Arushi Gupta

I am a senior at the California Institute of Technology studying computer science. My research broadly focuses on developing machine learning methods, with a particular focus on computer vision, to address critical challenges in healthcare and drive scientific discovery. My past work includes creating methods to predict the effectiveness of surgical feedback, accelerate MRI reconstruction for high-dimensional imaging, and evaluate the robustness of single-cell image representation models.

Throughout my undergraduate research career, I am grateful to have worked with Prof. Anima Anandkumar and Prof. Pietro Perona from Caltech, Dr. Alex Lu from Microsoft Research, Dr. Andrew Hung from Cedars-Sinai, and Prof. Shreyas Vasanawala from Stanford.

news

Dec 04, 2024	I am excited to announce that Multi-Modal Self-Supervised Learning for Surgical Feedback Effectiveness Assessment has won a best paper award at ML4H 2024!
Nov 17, 2024	Our latest paper, ”Multi-Modal Self-Supervised Learning for Surgical Feedback Effectiveness Assessment,” is now available on arXiv!
Nov 11, 2024	I am honored to have received the 2024 Patrick Hummel and Harry Gray Travel Award! The funding will enable me to attend future conferences, allowing me to share my research with the wider academic community.

Selected Publications

Multi-Modal Self-Supervised Learning for Surgical Feedback Effectiveness Assessment

Arushi Gupta^*, Rafal D. Kocielnik^*, Jiayun Wang, Firdavs Nasriddinov, Cherine Yang, Elyssa Wong, Anima Anandkumar, and Andrew J. Hung

Machine Learning for Health (ML4H) Symposium, 2024

BEST PAPER Abs arXiv

The paper won the best paper award at 2024 Machine Learning for Health Conference.

During surgical training, real-time feedback from trainers to trainees is important for preventing errors and enhancing long-term skill acquisition. Accurately predicting the effectiveness of this feedback, specifically whether it leads to a change in trainee behavior, is crucial for developing methods for improving surgical training and education. However, relying on human annotations to assess feedback effectiveness is laborious and prone to biases, underscoring the need for an automated, scalable, and objective method. Creating such an automated system poses challenges, as it requires an understanding of both the verbal feedback delivered by the trainer and the visual context of the real-time surgical scene. To address this, we propose a method that integrates information from transcribed verbal feedback and corresponding surgical video to predict feedback effectiveness. Our findings show that both transcribed feedback and surgical video are individually predictive of trainee behavior changes, and their combination achieves an AUROC of 0.70+/-0.02, improving prediction accuracy by up to 6.6%. Additionally, we introduce self-supervised fine-tuning as a strategy for enhancing surgical video representation learning, which is scalable and further enhances prediction performance. Our results demonstrate the potential of multi-modal learning to advance the automated assessment of surgical feedback.
Greedy Learning for Memory-Efficient Self-Supervised MRI Reconstruction

Arushi Gupta, Batu M. Ozturkler, Arda Sahiner, Tolga Ergen, Arjun D. Desai, Shreyas Vasanawala, John M. Pauly, Morteza Mardani, and Mert Pilanci

International Society for Magnetic Resonance in Medicine (ISMRM) Annual Meeting, 2023

Abs HTML

Deep learning (DL) has recently shown state-of-the-art performance for accelerated MRI reconstruction. However, supervised learning requires fully-sampled training data, and training these networks with end-to-end backpropagation requires significant memory for high-dimensional imaging. These challenges limit the use of DL in high-dimensional settings where access to fully-sampled data is unavailable. Here, we propose self-supervised greedy learning for memory-efficient MRI reconstruction without fully-sampled data. The method divides the end-to-end network into smaller network modules and independently calculates a self-supervised loss for each subnetwork. The proposed method generalizes as well as end-to-end learning without fully-sampled data with at least 7x less memory usage.