Cardiac Image Analysis
Currently, most widely used image-based biomarkers of cardiac structure and function have been limited to global indices, such as infarct mass, left ventricular (LV) mass, end-diastolic volume, and ejection fraction. With recent advancements in MRI technology to acquire high-resolution three-dimensional (3D) images of the heart and more accurate and quantitative techniques to image the infarct structure, there is a significant opportunity for realizing more sensitive and more localized measurements of cardiac structure and function. One of the main challenges to properly evaluate and to eventually translate this imaging technique for analysis of structure and function of the heart, is the lack of robust automated image analysis methods. The following is an image processing pipeline for building and simulating electrophysiological models of the heart.
Digital Histopathology Image Analysis
Hirschsprung's disease is a critical medical condition that affects the
large intestine and causes problems with passing stool in small children. The condition is present at birth as a result of missing
nerve cells in the muscles of the child's colon. Surgery to bypass the part of the colon that has no nerve cells is a major treatment
for Hirschsprung's disease. The decision for this surgery is based on qualitative inspection of density of nerve cells in the digital
histopathology images of excised specimens. However, even leading clinical experts cannot avoid human errors which can lead to catastrophic
results. Therefore, developing an automated and quantitative approach for processing of digital histopathology images of patients could
enormously increase the accuracy of decisions made by clinicians.
Forth Year Project Supervision
2016-2017
- Gesture Control-based Medical Image Navigation System: It is challenging for clinicians in the operating room to use a mouse to navigate through 3D medical images during the procedure. In this project, the students will investigate the use of gesture control (e.g. Microsoft Kinect) to navigate to the proper organ/plane in an medical image.
- "Writing in the Air" Guest Book Based on Kinect (Co-supervised with Dr. Sreeraman Rajan): The purpose of this project is to design and implement a guest book for the SCE 75 th anniversary project. The system will detect the fingertip path and either convert it into a binary image to preserve the handwriting, or convert it into text using properly interfaced OCR (optical character recognition) software. The system should also support moderation of the guest book entries: send them to the moderator and display them on the systems screen after validation.
- Web-based Medical Image Processing Software: While there are many standalone software for medical image analysis and segmentation, very few web-based medical image processing software tools are available. In this project, the expectation is for students to develop a web-based medical image processing software platform. A user should be able to load an image, then perform basic pre-processing, image segmentation such as, image thresholding and region growing, and visualization.