Challenge
The objective was to build an automated system capable of processing and classifying large-scale chest image datasets collected from multiple hospitals for breast cancer treatment analysis.
Internship – DCPT (Medical Image Classification Pipeline for Breast Cancer Research)
Process
Dataset Analysis
Investigated imaging datasets from multiple hospitals to understand inconsistencies in file naming, metadata availability, and image quality.
Pipeline Design
Designed modular Python pipelines responsible for preprocessing, metadata parsing, classification, and dataset restructuring.
Model Training
Trained and optimized a YOLOv8 classification model, experimenting with preprocessing and augmentation techniques to improve generalization.
Validation & Testing
Evaluated model performance using precision, recall, and F1-score, followed by manual verification across datasets from different hospitals.
Results
This project produced a reliable machine learning pipeline for automated medical image classification and dataset organization. By standardizing datasets from multiple hospitals and automating processing workflows, the system provides a scalable foundation for future BCCT research.
Stack
Conclusion
The project has demonstrated how modern technologies such as Python, YOLOv8, PyTorch can be combined to create Internship – DCPT (Medical Image Classification Pipeline for Breast Cancer Research)—a scalable and modular system offering high precision and user-friendliness.