Effective Date: May 2026 · PancreaTrack MedTech LLC
PancreaTrack uses artificial intelligence in several specific, bounded ways to help patients organize their health data, understand their meals, and prepare for clinical appointments. "AI" within this platform refers to machine learning models — specifically large language models (LLMs) and vision models — that process text or images and generate structured outputs.
AI features in PancreaTrack are designed to augment patient self-management and physician awareness — not to replace clinical judgment, diagnosis, or treatment decision-making. Every AI output the platform produces is labeled as AI-generated and accompanied by a reminder that it is not a substitute for professional medical advice.
PancreaTrack currently uses the following AI model infrastructure:
| Provider | Model(s) Used | Purpose |
|---|---|---|
| Google Gemini | Gemini 2.5 Flash & Gemini 2.5 Pro | Meal photo analysis, dietary trigger identification, appointment summaries, physician patient briefs, document Q&A |
| USDA FoodData Central | API (database lookup, not generative AI) | Nutritional data retrieval — fat, carbohydrate, protein, and calorie content for logged foods |
Google Gemini is a product of Google LLC. Use of the Gemini API is governed by Google's Generative AI Terms of Service. PancreaTrack does not use your health data to train AI models. Data submitted to Gemini is processed in accordance with Google's API data usage policies and is not retained by Google for training purposes under the API terms.
The following features within PancreaTrack use AI model inference:
Patients upload a photo of a meal. Gemini 2.5 Flash analyzes the image using computer vision, identifies likely food items, estimates macronutrients (fat, carbohydrates, protein, calories), and generates a pancreatitis-specific fat risk score. Nutritional estimates are approximations based on visual analysis and are not laboratory-verified values. Accuracy varies with image quality, portion visibility, and food preparation methods.
Gemini reviews up to 90 days of a patient's meal logs and pain scores to identify foods that statistically precede symptom flares. Results are presented as observations based on logged data — not clinical diagnoses of food intolerance or allergy. The analysis reflects patterns in what was logged and may not reflect all relevant dietary or non-dietary factors.
Gemini 2.5 Pro generates a structured narrative summarizing a patient's recent health data — pain trends, dietary patterns, severe episodes, and enzyme logs — in a format intended for sharing with a healthcare provider at an appointment. This summary is not a clinical assessment. It is an organizational tool that presents the patient's logged data in a readable format. The patient is responsible for verifying its accuracy before sharing.
Available in the physician portal, this feature generates a structured clinical brief from a patient's logged data to orient the physician before a review or visit. This brief is not a clinical record and does not replace physician chart review, clinical assessment, or professional judgment. It is an AI-generated summary of patient-reported data and carries the same limitations as all patient-reported outcomes: it reflects what was logged, not necessarily the complete clinical picture.
Patients upload medical documents (lab results, discharge summaries, imaging reports). Gemini 2.5 Pro reads the document and provides a plain-language summary, highlights abnormal values, defines medical terms, and suggests questions for the patient to raise with their physician. This feature is an educational translation tool only. It does not provide clinical interpretation, diagnosis, or treatment guidance. Patients are explicitly directed to discuss all findings with their healthcare provider.
When you use an AI-powered feature, PancreaTrack transmits a structured prompt to the Gemini API containing relevant portions of your logged health data. The following table describes what is included for each feature:
| Feature | Data Transmitted to Gemini |
|---|---|
| Meal Photo Analysis | The uploaded meal photograph (image data only — no personal identifiers) |
| Food Trigger Analysis | De-identified meal logs and pain scores from the past 90 days (food names, fat grams, pain scores, timestamps) |
| Pre-Appointment Summary | Pain statistics, meal summaries, severe episode counts, enzyme logs, and food tolerance data for the selected date range — no name, email, or contact information |
| Physician AI Patient Brief | Pain and meal statistics, food tolerance, care notes summary, and current medications — no government-issued ID or financial data |
| Document Q&A | The full content of the uploaded document (PDF, image, or scan) |
We do not transmit your name, email address, password, payment information, or government-issued ID to AI models. Prompts are structured to include only the health data necessary to generate the requested output.
PancreaTrack AI is a health information and self-management platform. It is not cleared or approved by the U.S. Food and Drug Administration (FDA) as a medical device. The AI features within PancreaTrack are designed to help patients organize, understand, and communicate their own health data — they are not intended to diagnose disease, recommend treatment, prescribe medication, or replace the clinical judgment of a licensed healthcare provider.
The platform's AI outputs are decision-support tools for patients and clinicians to consider alongside — not instead of — standard clinical assessment. Any clinical action taken based on PancreaTrack data remains the responsibility of the treating physician and the patient.
All physician access to patient data in PancreaTrack is governed by a two-layer identity verification process: NPI verification against the CMS NPPES National Provider Registry and government-issued photo ID verification via Stripe Identity. No AI feature generates or modifies patient clinical records — all data in the system is either patient-entered or explicitly labeled as AI-generated content.
Physicians using the Physician AI Patient Brief are reminded within the interface that the brief is an AI-generated summary of patient-reported data and is not a substitute for full chart review, direct patient assessment, or clinical judgment. The platform is designed on the principle that AI should flag genuine concerns based on available data — and should explicitly acknowledge when data is insufficient rather than generating confident outputs from sparse inputs.
If you have questions about how AI is used in PancreaTrack, concerns about a specific AI output, or requests related to your data, please contact us:
This disclosure is provided for transparency and patient understanding. It supplements but does not replace the Privacy Policy and Terms of Service, which govern data use and platform access. PancreaTrack reserves the right to update this disclosure as AI features evolve. Material changes will be reflected in the effective date above.