Cardiologist's Guide to the Heart

Track: Education

Why I built it

During Build Week, my cousin brother, who I spent much of my childhood with, was recovering from a pneumothorax. For the first time, understanding human anatomy stopped feeling optional or like something to memorise. I wanted to understand how the parts affect one another.

Medicine is often taught as isolated names and diagrams. The heart is really a chain of causes and effects: electrical timing changes contraction, contraction changes pressure, and pressure moves blood. Cardiologist's Guide to the Heart turns that chain into a spatial lesson learners can inspect.

What it does

Learners stand beside a beating 3D heart and move through a guided sequence covering:

• spatial orientation and cardiac anatomy;
• blood flow through the chambers and major vessels;
• electrical conduction and heartbeat timing;
• healthy coronary circulation;
• reversible coronary artery narrowing;
• the resulting changes in flow, pressure, and workload;
• restoration of healthy flow and a lesson recap.

The learner can rotate, zoom, pause, replay, isolate structures, and compare healthy and narrowed coronary states without losing the surrounding anatomy.

The simulation is deterministic. The lesson state machine, physiological transitions, visible measurements, and reversible stenosis sequence are computed in Swift. The AI tutor explains the state; it does not control or invent it.

Try it

Watch the demo: https://youtu.be/DFMHMAHIxrs

Source and setup: https://github.com/Arnie016/heart-field-vision

This is a native visionOS project, not a browser demo. The fastest hands-on path is the Apple Vision Pro Simulator on a Mac with Xcode and the visionOS simulator runtime installed.

git clone https://github.com/Arnie016/heart-field-vision.git
cd heart-field-vision
brew install xcodegen
xcodegen generate
open HeartFieldVision.xcodeproj

In Xcode:

1. Select the HeartFieldVision scheme.
2. Choose an Apple Vision Pro simulator.
3. Press Run or Command-R.
4. Begin the guided lesson from the control deck.

The deterministic lesson runs locally. The live tutor requires the user's own OpenAI Platform API key, entered inside the app and stored in Keychain. No API key is included in the repository. The live voice path is explicitly configured for gpt-realtime-2.1.

What judges can try

1. Complete the guided anatomy orientation.
2. Follow blood through the chambers and major vessels.
3. Reveal the electrical conduction sequence and compare it with contraction timing.
4. Activate coronary stenosis and inspect the resulting state changes.
5. Restore healthy flow and watch the causal chain reverse.
6. Connect the Realtime tutor and ask about the structure or state currently visible.

How I used Codex and GPT-5.6

Codex was my engineering partner throughout Build Week. It helped architect the visionOS application, implement the lesson state machine, connect spatial interactions to deterministic simulation states, build the reversible stenosis workflow, diagnose build and simulator issues, and create reproducible verification paths.

GPT-5.6 was used during development to challenge the lesson structure, improve explanations, review safety boundaries, and help turn physiological relationships into a sequence a learner could follow. It is not the numerical engine.

The live tutor uses gpt-realtime-2.1. It receives a bounded summary of the current lesson stage and visible simulation state, then explains a relationship or suggests the next observation. It cannot alter anatomy, modify physiological values, diagnose a patient, or provide treatment advice.

Why it matters

A flat diagram can show where a valve or artery is. It is much harder for that diagram to show why electrical timing, contraction, pressure, and coronary flow belong to one system. This project gives learners a shared, replayable spatial model for building that causal intuition.

Current status

The complete guided lesson and deterministic state transitions run in the Apple Vision Pro simulator. The project has a recorded demo and reproducible source setup. Physical Apple Vision Pro hardware and live microphone behavior remain device-dependent validation steps and are not claimed as complete.

git clone https://github.com/Arnie016/heart-field-vision.git cd heart-field-vision brew install xcodegen # skip if installed xcodegen generate open HeartFieldVision.xcodeproj

Medical disclaimer

This is an educational visualization. It is not medical advice, diagnostic software, clinical decision support, or a substitute for professional care.

Built With

• gpt-5.6
• gpt-realtime-2.1
• openai-codex
• openai-realtime-api
• realitykit
• swift
• swiftui
• visionos
• xcodegen