Statistical quantification of colocalization between interacting bio-molecules is an essential step in understanding inter-process associations. In this work we propose a statistical learning based technique to identify and evaluate the degree of statistical interaction between two spatial processes. We have just published two papers on this.

• Quantitative colocalization in microscopy, IEEE Signal Proc. Letters'20 [Preprint] [Paper]

• Cluster Core Correspondence Index (C3I) for point-cloud stability, IEEE TIP'20 [Preprint][Paper]

We developed hybrid deep learning based algorithm to segment the endometrium fundus from transvaginal ultrasound. This work was done at GE-GRC, and was published in IEEE ISBI'17 [Paper].

A related work was also published in IEEE ISBI'17 which integrates graph cuts with deep learning for 3D segmentation [paper] for segmenting lung nodules from 3D low dose CT scans.

We developed this algorithm to segment neurons from noisy (2D/3D) confocal microscopy images. It could be applied for any segmentation problems involving filamentous structures. Check out our paper in IEEE Transactions on Image Processing. The code is available too. [Paper][Matlab Code]

This segmentation technique generalizes Chan-Vese's method to segment object in presence of inhomogeneous intensity. Please check out our paper in IEEE Signal Processing Letters. The code is available too. [Paper][Matlab Code][Python code]

This project involves segmentation of blood vessel boundary from ultrasound to assist medical practitioners for performing phlebotomy application such as intravenous needle placement. The proposed solution to this problem involves the integration of a level set segmentation methodology with the dictionary learning framework. Please check out our paper in IEEE Signal Processing Letters. The code is available too. [Paper][Matlab Code]