50 AI in Medicine Project Ideas for High School Students

At the high school level, a good AI + medicine project teaches you something a biology class never can: how a model actually performs against real, messy clinical data, not just a clean textbook example. Most students start by trying to build a general disease predictor, and that is fine as a first exercise, but the projects that actually build skill, and the ones that stand out in a portfolio or research paper, ask a narrower, more specific question using real medical data. We compiled this list of 50 AI in medicine project ideas for high school students, organized by domain, from medical imaging to genomics to clinical natural language processing.

If you want structured guidance, turning any of these into a polished, defensible research project rather than a tutorial clone, a mentored program like Veritas AI pairs you directly with an AI researcher to build something original. 

What Makes an AI + Medicine Project Worth Doing at the High School Level?

Before the list, a quick framing question that will save you a lot of wasted effort: are you trying to answer something you don't already know, or are you trying to demonstrate a technique? Both are valid reasons to do a project, but they should be approached differently.

If you are learning a technique, pick a clean, well-documented medical dataset and focus on getting the mechanics right. If you are trying to answer a real question, like whether a specific imaging feature actually predicts an outcome, expect the data to be messier, expect your first model to need revision, and expect the most interesting part of the project to be explaining what you found and where your model's limitations are, since medical claims deserve real caution.

The second kind of project is what makes a portfolio, research paper, or competition submission memorable, as long as you are always honest that it is a research exercise, not a diagnostic tool.

Medical Imaging and Diagnosis Projects

These projects use image classification to detect patterns in medical scans, and they are the most common entry point into AI + medicine, so a genuinely original angle matters more here than elsewhere on this list.

1. Building a model to detect pneumonia from chest X-ray images using a public medical imaging dataset

2. Classifying skin lesion images as benign or potentially malignant using a dermatology image dataset

3. Detecting diabetic retinopathy from retinal scan images and analyzing where the model is least confident

4. Building a model to identify bone fractures in X-ray images and comparing its errors to common radiologist mistakes

5. Segmenting tumor regions in MRI scans and evaluating segmentation accuracy against expert-labeled data

Disease Prediction and Risk Modeling Projects

1. Predicting diabetes risk from clinical and lifestyle features using a public health dataset

2. Building a model to predict heart disease risk from patient vitals and history, and analyzing which features matter most

3. Predicting hospital readmission risk within 30 days using de-identified patient intake data

4. Investigating whether a stroke risk model shows different accuracy across age or demographic groups

5. Building a model to predict sepsis onset hours before clinical diagnosis using vital sign trends

Clinical NLP and Electronic Health Records Projects

1. Extracting medication names and dosages from unstructured clinical notes using named entity recognition

2. Building a model that classifies clinical notes by likely diagnosis category

3. Summarizing lengthy discharge notes into a short, plain-language summary of key information

4. Detecting mentions of adverse drug reactions in unstructured patient forum text

5. Building a model that flags clinical notes with ambiguous or contradictory information for human review

Drug Discovery and Chemistry Projects

1. Predicting a candidate molecule's binding affinity to a target protein using public chemistry datasets

2. Building a model to predict drug side effects from molecular structure using public pharmacology data

3. Classifying molecules by toxicity risk using public chemical structure and toxicity datasets

4. Predicting drug-drug interaction risk from known pharmacological features

5. Building a model to identify structurally similar compounds to a known effective drug

Genomics and Personalized Medicine Projects

1. Building a model to classify gene expression patterns associated with a specific cancer subtype

2. Predicting disease risk from genetic variant data using a public genomics dataset

3. Investigating which genetic markers a model relies on most when predicting a specific condition

4. Building a model to cluster patients into subgroups based on genetic and clinical profile similarity

5. Studying how well a genomics-based prediction model generalizes across different population groups represented in the data

Epidemiology and Public Health Modeling Projects

1. Modeling and forecasting disease spread through a population using an SIR-style epidemic model

2. Analyzing vaccination rate trends and their correlation with disease incidence at the county level

3. Building a model to predict flu season severity using historical public health surveillance data

4. Investigating the relationship between air quality data and asthma-related hospital visits

5. Analyzing how a public health intervention's timing affected case counts using historical outbreak data

Wearables and Remote Patient Monitoring Projects

1. Building a model to detect irregular heart rhythms from wearable device heart rate data

2. Classifying physical activity type from smartphone or wearable accelerometer data

3. Detecting early signs of illness from resting heart rate and sleep pattern changes captured by a wearable

4. Building a fall detection model using accelerometer data from a wearable or phone sensor

5. Analyzing heart rate variability data to identify patterns associated with stress or recovery

Mental Health and Behavioral Prediction Projects

1. Building a sentiment classifier for anonymized social media posts as a proxy for detecting language associated with distress

2. Analyzing survey data to identify the strongest predictors of reported adolescent anxiety

3. Building a model that detects linguistic markers of depression in de-identified text, with careful attention to its real limitations

4. Investigating how sleep pattern data correlates with self-reported mood in a public dataset

5. Studying how screen time patterns relate to self-reported wellbeing using survey data

Medical AI Ethics, Bias, and Interpretability Projects

1. Auditing a public diagnostic model for accuracy differences across demographic groups

2. Investigating which features a medical prediction model relies on most, using an interpretability technique like SHAP

3. Studying how removing a demographic feature, like race or zip code, changes a model's accuracy and its fairness

4. Building a simple tool that explains a medical model's prediction in plain language for a non-technical reader

5. Comparing a black-box model's accuracy against a more interpretable model on the same medical prediction task

Hospital Operations and Healthcare Systems Projects

1. Building a model to predict emergency room wait times based on historical patient volume data

2. Forecasting hospital bed demand using historical admission and discharge data

3. Analyzing appointment no-show patterns and building a model to predict which appointments are at risk

4. Building a model to optimize nurse or staff scheduling based on predicted patient volume

5. Studying how a hospital's patient flow changed during a historical surge event using operational data

How Should You Choose the Right Project?

With this many options, the hardest part is often deciding where to start. A few questions worth asking yourself:

1. Does this project use a skill I haven't tried yet, or does it deepen a skill I'm still shaky on?

2. Is there a real, accessible medical dataset available for this, and is it clean enough, or appropriately messy, for what I want to do?

3. Can I state, in one sentence, the specific question I'm trying to answer? If you cannot answer that last question clearly, the project needs more thinking before you open a notebook. The strongest AI + medicine work, at any level, starts with a sharp, specific question and stays honest about being a research exercise rather than a clinical tool.

How Can You Get Feedback on Your Work?

The single biggest difference between a project that looks like a tutorial exercise and one that looks like real research is feedback from someone who actually knows what they are looking at. A mentor can tell you when your model's high accuracy is actually a sign of data leakage, when your dataset is too small to support the claim you are making, or when an interesting result needs one more robustness check before you can trust it.

While a mentor can be a teacher, an online community, or an older student further along than you, a structured program built around mentorship is usually the most reliable way to get that feedback consistently rather than occasionally.

Veritas AI offers two paths depending on where you are. The Deep Dive: AI + Medicine program is a project-based course covering AI applications in healthcare, including disease diagnosis and medical imaging, with mentorship from researchers experienced in the field. If you are ready to take a specific idea from this list further, the AI Fellowship pairs you one-on-one with a mentor, often a PhD or PhD candidate from a school like Harvard, Stanford, or Oxford, for twelve to fifteen weeks of original research, with support toward publication if the work merits it.If one of the ideas on this list is the one you keep coming back to, that is usually a sign it is worth doing properly, with someone who can help you do it right.

Frequently Asked Questions About AI in Medicine Project Ideas

1. Do I need a background in biology or medicine to start one of these projects?

No, though it helps to build as you go. Most of these projects are approachable with a working knowledge of Python and basic statistics, and medical concepts can be learned alongside the project as you need them, rather than as a prerequisite.

2. Where can I find real medical data for these projects?

Public sources like Kaggle host cleaned medical imaging and clinical datasets, PhysioNet provides de-identified clinical and physiological data specifically for research, and the CDC and NIH both publish public health and genomics datasets that are free to use.

3. What is the biggest mistake students make with AI medicine projects?

Presenting a model as diagnostically reliable without acknowledging its real limitations. Medical claims carry real weight, and a project that is upfront about small sample sizes, dataset bias, or the gap between a research model and an actual clinical tool reads as far more credible than one that overstates what it found.

4. How long should an AI + medicine project take?

A focused classification or exploratory project can be completed in one to two weeks. Projects involving model comparison or interpretability analysis typically take two to four weeks. The most ambitious original research projects, suitable for a journal submission, often take two to three months, including iteration and write-up.

5. Can a high school AI + medicine project be published?

Yes. Journals like the Journal of Emerging Investigators accept rigorous research from high school students, and a project with a clear question, honest methodology, and a defensible conclusion has a real shot, particularly with mentorship from someone who can help you meet the bar for publication.