Arunkumar Kannan

I am a first second third-year PhD student in the Department of Electrical and Computer Engineering at Johns Hopkins University working with Prof. Brian Caffo. My research interests focus on the integration of deep learning algorithms with statistical models to provide insights into black-box predictions, particularly in the context of analyzing structural and functional brain imaging data. Additionally, I have a keen interest in exploring generative models like diffusion models in the context of medical imaging applications, where these models hold potential for generating counterfactual data, which is highly valuable in scenarios with limited available data.

Prior to joining Hopkins, I received my master's degree from the School of Biomedical Engineering in University of British Columbia, Vancouver, where I worked at Biomedical Signal and Image Computing Laboratory (BiSICL) with Prof. Rafeef Garbi. At BiSICL, I worked on applying deep learning models for automating the diagnosis and management of developmental dysplasia of the hip (DDH), specifically proposing novel diagnostic frameworks that provide model uncertainty measure to clinicians to promote confidence over the predicted DDH metrics. I received my bachelor's degree (with distinction) in Biomedical Engineering from the Faculty of Information and Communication Engineering, Anna University, India, with a focus on biomedical signal processing and biostatistics.

Email  /  CV  /  Google Scholar  /  Twitter  /  LinkedIn  /  Github

I'm interested in exploring diffusion models for medical imaging applications, ML explainability, and high-dimensional feature representation learning frameworks, with its applications to structural and functional brain connectomics data.

This article provides an overview of the current landscape in functional brain measurement, with a particular focus on fMRI, highlighting key developments in the past decade. Furthermore, it looks ahead to the future, discussing unresolved research questions in the community and outlining potential research topics for the future.

Project Website | Poster

GAMMA-FACE introduces a novel approach to address bias amplification in face-generative diffusion models using Gaussian Mixture Models (GMMs). By leveraging GMMs, we disentangle facial attributes and localize the means within the latent space of the diffusion model, effectively reducing bias on-the-fly without the need for retraining.

Code (Coming Soon) | Poster

We introduce a novel and easily implemented analysis approach aimed at explaining the variation in functional connectivity of the brain by integrating important structural factors such as anatomical morphology summaries, voxel intensity, diffusion-weighted information, and geographic distance.

Gaussian Harmony introduces a novel approach to addressing bias amplification in diffusion models for face generation by balancing facial attributes through the localization of attribute means in the latent space using Gaussian mixture models.

We proposed a technique to quantify confidence in the segmentation process that incorporates voxel-wise uncertainty into the binary loss function used in the training regime of a deep neural network, which encourages the network to concenterate its training effort on its least certain predictions.

We proposed interpretable uncertainty measures that can simultaneously measure bone segmentation reliability and quantify scan adequacy in clinical DDH assessment from 3D Ultrasound.

Website template courtesy: Jon Barron