Fall Research Expo 2020

Unsupervised Method of Automatic Lesion Segmentation and Quantification of Pathological COVID-19 Lung Tissue on Computed Tomography Scans

The COVID-19 pandemic is perhaps the greatest tragedy of 2020. Late spring, PURM was announced to take place remotely in the unconventional virtual format. Nevertheless, this experience I had this summer was no doubt one of the most invaluable I have had. Over the course of this summer, I worked with Dr. Maurizio Cereda, the co-director of HUP’s surgical ICU, and his team on research about acute respiratory distress syndrome and ventilator induced diaphragm dysfunction. The original plan was for this study to be conducted on pigs and mice to access the correlation between regional lung injury progression and treatment methods such as different mechanical ventilation modes and prone positioning. Due to the close association between COVID-19 and Dr. Cereda’s research, our project for this summer was modified to have a heavy clinical focus on COVID-19 patients. Our study trained an unsupervised program that quantitatively characterizes CT scans of COVID-19 patients to improve the efficiency of image analysis and better inform clinicians about prognostication and treatment plans. Our team works closely with Italy, one of the epicenters of the pandemic. To train the program, we processed CT scans of patients with SARS-CoV-2 infections from the Hospital of the University of Milano-Bicocca and the Hospital of San Gerardo. Throughout the summer, I helped paint and build the 3D lung models of many of these de-identified patients. As of today, our team continues to improve the model with more clinical data to help lower mortality in the ICU during this unprecedented condition. In addition to this research, Dr. Cereda and his assistant Yi Xin also hosted a weekly learning session in which we read and present medical journals about ventilator induced diaphragm dysfunction to prepare us for the re-opening of the animal lab this fall. These journal meetings were very educational parts of my PURM experience that helped me to gain deeper understanding and greater interest in this field of study. 

 

PRESENTED BY
PURM - Penn Undergraduate Research Mentoring Program
College of Arts & Sciences 2023
Advised By
Dr. Maurizio Cereda
Co-Director, Surgical Intensive Care Unit
Join Jiacheng for a virtual discussion
PRESENTED BY
PURM - Penn Undergraduate Research Mentoring Program
College of Arts & Sciences 2023
Advised By
Dr. Maurizio Cereda
Co-Director, Surgical Intensive Care Unit

Comments

Probably one of the coolest projects I've seen on here. What a creative way to approach this problem! I wonder if CycleGANs can be used in other cases as well. Is your code open source? Would love a link!

Important study amidst and for this current pandemic. I know you said your team used CT scan to help build model and create an unsupervised lesion segmentation. Have your team given a thought in using FDG-PET scan in doing something similar?