Cross-Case Fraud Indicator Correlation

INSPIRATION : The idea began because of a frustration : Liver disease is one of the most common health conditions and it has almost no warning system. MASLD, which means fatty liver disease, stays quiet until it is already advanced. By the time it appears on a blood test or scan, damage has usually been accumulating for years.

Wearables already provide monitoring for things such as heart rate and sleep. However, nobody has linked that everyday tracking to liver‑health risk. Moreover, no one has explained why that risk is changing instead of merely giving a single number.

When we discovered Eco Circuit's brief, it raised an equally important question : why must health monitoring create more electronic waste? Continuous glucose monitors and cardiac patches are discarded every one to two weeks. We wanted to create a device that solves both problems together: a wearable that gives you information about a silent disease and a device that does not leave electronic waste each time it is removed.

WHAT IT DOES :

LiverShield Patch is a two‑part system : ~ A biodegradable adhesive patch made from cellulose and PLA with printed bio‑ink electrodes. It stays on the skin for days, collects PPG signals that relate to heart rate and motion signals, and then is peeled off and composted. ~ A small reusable electronics module built around an ESP32. It snaps onto each patch and travels with the user from one wear cycle to the next. Thus only the sensing layer is discarded.

Sensor data from every cycle, with a few lifestyle inputs such as sleep, activity and diet, goes into an AI layer. This layer examines trends across cycles instead of single‑day snapshots. It produces a risk trajectory that can be stable, improving or increasing. It also explains the behavioral factors that drive that trend and suggests a clear non‑prescriptive lifestyle action. This device is not a diagnostic tool. It is an early‑warning layer that comes before clinical care and does not replace it.

HOW WE BUILT IT :

We split the system into three layers and built them somewhat in parallel

Sensing hardware : We used a MAX30102 for PPG and an MPU6050 for IMU, both connected to an ESP32. We first made a wearable to test the sensing pipeline before moving to the biodegradable substrate.

Biodegradable substrate research : We based our material choices on existing bioelectronics studies instead of guessing. EKG and temperature sensors that use graphene and PEDOT:PSS electrodes have shown real‑world feasibility and much lower environmental impact than normal electrodes. Compostable printed circuits on paper or bioplastic substrates have remained stable for more than a year before degrading. Those findings gave us a base for designing the disposable patch electrodes.

AI risk‑trend engine : We began with simple models such as logistic regression and random forests using a rolling window of readings. We avoided a black‑box model because explainability was a core requirement, not an afterthought.

Intervention layer : We built this part to answer the question "Why is my risk changing?" with human‑readable factors. It then links those factors to a lifestyle suggestion.

CHALLENGES WE RAN INTO :

Reconciling monitoring with disposable hardware : Most wearable health ideas assume the device stays on the body forever. We redesigned the system around discrete wear cycles. We proved that a risk trajectory can still be meaningful when it is assembled from short segments instead of one continuous stream.

Biodegradable materials do not automatically mean the device is safe. We discovered that some transient electronics can leave non‑degradable byproducts even when the main material dissolves as expected. This changed our validation plan. We now treat the question of "what does it degrade into?" as a priority, not an afterthought.

Avoiding overclaiming about benefits : It would have been simple to present this as "AI that detects liver disease." We chose to limit the claim to risk‑trend detection. No wearable signal directly measures liver pathology. This is a more honest story to tell.

Scope discipline under hackathon time limits : Building firmware, a biodegradable sensing layer, an AI engine and two dashboards is a lot to do in a short time. We had to decide what to build for real — the sensing pipeline — and what to show through simulation — the multi‑cycle risk trajectory.

WHAT WE LEARNED :

The interesting engineering problem was not the AI but the material science. Making a sensor that works well and has low impact when it is discarded is a much harder challenge than either goal alone. This shifted our thinking about the system. Reusability is not a sustainability checkbox; it is an architectural choice that must be made before designing anything else.