm to match recorded videos with corresponding exams in the EHR based upon procedure time and room and subsequently extract frames of interest. Results Among 28,611 total colonoscopies performed over the study period, 21,170 colonoscopy videos in 20,420 unique patients (54.2â% male, median age 58) were matched to EHR data. Of 100 randomly sampled videos, appropriate matching was manually confirmed in all. In total, these videos represented 489,721 minutes of colonoscopy performed by 50 endoscopists (median 214 colonoscopies per endoscopist). The most common procedure indications were polyp screening (47.3â%), surveillance (28.9â%) and inflammatory bowel disease (9.4â%). From these videos, we extracted procedure highlights (identified by image capture; mean 8.5 per colonoscopy) and surrounding frames. Conclusions We report the successful merging of a large database of endoscopy videos stored with limited identifiers to rich patient-level data in a highly accurate manner. This technique facilitates the development of ML algorithms based upon relevant patient outcomes.
ABSTRACT
High-resolution manometry (HRM) is the primary method for diagnosing esophageal motility disorders and its interpretation and classification are based on variables (features) from data of each swallow. Modeling and learning the semantics directly from raw swallow data could not only help automate the feature extraction, but also alleviate the bias from pre-defined features. With more than 32-thousand raw swallow data, a generative model using the approach of variational auto-encoder (VAE) was developed, which, to our knowledge, is the first deep-learning-based unsupervised model on raw esophageal manometry data. The VAE model was reformulated to include different types of loss motivated by domain knowledge and tuned with different hyper-parameters. Training of the VAE model was found sensitive on the learning rate and hence the evidence lower bound objective (ELBO) was further scaled by the data dimension. Case studies showed that the dimensionality of latent space have a big impact on the learned semantics. In particular, cases with 4-dimensional latent variables were found to encode various physiologically meaningful contraction patterns, including strength, propagation pattern as well as sphincter relaxation. Cases with so-called hybrid L2 loss seemed to better capture the coherence of contraction/relaxation transition. Discriminating capability was further evaluated using simple linear discriminative analysis (LDA) on predicting swallow type and swallow pressurization, which yields clustering patterns consistent with clinical impression. The current work on modeling and understanding swallow-level data will guide the development of study-level models for automatic diagnosis as the next stage.