EECE Assistant Professor Priya Deshpande is leading two projects at the intersection of heterogenous database systems, AI, and medicine and achieving milestones in her research projects.

The first project focuses on developing predictive text-generation for patients with aphasia, a language disorder resulting from brain injury. The objective of this project is to train Natural Language Processing models, using the comprehensive AphasiaBank database, to further enable them to predict subsequent words or form complete sentences from partial inputs. This approach aims to enhance communication for aphasia patients, reducing the effort needed to construct sentences. In the future, the team aspires to incorporate brain signals or fMRI data to predict their intended speech.

The second project explores the identification and precise localization of brain metastases in MRI scans using generative adversarial networks. Brain metastases occur when cancer cells spread from a primary tumor to the brain. Dr. Deshpande鈥檚 team applies fully convolutional neural networks (FCNNs) to segment metastatic lesions within brain MRI images, aiming to streamline and enhance the diagnostic process for clinicians. Their findings are detailed in the article 鈥淔ully Convolutional Neural Network Based Segmentation of Brain Metastases,鈥� co-authored with her student and published in Neural Computing and Applications.

In recognition of her talent, Dr. Deshpande and her collaborators received a $100K from The Northwestern Mutual Data-Science Institute (NMDSI) to lead a project focusing on developing a comprehensive framework for identifying and assessing biases in large language models within the health sector. Additionally, she received a 不良研究所 SEED Grant for her project, 鈥淚nvestigating Large Language Models for Improving Aphasia Patient鈥檚 Care and Speech Recognition.鈥� This one-year grant supports collaboration with Dr. William Gross from the Medical College of Wisconsin to further develop speech recognition tools for aphasia patients.