Tushar Nayak

Pittsburgh PA

I am a graduate student at Carnegie Mellon University, pursuing a research-oriented master’s in the Biomedical Engineering department, advised by Professor Kenji Shimada at the Computational Engineering & Robotics Lab, where I work on computer vision for image-guided robotic intervention.

My research combines medical imaging, geometric vision, and physics-aware learning models to recover structure, motion, and deformation from sparse or noisy clinical data.

I build these methods for surgical and biomedical settings including endovascular robotics, 3D reconstruction, image-guided navigation, and learning systems grounded in real clinical workflows. Always happy to connect. Reach me at tusharn [at] andrew [dot] cmu [dot] edu or at the Steffey Robotics Lab space at Scaife Hall.

Surgical Robotics

Vision systems for endovascular robotics, tool tracking, and deformable operative scenes.

Medical Imaging

Learning-based analysis of MRI, CT, and other clinical scans for diagnosis and planning.

3D Vision & Reconstruction

Geometry recovery from sparse views, point clouds, and neural scene representations.

Computational Pathology

Histopathology models for cancer grading, subtype detection, and slide-level reasoning.

Vision-Language Models

Open-vocabulary querying and semantic grounding for surgical and biomedical scenes.

Physics-Informed Neural Networks

Models that bake in deformation, dynamics, and other physical constraints when the data alone is not enough.

featured research

ongoing Surgical Robotics Medical Imaging

Physics-Informed Endovasculature Deformation Estimation And Registration

Research at CERLAB (master's thesis)

paused Medical Imaging Neural ODEs

Spatiotemporal Glioblastoma Evolution Visual Prediction

Research at The ∀ Lab & Image Science Lab

featured projects

active prototype Surgical Robotics Tool Tracking

Open-Vocabulary Surgical Tool Detection and Tracking

A real endoscopy pipeline that turns text prompts into surgical tool boxes, masks, tracking overlays, and evaluation metrics using Grounding DINO and SAM2.

active extension 3D Vision Medical Imaging

Gaussian Occupancy Cardiac Reconstruction

A sparse-view echocardiography project that recovers 3D cardiac anatomy with Gaussian occupancy fields, differentiable slice supervision, and mesh evaluation.

completed Medical Imaging Mammography

Attention-augmented Dual-View Mammogram Alignment for Enhanced BI-RADS Classification

Course project for Projects in Biomedical AI

completed 3D Vision Medical Imaging

Few-Shot 2D Echo to 3D Cardiac Reconstruction via Neural Implicit Priors

Course Project for Learning for 3D Vision

I work on the vision sub-system of a haptically enabled endovascular robotic tele-surgerical platform as part of my masters thesis with doctoral candidate Rishi Basdeo. My main area of focus is investigating physics informed neural networks and neural ordinary differential equations to quantify deformation from 2D fluroscopy angiograms and register the deformation to the pre-operative computer tomography angiographs.

Before Pittsburgh, I completed my undergrad at Manipal Institute of Technology with a major in Biomedical Engineering, focusing on pattern recognition and image processing, and a minor in Data Science. Over the second half of my undergrad, I worked under Professor Niranjana S and Dr. Krishnaraj Chadaga at the Biomedical Computing Lab on projects ranging from skin lesion image-based viral infection detection to multi-stage and multi-modal cancer detection.

Between my bachelors and masters, I also spent a year as a researcher at IIT Hyderabad working on motion-capture & electromyography analysis for exercise and the Indian Council of Medical Research where I worked on an ultrasound-based fetal anomaly system.

Besides lab-work, I'm also usually giving/boring people with my usual spiel as one of the BME department's ambassador, going trailbiking, playing the piano or continuing work on this website!

news

May 20, 2026	I will be spending my summer as a researcher at Surreality Lab with Rishi Basdeo and PIs Professor Edward Andrews and Professor Jacob Biehl
Jan 16, 2026	Began serving as a Teaching Assistant for Professor Newell Washburn’s course, Machine Learning Applications in Experimental Biomedical Research, at Carnegie Mellon University. A full-circle moment, since I took the course myself in Spring 2025.

latest posts

Jun 02, 2026
May 27, 2026	sparse-view clinical reconstruction: explicit vs implicit representations

selected publications

lung-2024

Automated histopathological detection and classification of lung cancer with an image pre-processing pipeline and spatial attention with deep neural networks

Tushar Nayak, Nitila Gokulkrishnan, Krishnaraj Chadaga, and 3 more authors

Cogent Engineering, 2024

Abs DOI PDF

Lung Cancer is a major cancer in the world and specifically India. Histopathological examination of tumorous tissue biopsy is the gold standard method used to clinically identify the type, sub-type, and stage of cancer. Two of the most prevalent forms of lung cancer: Adenocarcinoma & Squamous Cell Carcinoma account for nearly 80% of all lung cancer cases, which makes classifying the two subtypes of high importance. Proposed in this study is a data pre-processing pipeline for the H&E-stained lung biopsy images along with a customized EfficientNetB3-based Convolutional Neural Network employing spatial attention, trained on a public three-class lung cancer histopathological image dataset. The pre-processing pipeline employed before training, validation and testing helps enhance features of the histopathological images and removes biases due to stain variations for increased model robustness. The usage of a pre-trained CNN helps the deep learning model generalize better with the pre-trained weights, while the attention mechanism On three-fold validation, the classifier bagged accuracies of 0.9943 ± 0.0012 and 0.9947 ± 0.0018 and combined F1-Scores of 0.9942 ± 0.0042 and 0.9833 ± 0.0216 over the validation and testing data respectively. The high performance of the model combined with its computational efficiency could enable easy deployment of our model without necessitating infrastructure overhaul.

What I work on