Alzheimer Detection Paradigm

Inspiration- I chose this project because I wanted to work on something both meaningful and challenging. Alzheimer’s is a condition that affects millions of people, and the idea of using technology to support earlier detection motivated me to explore this problem. Build- I built an end-to-end system that takes raw MRI scans and trains a model to classify different stages of Alzheimer’s. I began by visualizing the MRI data to better understand the images, then cleaned and standardized them by resizing the scans and mapping each label to the correct disease stage. Using this prepared data, I trained a convolutional neural network (CNN) and validated it with 5-fold cross-validation to ensure consistent and reliable performance, saving the best-performing model across all folds. To make the project usable beyond just training a model, I created an interactive tool that allows users to upload an MRI scan and receive a clear stage prediction with a confidence score. Challenges- One challenge was differentiating between mild and very mild Alzheimer’s, as visual differences can be subtle, leading to occasional overlap in predictions. Takeaways- Through this project, I learned that data quality is the backbone of any AI system, the more precise and Alzheimer’s-specific the data, the more accurate the results. Further optimizations- Moving forward, I hope to reduce bias, improve stage differentiation, and incorporate additional patient factors and brain-region analysis to strengthen predictions. I also plan to create a user-friendly upload GUI/app that allows patients to upload and see the outcome. How to test my model? ** Click on the colab link below, scroll to the headline **Implement Interactive Image Upload and Prediction. To the following prompt enter upload. Upload your image and wait for the result to display.

Built With

• gemini
• google-colab
• python
• test.parquet
• train.parquet