Research Background

Idea

My current research focuses on Few-Shot Learning for Medical Image Analysis. In the medical domain, the availability of large image datasets is often limited, making traditional supervised machine learning approaches less effective. Meta-Learning has emerged as a promising technique for addressing this challenge, as it enables models to learn new tasks in novel domains from only a few samples. Meta-Learning methods are widely applied to solve few-shot learning problems in data-scarce fields like medical imaging.

Our research mainly aims at designing efficient few shot learning algorithms to address the common problems in medical imaging domain. We have basically targeted the main crux of meta learning and few shot learning and have tried to come up with our research objectives.

Basically we can divide our research objectives into the following:

• Domain specific prototype creation in prototypical netowrks for medical imaging perspective

• Adaptive Sampling over randomized sampling of classes as well as instance in few shot paradigm for medical image classification

• Novel Ways of creating class representative for unseen classes in zero shot classification paradigm.

Domains Addressed

In our phd works we have tried to solve few shot classification problems in various medical imaging domain which are displayed below :

• Dermatoscopic Images

1.  Derm7pt Dataset

2.  SD-198 Dataset

3.  ISIC-2018 Dataset

• Breast Cancer Histopathological Image Dataset(BreakHis Dataset)

• Blood Cell MNIST Dataset

• Pathology MNIST Dataset

• Chest XRay MNIST Dataset

• NIH Chest XRay Dataset

• Diabetic Retinopathy Dataset