Leveraging Large Language Models (LLMs) and OCR for Real-Time Nutritional Risk Assessment in Specialized Diets

Apoorva Pandey; Anusha Marada

Leveraging Large Language Models (LLMs) and OCR for Real-Time Nutritional Risk Assessment in Specialized Diets

Author(s)	Ms. Apoorva Pandey, Prof. Anusha Marada
Country	India
Abstract	When it comes to managing long-term health issue like diabetes or hypertension, people always check what are they consuming, and food labels make it worse. During the course of this research, it became clear that how confusing packaged food labels actually are. Tiny fonts. Reflective plastic. Ingredient names that are rarely recognizable to the customer. It becomes very difficult for patients as they are already managing their complex medical schedules, understanding packaged food label in a market is not feasible and is unreasonable. This gap is addressed by the present study. Existing apps fall short. Most rely on static food databases that miss the regional Indian products entirely, or they demand manual entry that leads to its own individual errors. They also treat every user the same. Which is the real problem when someone has both diabetes and hypertension simultaneously, because the risks of certain ingredients compound in ways no simple calorie tracker accounts for. The proposed system takes a different approach. EasyOCR extracts text directly from food packaging images, which is then passed to Gemini 1.5 Pro for contextual medical reasoning. Rather than simply listing ingredients, the model interprets them against the user’s individual health profile. Testing was done on eight packaged food products taken from the Indian consumer market – including instant noodles, savoury snacks, and processed desserts – reflecting real supermarket conditions rather than controlled laboratory samples. A noteworthy observation was recorded during testing, the model demonstrated the ability to self-correct character level OCR errors prior to analysis. For example, mapping the jumbled output “Soclum” to “Sodium” without explicit instruction. This proved the integrity of medical reasoning. The result showed that the system successfully identifies compounding risks that the other tools missed entirely– like how sodium and refined carbohydrates together create a far more serious vascular threat for someone with both hypertension and diabetes than either ingredient does alone. This research demonstrates that the integration of LLMs into food safety systems can significantly enhance dietary protection and personalized healthcare systems.
Keywords	Optical Character Recognition (OCR), Large Language Models (LLMs), Gemini 1.5 Pro, Specialized Diets, Comorbidity Reasoning
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 17, Issue 2, April-June 2026
Published On	2026-04-08

View / Download PDF File

doi

CrossRef DOI is assigned to each research paper published in our journal.

IJSAT DOI prefix is
10.71097/IJSAT

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 17 Isu 1 Cover Page Vol 16 Isu 3 Cover Page Vol 16 Isu 2

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About IJSAT Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us	Message on WhatsApp	+91-9687-182-185	editor@ijsat.org

International Journal on Science and Technology

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Leveraging Large Language Models (LLMs) and OCR for Real-Time Nutritional Risk Assessment in Specialized Diets

Share this