Table of Contents

  1. Introduction
  2. System Architecture & Conceptual Design
  3. Conclusion
  4. Links

Introduction

The Medical Image Analysis Research Group at the Teaching Hospital Peradeniya handles critical diagnostic imaging cases (e.g., tumors, fractures, internal anomalies) and radiographic MRI/CT examinations. Currently, the department relies on manual visual inspection, physical radiological films, and human-based Diagnostic Imaging Reports (DIR).

This manual workflow creates critical operational bottlenecks:

MedImgSys is a deep learning system engineered specifically to automate, secure, and streamline the diagnostic imaging workflows of the Radiology Department. The primary objective of this project is to implement a robust, convolutional neural network pipeline that guarantees high accuracy, enforces precise segmentation constraints, maintains performance logging via metric tracking, and optimizes complex model architectures for diagnostic report compilation.


System Architecture & Conceptual Design

The core of this system is designed around strict machine learning principles to ensure minimal false negatives, high diagnostic integrity, explicit feature extraction via specialized network architectures, and structured model optimization.

System Architecture Breakdown


Conclusion

The MedImgSys delivers a computer vision infrastructure engineered for the Medical Image Analysis Research Group at Teaching Hospital Peradeniya. By shifting from manual visual inspections to an automated deep learning pipeline with explicit classification/segmentation architectures, the system guarantees diagnostic integrity, preserves precise spatial boundaries for lesions and anomalies, enforces consistent evaluation over sensitive data, and speeds up diagnostic report aggregation.

Key Architecture Achievements: