Blogs | Debesh Jha

ColonSegNet for Real-time Polyp Segmentation

February 6, 2023

ColonSegNet is a lightweight polyp segmentation architecture utilized by NVIDIA in the Clara Holoscan Sample App for colonoscopy polyp segmentation. With an image size of 512×512, it achieves a dice coefficient of 82.06% for segmentation tasks and an average precision of 80.00% for detection tasks.

DoubleUNet for Medical Image Segmentation

February 3, 2023

DoubleUNet is a novel architecture that improves upon the traditional U-Net model for medical image segmentation tasks. It introduces a dual encoder-decoder structure to enhance feature extraction and segmentation accuracy.

Explainability of Deep Learning Models

January 24, 2023

Deep learning models have achieved remarkable success, but their black-box nature raises concerns. This blog discusses various techniques to interpret and explain deep learning models, enhancing their transparency and trustworthiness.

Artificial Intelligence in Radiation Therapy for Cancer Treatment

January 24, 2023

Artificial Intelligence (AI) is revolutionizing cancer treatment, including radiation therapy. This blog explores how AI enhances treatment planning, delivery, and outcome prediction in radiation therapy.

Introduction to Transformer Model

January 20, 2023

Transformers have transformed natural language processing tasks. This blog introduces the transformer architecture, its components, and its applications in various NLP tasks.

Machine Learning-based Classification, Detection, and Segmentation of Medical Images (PhD Thesis Summary)

January 20, 2023

This blog summarizes my PhD thesis, focusing on machine learning approaches for classification, detection, and segmentation of medical images, particularly in gastrointestinal endoscopy.

ResUNet++

January 20, 2023

ResUNet++ is an advanced architecture for medical image segmentation, incorporating residual blocks, squeeze and excitation blocks, ASPP, and attention mechanisms to enhance performance.