Ved Thakker

DigitalTwinforAlzheimer’sDiseaseProgressionandPrediction

Made in: 11/2025Purpose: Mini Project - SEM V

A multimodal digital twin system designed to model and predict Alzheimer’s disease progression by combining clinical data, MRI imaging, and biomarker–genotype risk analysis. The platform integrates multiple machine learning models into a unified pipeline for longitudinal disease monitoring and real-time inference.

Tech Stack

Python
Python
FastAPI
FastAPI
Next.js
Next.js
React.js
React.js
PostgreSQL
PostgreSQL

ADDITIONAL TOOLS

CatBoost
XGBoost
Random Forest
MLP
ResNet34
Optuna
SMOTE
PCA
ADNI Dataset

Features

  • 1

    Multimodal Diagnostic Pipeline

    Combines a CatBoost clinical classifier, a ResNet34 MRI model, and a heterogeneous ensemble to capture complementary disease signals.

  • 2

    Digital Twin Modeling

    Constructs a patient-specific digital twin to simulate and track longitudinal Alzheimer’s disease progression.

  • 3

    Ensemble Learning Optimization

    Trains XGBoost, Random Forest, and MLP models on PCA-reduced features and optimizes ensemble performance using Optuna.

  • 4

    Imbalanced Data Handling

    Applies SMOTE to address class sparsity and imbalance in the ADNI dataset.

  • 5

    Real-Time Inference Backend

    Implements a FastAPI backend to serve trained models for low-latency predictions.

  • 6

    Interactive Visualization Frontend

    Provides a Next.js-based interface for data ingestion, result visualization, and model interaction.