Lung Nodule Classifier (LIDC-IDRI)

This project uses a Convolutional Neural Network (CNN) to classify lung nodules as benign or malignant using the LIDC-IDRI dataset.

#📁 Dataset

• Source: LIDC-IDRI - The Cancer Imaging Archive (TCIA)
• Format: Preprocessed PNG images (e.g., slice-0.png)
• Directory structure:

Model

CNN with:

Conv2D, MaxPooling, BatchNorm, Dense layers

Input size: 128x128 grayscale

Augmentation: rotation, flip, shift

Output: 2 classes (binary classification)

Output

Trained model saved as: lidc_final_model.h5

Accuracy: ~62% (can be improved with proper labels)

Notes

Currently uses random labels. Replace with true labels for real medical classification.

Works with PNG slices derived from CT scans.

Download model: Model file