Catheter Position Classification in Chest X-Rays using ResNext50

Deep learning model for automated classification of catheter and tube positions in chest X-rays, achieving 93% accuracy across 11 position categories.

Overview

This project implements a ResNext50-based deep learning model to classify the positions of medical devices (ETT, NGT, CVC, Swan Ganz Catheter) in chest X-rays as Normal, Borderline, or Abnormal. The solution incorporates novel anatomical masking techniques and robust data preprocessing to enhance model performance.

Project Structure

Setup

1. Clone the repository:

git clone https://github.com/sankalp112kashyap/ChestX-RayMultiLabelClassification.git
cd ChestX-RayMultiLabelClassification