Congrats to our Undergraduate Research Scholars

This past academic year three undergraduate students participated in the undergraduate research scholars (URS) program with MIMRTL. Here are more details about their projects:

Analysis of Probabilistic Classifiers for Disease Detection in Medical Imaging
Camilla Appiani
Project Description: The medical field is constantly growing with regards to new applications of artificial intelligence and machine learning. Current implementations include using neural networks with multiple layers to classify medical scans, to suggest diagnoses from medical images. This project analyzes probabilistic classifiers that associate a confidence rating to diagnostic classifications, where our goal is to understand how reliable and consistent probabilistic classifiers are. Using large datasets of chest x-ray images, we aim to test deep generative and Monte Carlo dropout classifiers and compare their diagnostic performance. If successful in demonstrating the consistency of probabilistic classifiers, it will promote further work to build and implement probabilistic classifiers in automated disease detection from medical images.

MRI Features of Immunotherapy Use in Lung Cancer Brain Metastasis Patients
Chengnan Li
Project Description: Immunotherapy treatments like immune checkpoint inhibitors (ICIs) are now routinely used for treatment of metastatic non-small cell lung cancer (NSCLC). Brain MRI changes in response to ICI and stereotactic radiosurgery (SRS) can sometimes be challenging to interpret. This project uses radiomic analysis to objectively characterize the differences between NSCLC brain metastases treated with and without immunotherapy after SRS. Post-SRS radiomic features were analyzed with a machine learning model using extreme gradient boosting algorithms to predict the presence of immunotherapy treatment. Preliminary results yielded a receiver operator curve AUC of 0.84 ± 0.12, but will be expanded on with more patients and MR imaging modalities. Radiomics and machine learning analysis can provide insight into assessing effectiveness of immunotherapy treatments and predicting patient outcomes.

Artificial Intelligence and Deep Learning in Medical Imaging
Paridhi Gupta
Project Description: Many diseases are time-sensitive, i.e., they can become much worse and even fatal if they do not get detected and treated in time. Detecting them using medical images requires careful observations, which takes a lot of focused time. This research aims to speed up the analysis process of whole body images, making it more reliable, efficient and accurate. It aims to use deep learning techniques to create an artificial intelligence to analyze and detect abnormalities and diseases from segmentations of medical images. It builds on the previously created technique of training Convolutional Neural Networks (CNN) to identify organs and abnormalities in from whole body images. The program trains, validates, tests, and then provides its accuracy rate for each organ.

Dr. McMillan also won an award for mentorship. More details are available on the Department of Radiology website: