RatioLogo
Back

The OBESEYE System: AI for Precision Nutrition

The standard "one-size-fits-all" diet can be dangerously unhelpful for patients in regions like Northern Dhaka, who often manage overlapping chronic conditions. To bridge this gap, researchers are developing OBESEYE, a system leveraging machine learning and Explainable AI (XAI) to create a "Precision Nutrition" model tailored to patients in developing regions.

The Problem & The Purpose

The Challenge of Multimorbidity

A patient isn't just a "diabetic"; they are an individual managing Diabetes Mellitus alongside Chronic Kidney Disease and other illnesses. Most health apps use Western data, failing to account for regional diets or the prevalence of multiple concurrent diseases (multimorbidity).

OBESEYE's Goal

By inputting a patient's demographic data and clinical markers (like serum creatinine and blood glucose), OBESEYE generates a nutrient-specific prescription. It aims to act as a digital nutritionist where human specialists are scarce, moving away from "black box" algorithms toward transparent, actionable advice.

How It Works: Algorithms & Accuracy

The study analyzed 146 NCD patients aged 18–95 years, testing six algorithms to find the best predictors for nutritional needs.

Prediction Performance by Nutrient

  • Carbohydrates: Predicted via Random Forest with 86.99% accuracy.
  • Fluids: Calculated most stably through Linear Regression, yielding an RMSE of 0.39 L.
  • Fats: Prediction accuracy was 64.96%, with an RMSE of 15.09g. Nutritionists deemed this "tolerable" within clinical margins, but it highlights the complex challenge of mapping every metabolic variable.

The "Explainable" in Explainable AI

Transparency is core to this innovation. Using LIME and SHAP values, the system provides clinical justifications for its recommendations. For example, if it suggests cutting protein, it will specifically point to the patient's elevated serum creatinine levels indicating Chronic Kidney Disease. This allows doctors to trust the machine's logic, not just follow a blind command.

Current Limitations & Future Steps

Recognized Hurdles

While promising, the system has limitations:

  • The cohort of 146 patients is relatively small.
  • Data is localized to a single hospital in Dhaka; the model may not perfectly translate to non-Oriental diets.
  • The system tracks macronutrients (carbs, fats, protein) but does not yet account for essential micronutrients like vitamins and minerals.

Conclusion: Despite these challenges, OBESEYE represents a significant step toward a future where AI handles the complex math of personalized medicine, allowing clinicians to focus more on the patient.

Reference: Roy, M., Das, S., & Protity, A. T. (2023). OBESEYE: Interpretable Diet Recommender for Obesity Management using Machine Learning and Explainable AI. International Journal of Recent Advances in Multidisciplinary Topics, Volume 4, Issue 6.