Introduction
As the Veterans Administration (VA) continues to innovate in the realm of healthcare, artificial intelligence (AI) is playing an increasingly pivotal role in enhancing the quality and efficiency of care for veterans. AI technologies are being integrated across a range of applications, from advanced diagnostic tools to predictive analytics, addressing various aspects of veteran health and well-being. These innovations span diverse use cases including physical therapy, cardiac surgery, medication adherence, and mental health monitoring. By leveraging AI, the VA aims to provide more personalized, accurate, and timely care, ultimately improving patient outcomes and operational efficiencies. The following use cases highlight key AI-driven projects within the VA, showcasing how these technologies are transforming healthcare delivery and supporting the unique needs of veterans.
Use Cases
-
1. Artificial Intelligence physical therapy app
The Artificial Intelligence Physical Therapy App serves as a supportive tool for physical therapists, designed to be data source agnostic. It accepts input from various wearable sensors and analyzes the collected data to provide actionable feedback to therapists in a clear and understandable format. This app aims to enhance the effectiveness of physical therapy by enabling therapists to make informed decisions based on real-time data from their patients.
-
2. Artificial intelligence coach in cardiac surgery
The Artificial Intelligence Coach in Cardiac Surgery is designed to identify misalignments in the mental models of team members during complex surgical tasks. This project focuses on safety-critical domains, such as healthcare and aviation, where discrepancies in shared mental models can result in preventable errors. By detecting these misalignments, the AI coach lays the groundwork for computer-assisted interventions that aim to enhance teamwork and support cognitive functions in the operating room, ultimately improving patient safety and surgical outcomes.
-
3. AI Cure
AICURE is a mobile application designed to monitor patient adherence to orally prescribed medications during clinical trials or pharmaceutical-sponsored drug studies. By tracking medication intake, the app aims to improve compliance and provide valuable data to researchers, enhancing the overall effectiveness of clinical studies and ensuring that patients receive the intended therapeutic benefits.
-
4.Acute kidney injury (AKI)
The Acute Kidney Injury (AKI) project is a collaborative effort with Google DeepMind aimed at developing AI solutions for the early detection of AKI, which can range from minor kidney function loss to complete kidney failure. The AI system is designed to identify cases of AKI that may arise as a complication of other illnesses, facilitating timely intervention and improving patient outcomes through enhanced monitoring and diagnosis.
-
5. Assessing lung function in health and disease
The Assessing Lung Function in Health and Disease project utilizes artificial intelligence to help healthcare professionals identify predictors of both normal and abnormal lung function, as well as sleep parameters. By analyzing relevant data, the AI system aims to enhance the understanding of respiratory health, enabling more accurate assessments and interventions for patients with varying lung function profiles.
-
6. Automated eye movement analysis and diagnostic prediction of neurological disease
The Automated Eye Movement Analysis and Diagnostic Prediction project utilizes artificial intelligence to recursively analyze previously collected data, enhancing the quality and accuracy of automated algorithms. This system is designed to screen for markers of various neurological diseases, including traumatic brain injury, Parkinson’s disease, and stroke, thereby aiding in early diagnosis and intervention.
-
7. Automatic speech transcription engines to aid scoring neuropsychological tests
The Automatic Speech Transcription Engines project focuses on analyzing the cognitive decline of older Veterans Affairs (VA) patients through automated speech transcription. Digitally recorded speech responses are processed using multiple AI-based speech-to-text engines, and the resulting transcriptions are combined to minimize or eliminate the need for manual transcription. This automation streamlines the scoring of neuropsychological tests, improving efficiency and accuracy in assessing cognitive health.
-
8. CuraPatient
CuraPatient is an innovative remote tool designed to empower patients in managing their health conditions without the necessity of in-person provider visits. Utilizing artificial intelligence, the platform enables users to create personalized profiles for tracking their health, enrolling in relevant programs, managing insurance details, and scheduling appointments. This tool aims to enhance patient engagement and streamline healthcare management.
-
9. Digital command center
The Digital Command Center initiative aims to centralize all data within a medical center and leverage predictive prescriptive analytics to enhance hospital performance. By consolidating information, healthcare leaders can make informed decisions that optimize operations, improve patient care, and increase overall efficiency within the facility.
-
10. Disentangling dementia patterns using artificial intelligence on brain imaging and electrophysiological data
The Disentangling Dementia Patterns project is a collaborative effort aimed at creating a deep learning framework to identify and predict various dementia patterns observed in MRI and EEG data. This initiative also explores the potential of these imaging modalities as biomarkers for different types of dementia and epilepsy disorders. The VA is conducting retrospective chart reviews to support this research, enhancing understanding of dementia and improving diagnostic capabilities.
-
11. Machine learning (ML) for enhanced diagnostic error detection and ML classification of protein electrophoresis text
The Machine Learning for Enhanced Diagnostic Error Detection project involves researchers conducting chart reviews to gather true/false positive annotations for patient records. This data is then used to create vector embeddings, enabling similarity-based retrieval of unlabeled records that are similar to labeled ones in a semi-supervised approach. The goal is to utilize machine learning as a filtering mechanism following rules-based retrieval, thereby improving diagnostic specificity. The inputs for embedding will include high-value structured data related to stroke risk and relevant prior text notes.
-
12. Behavidence
Behavidence is a mental health tracking application designed for veterans. By downloading the app onto their smartphones, users can monitor their phone usage patterns, which are then compared to a digital phenotype representing individuals with confirmed mental health diagnoses. This comparison aims to provide insights into mental health status and promote awareness, potentially facilitating early intervention and support for veterans.
-
13. Machine learning tools to predict outcomes of hospitalized VA patients
The Machine Learning Tools to Predict Outcomes of Hospitalized VA Patients project is an Institutional Review Board (IRB)-approved study that investigates the application of machine learning techniques to forecast health outcomes for VA patients. The study specifically targets predictions related to Alzheimer’s disease, rehospitalization rates, and Clostridioides difficile infections, aiming to enhance patient care and resource allocation through improved predictive analytics.
-
14. Nediser reports QA
Nediser is an advanced artificial intelligence system designed to function as a “radiology resident,” continuously trained to assist radiologists in verifying X-ray properties within their reports. This AI tool can select normal templates, detect hardware issues, assess patella alignment, evaluate leg length discrepancies, and measure Cobb angles. By providing these capabilities, Nediser enhances the accuracy and efficiency of radiological assessments, supporting radiologists in their diagnostic processes.
-
15. Precision medicine PTSD and suicidality diagnostic and predictive tool
The Precision Medicine PTSD and Suicidality Diagnostic and Predictive Tool is designed to analyze various real-time inputs to predict and diagnose episodes of post-traumatic stress disorder (PTSD) and suicidality. This model aims to provide early warnings for potential episodes, facilitating timely and accurate diagnoses. Additionally, it seeks to enhance understanding of the short- and long-term effects of stress, particularly in extreme situations, contributing to better management and treatment of PTSD among veterans.
-
16. Prediction of Veterans' Suicidal Ideation following Transition from Military Service
The Prediction of Veterans’ Suicidal Ideation project employs machine learning techniques to identify factors that predict suicidal thoughts among veterans transitioning from military service. The analysis is based on data collected from a web-based survey that captures veterans’ experiences within three months of separation and continues every six months for the first three years post-service. This initiative aims to enhance understanding of the mental health challenges faced by veterans and inform preventive measures.
-
17. PredictMod
PredictMod is an innovative project that leverages artificial intelligence to explore the potential for predicting diabetes risk based on the composition of the gut microbiome. By analyzing microbiome data, the project aims to uncover relationships between gut health and diabetes, potentially leading to new preventive strategies and personalized treatment options for patients at risk.
-
18. Predictor profiles of OUD and overdose
The Predictor Profiles of Opioid Use Disorder (OUD) and Overdose project utilizes machine learning models to assess the interactions between established and emerging risk factors associated with OUD and overdose among Post-9/11 veterans. By employing various classification-tree modeling techniques, the project aims to develop comprehensive predictor profiles that can inform targeted interventions and support efforts to combat the opioid crisis within this population.
-
19. Provider directory data accuracy and system of record alignment
The Provider Directory Data Accuracy and System of Record Alignment project employs artificial intelligence to enhance identity resolution and linking within healthcare data systems. AI functions as a transactor, facilitating intelligent state reconstruction over time and enabling real-time discrepancy detection. Additionally, it synchronizes data through intelligent propagation and semi-automated resolution of discrepancies. The use of AI adapters for inference via OWL and logic programming further enhances data integration. The system also features long-term storage capabilities, akin to a “black box flight recorder,” supporting extensive machine learning and business intelligence applications.
-
20. Seizure detection from EEG and video
The Seizure Detection from EEG and Video project utilizes machine learning algorithms to analyze electroencephalogram (EEG) and video data collected from a Veterans Health Administration (VHA) epilepsy monitoring unit. This system is designed to automatically identify seizure events without the need for human intervention, enhancing the efficiency and accuracy of seizure monitoring and diagnosis in patients with epilepsy.
-
21. SoKat Suicidal Ideation Detection Engine
The SoKat Suicidal Ideation Detection Engine (SSIE) employs natural language processing (NLP) techniques to enhance the identification of suicidal ideation among veterans. By analyzing survey data collected by the Office of Mental Health (OMH) through the Veteran Crisis Line (VCL) support team, the SSIE aims to improve the accuracy and timeliness of detecting veterans at risk of suicide, ultimately supporting better mental health interventions.
-
22. Using machine learning to predict perfusionists’ critical decision-making during cardiac surgery
This project focuses on utilizing machine learning to develop predictive models that analyze the decision-making processes of perfusionists during critical moments in the cardiopulmonary bypass phase of cardiac surgery. The insights gained from this analysis may inform the creation of computerized clinical decision support tools that can be integrated into the operating room, enhancing patient safety and improving surgical outcomes through better-informed decision-making.
-
23. Gait signatures in patients with peripheral artery disease
The Gait Signatures in Patients with Peripheral Artery Disease (PAD) project employs machine learning techniques to enhance the treatment of functional issues associated with PAD. By analyzing previously collected biomechanics data, the project aims to identify characteristic gait signatures of PAD patients. Additionally, it investigates the effectiveness of limb acceleration measurements in modeling significant biomechanical metrics related to PAD, ultimately improving patient assessment and treatment strategies.
-
24. Medication Safety (MedSafe) Clinical Decision Support (CDS)
The Medication Safety (MedSafe) Clinical Decision Support (CDS) system leverages electronic clinical data from the VA to analyze the management of conditions such as diabetes, hypertension, and chronic kidney disease. By providing patient-specific, evidence-based recommendations to primary care providers, the system enhances clinical decision-making. It utilizes knowledge bases that encode clinical practice guidelines and employs an automated execution engine to assess various factors, including comorbidities, laboratory results, medications, and past adverse drug events, to generate tailored recommendations for patient care.
-
25. Prediction of health outcomes, including suicide death, opioid overdose, and decompensated outcomes of chronic diseases
The Prediction of Health Outcomes project utilizes electronic health records (EHR), incorporating both structured and unstructured data, to generate deep phenotypes and predict various health outcomes. These outcomes include suicide death, opioid overdose, and decompensated conditions related to chronic diseases. By analyzing comprehensive patient data, this tool aims to enhance risk assessment and inform preventive strategies, ultimately improving patient care and outcomes within the veteran population.
-
26. VA-DoE Suicide Exemplar Project
The VA-DoE Suicide Exemplar Project leverages artificial intelligence to enhance the Veterans Affairs (VA) capacity to identify veterans at risk for suicide. This initiative encompasses three interconnected projects that involve collaboration with the Department of Energy, aiming to develop more effective predictive models and interventions for veteran suicide prevention.
-
27. Machine learning models to predict disease progression among veterans with hepatitis C virus
This project employs machine learning models to forecast the progression of hepatitis C virus (HCV) infection among veterans. By analyzing relevant patient data, the model aims to provide insights into disease trajectories, enabling timely interventions and improved management of HCV among the veteran population.
-
28. Prediction of biologic response to thiopurines
The Prediction of Biologic Response to Thiopurines project utilizes artificial intelligence to analyze data from the Computerized Patient Record System (CPRS) and Clinical Data Warehouse (CDW) to predict how veterans with irritable bowel disease (IBD) will respond to thiopurine medications. This predictive capability aims to optimize treatment plans and improve patient outcomes by tailoring therapies to individual responses.
-
29. Predicting hospitalization and corticosteroid use as a surrogate for IBD flares
The project focuses on predicting hospitalizations and corticosteroid use as indicators of irritable bowel disease (IBD) flare-ups among 20,368 patients diagnosed with IBD in the Veterans Health Administration (VHA) between 2002 and 2009. By utilizing longitudinal laboratory data and various predictors, random forest models are developed to enhance the understanding of IBD management and anticipate exacerbations, ultimately improving patient care.
-
30. Predicting corticosteroid-free endoscopic remission with Vedolizumab in ulcerative colitis
The project aims to predict the likelihood of achieving corticosteroid-free endoscopic remission in patients with ulcerative colitis receiving Vedolizumab treatment. Utilizing random forest modeling, the study analyzes data from 594 patients to forecast outcomes at week 52, based on baseline characteristics and data collected during the first six weeks of therapy. This predictive modeling seeks to inform treatment decisions and improve patient outcomes in ulcerative colitis management.
-
31. Use of machine learning to predict surgery in Crohn’s disease
This project employs machine learning techniques to analyze data from adult patients in the Veterans Integrated Service Networks (VISN) 10 cohort, focusing on demographics, medication usage, and longitudinal laboratory values collected between 2001 and 2015. The goal is to predict the likelihood of surgery for patients with Crohn’s disease and model potential surgical outcomes within one year, thereby enhancing clinical decision-making and patient management strategies.
-
32. Reinforcement learning evaluation of treatment policies for patients with hepatitis C virus
This project utilizes reinforcement learning techniques to evaluate treatment policies for veterans with hepatitis C virus (HCV). By predicting disease progression, the model aims to inform and optimize treatment strategies, ultimately improving health outcomes for veterans affected by HCV.
-
33. Predicting hepatocellular carcinoma in patients with hepatitis C
The Predicting Hepatocellular Carcinoma project focuses on assessing the risk of developing hepatocellular carcinoma (HCC) in patients with hepatitis C virus (HCV)-related cirrhosis within the Veterans Health Administration. By analyzing data from patients with at least three years of follow-up, the study compares the performance of deep learning recurrent neural network (RNN) models, which utilize raw longitudinal data from electronic health records, against traditional regression models. The goal is to enhance predictive accuracy for HCC risk, facilitating early intervention and improved patient outcomes.
-
34. Computer-aided detection and classification of colorectal polyps
The Computer-Aided Detection and Classification of Colorectal Polyps project explores the application of artificial intelligence models to enhance the clinical management of colorectal polyps. The AI models analyze video frames from colonoscopy procedures in real time, focusing on two key tasks: detecting the presence of polyps in the frames and predicting their potential malignancy. This technology aims to improve diagnostic accuracy and support timely interventions for patients undergoing colonoscopy.
-
35. GI Genius (Medtronic)
The Medtronic GI Genius is an advanced artificial intelligence system designed to assist in the detection of colon polyps during gastrointestinal examinations. By leveraging AI technology, the GI Genius enhances the accuracy and efficiency of polyp detection, ultimately supporting healthcare providers in delivering better patient care during colonoscopy procedures.
-
36. Extraction of family medical history from patient records
The Extraction of Family Medical History from Patient Records project is a pilot initiative that utilizes TIU (Text Integration Utility) documentation to gather family medical history data specifically from African American veterans aged 45 to 50. The primary goal is to identify veterans who may be at increased risk for prostate cancer and have not yet undergone screening for the disease. This project aims to enhance early detection and preventive care for prostate cancer within this demographic.
-
37. VA /IRB approved research study for finding colon polyps
The VA/IRB Approved Research Study for Finding Colon Polyps is an initiative that employs a randomized trial design to investigate the effectiveness of artificial intelligence in detecting colon polyps. This study, approved by the Institutional Review Board (IRB), aims to enhance the accuracy and efficiency of polyp detection during colonoscopy procedures, ultimately improving patient outcomes in colorectal cancer prevention.
-
38. Interpretation/triage of eye images
The Interpretation/Triage of Eye Images project leverages artificial intelligence to assist in the triage of eye patients receiving care through telehealth services. The AI system interprets retinal images and evaluates health risks associated with various eye conditions, including glaucoma, macular degeneration, and diabetic retinopathy. The primary objective is to enhance diagnostic accuracy and facilitate timely interventions for patients with eye health concerns.
-
39. Screening for esophageal adenocarcinoma
The Screening for Esophageal Adenocarcinoma project utilizes national Veterans Health Administration (VHA) administrative data to refine and adapt predictive tools that analyze electronic health records. The aim is to assess the risk of developing esophageal adenocarcinoma among veterans, thereby enabling early detection and intervention strategies for this serious condition.
-
40. Social determinants of health extractor
The Social Determinants of Health Extractor project employs artificial intelligence to analyze clinical notes and extract information related to social determinants of health (SDOH). By identifying these variables, the project aims to facilitate health-related analyses that explore the impact of SDOH on disease risk and healthcare disparities. This initiative seeks to enhance understanding of how social factors contribute to health outcomes among veterans, ultimately informing targeted interventions and policy decisions.