Office of Research & Development |
![]() ![]() |
The Department of Veterans Affairs (VA) shall, when designing, developing, acquiring, and using artificial intelligence (AI), adhere to the nine principles for use of AI in government as outlined in Executive Order (EO) 13960: Promoting the Use of Trustworthy Artificial Intelligence in Federal Government (Sec. 3). These nine principles describe trustworthy AI as: (1) lawful and respectful of our Nation’s values, (2) purposeful and performance-driven, (3) accurate, reliable, and effective, (4) safe, secure, and resilient, (5) understandable, (6) responsible and traceable, (7) regularly monitored, (8) transparent, and (9) accountable.
In accordance with the directives of this EO, the VA has published the following AI use cases:
|
AI use case name |
Summary |
1 |
Artificial Intelligence physical therapy app |
This app is a physical therapy support tool. It is a data source agnostic tool which takes input from a variety of wearable sensors and then analyzes the data to give feedback to the physical therapist in an explainable format. |
2 |
Artificial intelligence coach in cardiac surgery |
The artificial intelligence coach in cardiac surgery infers misalignment in team members’ mental models during complex healthcare task execution. Of interest are safety-critical domains (e.g., aviation, healthcare), where lack of shared mental models can lead to preventable errors and harm. Identifying model misalignment provides a building block for enabling computer-assisted interventions to improve teamwork and augment human cognition in the operating room. |
3 |
AI Cure |
AICURE is a phone app that monitors adherence to orally prescribed medications during clinical or pharmaceutical sponsor drug studies. |
4 |
Acute kidney injury (AKI) |
This project, a collaboration with Google DeepMind, focuses on detecting acute kidney injury (AKI), ranging from minor loss of kidney function to complete kidney failure. The artificial intelligence can also detect AKI that may be the result of another illness. |
5 |
Assessing lung function in health and disease |
Health professionals can use this artificial intelligence to determine predictors of normal and abnormal lung function and sleep parameters. |
6 |
Automated eye movement analysis and diagnostic prediction of neurological disease |
Artificial intelligence recursively analyzes previously collected data to both improve the quality and accuracy of automated algorithms, as well as to screen for markers of neurological disease (e.g. traumatic brain injury, Parkinson's, stroke, etc). |
7 |
Automatic speech transcription engines to aid scoring neuropsychological tests. |
Automated speech transcription engines analyze the cognitive decline of older VA patients. Digitally recorded speech responses are transcribed using multiple artificial intelligence-based speech-to-text engines. The transcriptions are fused together to reduce or obviate the need for manual transcription of patient speech in order to score the neuropsychological tests. |
8 |
CuraPatient |
CuraPatient is a remote tool that allows patients to better manage their conditions without having to see a provider. Driven by artificial intelligence, it allows patients to create a profile to track their health, enroll in programs, manage insurance, and schedule appointments. |
9 |
Digital command center |
The Digital Command Center seeks to consolidate all data in a medical center and apply predictive prescriptive analytics to allow leaders to better optimize hospital performance. |
10 |
Disentangling dementia patterns using artificial intelligence on brain imaging and electrophysiological data |
This collaborative effort focuses on developing a deep learning framework to predict the various patterns of dementia seen on MRI and EEG and explore the use of these imaging modalities as biomarkers for various dementias and epilepsy disorders. The VA is performing retrospective chart review to achieve this. |
11 |
Machine learning (ML) for enhanced diagnostic error detection and ML classification of protein electrophoresis text |
Researchers are performing chart review to collect true/false positive annotations and construct a vector embedding of patient records, followed by similarity-based retrieval of unlabeled records "near" the labeled ones (semi-supervised approach). The aim is to use machine learning as a filter, after the rules-based retrieval, to improve specificity. Embedding inputs will be selected high-value structured data pertinent to stroke risk and possibly selected prior text notes. |
12 |
Behavidence |
Behavidence is a mental health tracking app. Veterans download the app onto their phone and it compares their phone usage to that of a digital phenotype that represents people with confirmed diagnosis of mental health conditions. |
13 |
Machine learning tools to predict outcomes of hospitalized VA patients |
This is an IRB-approved study which aims to examine machine learning approaches to predict health outcomes of VA patients. It will focus on the prediction of Alzheimer's disease, rehospitalization, and Chlostridioides difficile infection. |
14 |
Nediser reports QA |
Nediser is a continuously trained artificial intelligence “radiology resident” that assists radiologists in confirming the X-ray properties in their radiology reports. Nediser can select normal templates, detect hardware, evaluate patella alignment and leg length and angle discrepancy, and measure Cobb angles. |
15 |
Precision medicine PTSD and suicidality diagnostic and predictive tool |
This model interprets various real time inputs in a diagnostic and predictive capacity in order to forewarn episodes of PTSD and suicidality, support early and accurate diagnosis of the same, and gain a better understanding of the short and long term effects of stress, especially in extreme situations, as it relates to the onset of PTSD. |
16 |
Prediction of Veterans' Suicidal Ideation following Transition from Military Service |
Machine learning is used to identify predictors of veterans' suicidal ideation. The relevant data come from a web-based survey of veterans’ experiences within three months of separation and every six months after for the first three years after leaving military service. |
17 |
PredictMod |
PredictMod uses artificial intelligence to determine if predictions can be made about diabetes based on the gut microbiome. |
18 |
Predictor profiles of OUD and overdose |
Machine learning prediction models evaluate the interactions of known and novel risk factors for opioid use disorder (OUD) and overdose in Post-9/11 Veterans. Several machine learning classification-tree modeling approaches are used to develop predictor profiles of OUD and overdose. |
19 |
Provider directory data accuracy and system of record alignment |
AI is used to add value as a transactor for intelligent identity resolution and linking. AI also has a domain cache function that can be used for both Clinical Decision Support and for intelligent state reconstruction over time and real-time discrepancy detection. As a synchronizer, AI can perform intelligent propagation and semi-automated discrepancy resolution. AI adapters can be used for inference via OWL and logic programming. Lastly, AI has long term storage (“black box flight recorder”) for virtually limitless machine learning and BI applications. |
20 |
Seizure detection from EEG and video |
Machine learning algorithms use EEG and video data from a VHA epilepsy monitoring unit in order to automatically identify seizures without human intervention. |
21 |
SoKat Suicidial Ideation Detection Engine |
The SoKat Suicide Ideation Engine (SSIE) uses natural language processing (NLP) to improve identification of Veteran suicide ideation (SI) from survey data collected by the Office of Mental Health (OMH) Veteran Crisis Line (VCL) support team (VSignals). |
22 |
Using machine learning to predict perfusionists’ critical decision-making during cardiac surgery |
A machine learning approach is used to build predictive models of perfusionists’ decision-making during critical situations that occur in the cardiopulmonary bypass phase of cardiac surgery. Results may inform future development of computerized clinical decision support tools to be embedded into the operating room, improving patient safety and surgical outcomes. |
23 |
Gait signatures in patients with peripheral artery disease |
Machine learning is used to improve treatment of functional problems in patients with peripheral artery disease (PAD). Previously collected biomechanics data is used to identify representative gait signatures of PAD to 1) determine the gait signatures of patients with PAD and 2) the ability of limb acceleration measurements to identify and model the meaningful biomechanics measures from PAD data. |
24 |
Medication Safety (MedSafe) Clinical Decision Support (CDS) |
Using VA electronic clinical data, the Medication Safety (MedSafe) Clinical Decision Support (CDS) system analyzes current clinical management for diabetes, hypertension, and chronic kidney disease, and makes patient-specific, evidence-based recommendations to primary care providers. The system uses knowledge bases that encode clinical practice guideline recommendations and an automated execution engine to examine multiple comorbidities, laboratory test results, medications, and history of adverse drug events in evaluating patient clinical status and generating patient-specific recommendations |
25 |
Prediction of health outcomes, including suicide death, opioid overdose, and decompensated outcomes of chronic diseases. |
Using electronic health records (EHR) (both structured and unstructured data) as inputs, this tool outputs deep phenotypes and predictions of health outcomes including suicide death, opioid overdose, and decompensated outcomes of chronic diseases. |
26 |
VA-DoE Suicide Exemplar Project |
The VA-DoE Suicide Exemplar project is currently utilizing artificial intelligence to improve VA's ability to identify Veterans at risk for suicide through three closely related projects that all involve collaborations with the Department of Energy. |
27 |
Machine learning models to predict disease progression among veterans with hepatitis C virus |
A machine learning model is used to predict disease progression among veterans with hepatitis C virus. |
28 |
Prediction of biologic response to thiopurines |
Using CPRS and CDW data, artificial intelligence is used to predict biologic response to thiopurines among Veterans with irritable bowel disease. |
29 |
Predicting hospitalization and corticosteroid use as a surrogate for IBD flares |
This work examines data from 20,368 Veterans Health Administration (VHA) patients with an irritable bowel disease (IBD) diagnosis between 2002 and 2009. Longitudinal labs and associated predictors were used in random forest models to predict hospitalizations and steroid usage as a surrogate for IBD Flares. |
30 |
Predicting corticosteroid free endoscopic remission with Vedolizumab in ulcerative colitis |
This work uses random forest modeling on a cohort of 594 patients with Vedolizumab to predict the outcome of corticosteroid-free biologic remission at week 52 on the testing cohort. Models were constructed using baseline data or data through week 6 of VDZ therapy. |
31 |
Use of machine learning to predict surgery in Crohn’s disease |
Machine learning analyzes patient demographics, medication use, and longitudinal laboratory values collected between 2001 and 2015 from adult patients in the Veterans Integrated Service Networks (VISN) 10 cohort. The data was used for analysis in prediction of Crohn’s disease and to model future surgical outcomes within 1 year. |
32 |
Reinforcement learning evaluation of treatment policies for patients with hepatitis C virus |
A machine learning model is used to predict disease progression among veterans with hepatitis C virus. |
33 |
Predicting hepatocellular carcinoma in patients with hepatitis C |
This prognostic study used data on patients with hepatitis C virus (HCV)-related cirrhosis in the national Veterans Health Administration who had at least 3 years of follow-up after the diagnosis of cirrhosis. The data was used to examine whether deep learning recurrent neural network (RNN) models that use raw longitudinal data extracted directly from electronic health records outperform conventional regression models in predicting the risk of developing hepatocellular carcinoma (HCC). |
34 |
Computer-aided detection and classification of colorectal polyps |
This study is investigating the use of artificial intelligence models for improving clinical management of colorectal polyps. The models receive video frames from colonoscopy video streams and analyze them in real time in order to (1) detect whether a polyp is in the frame and (2) predict the polyp's malignant potential. |
35 |
GI Genius (Medtronic) |
The Medtronic GI Genius aids in detection of colon polyps through artificial intelligence. |
36 |
Extraction of family medical history from patient records |
This pilot project uses TIU documentation on African American Veterans aged 45-50 to extract family medical history data and identify Veterans who are are at risk of prostate cancer but have not undergone prostate cancer screening. |
37 |
VA /IRB approved research study for finding colon polyps |
This IRB approved research study uses a randomized trial for finding colon polyps with artifical intelligence. |
38 |
Interpretation/triage of eye images |
Artificial intelligence supports triage of eye patients cared for through telehealth, interprets eye images, and assesses health risks based on retina photos. The goal is to improve diagnosis of a variety of conditions, including glaucoma, macular degeneration, and diabetic retinopathy. |
40 |
Screening for esophageal adenocarcinoma |
National VHA administrative data is used to adapt tools that use electronic health records to predict the risk for esophageal adenocarcinoma. |
41 |
Social determinants of health extractor |
AI is used with clinical notes to identify social determinants of health (SDOH) information. The extracted SDOH variables can be used during associated health related analysis to determine, among other factors, whether SDOH can be a contributor to disease risks or healthcare inequality. |
Contact: NAII@va.gov