A machine learning-based web application that classifies cognitive status (CN, EMCI, LMCI, MCI) using only minimal MRI metadata — specifically age, biological gender and scan description — without requiring full image data.
This project leverages structured metadata from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to build a lightweight, deployable screening tool for early cognitive impairment detection. The app is built with Streamlit, making it easy for clinicians or researchers to input basic patient info and receive real-time predictions using a trained Gradient Boosting Classifier.
Classifies cognitive state into:
-
CN (Cognitively Normal)
-
EMCI (Early Mild Cognitive Impairment)
-
LMCI (Late Mild Cognitive Impairment)
-
MCI (General MCI)
-
Trained on real clinical metadata (from ADNI-derived datasets)
-
Includes preprocessing, model training, evaluation and deployment
-
Web-based interface via Streamlit
-
Model performance evaluated using accuracy, precision, recall, and F1-score
Current ML tools for Alzheimer’s rely on complex neuroimaging. This project demonstrates that even low-dimensional features can provide meaningful classification, enabling scalable triage tools for use in resource-limited settings.
To run locally:
pip install -r requirements.txt
streamlit run streamlit_app.py
-
app.py: Model training and evaluation script
-
streamlit_app.py: Streamlit interface
-
model.pkl, model_columns.pkl: Trained model and expected features
-
Data: Contains the CSV files for CN, EMCI, LMCI and MCI classes
This project used Alzehimer dataset from Kaggle
Link: https://www.kaggle.com/datasets/dilipharish/alzehimercsvdatas