Micah V
- Research Program Mentor
PhD candidate at University of California Los Angeles (UCLA)
Expertise
AI for precision medicine, software engineering, biomedical engineering, deep learning, medical imaging, computer vision
Bio
Hi there! I'm Micah, a Ph.D. student in Bioengineering at UCLA specializing in AI for medical imaging, with 4+ years of experience in deep learning, multimodal data extraction/integration, biomarker discovery, and medical imaging analysis. I'm experienced in developing end-to-end medical imaging pipelines in ophthalmology, radiology, and neurology for AS-OCT, MRI, DTI, CT, and fluoroscopy imaging, including tasks of segmentation, classification, registration, and detection. I'm passionate about translating medical imaging research into scalable, explainable AI solutions for healthcare applications. In my free time, I enjoy bike riding, running (I used to do cross country!), classical guitar, board & video games, reading fantasy novels, and listening to music (mostly soundtracks & classical). I'm always interested in anything that tells a great story, whether it be through a book, movie, musical piece, or any other medium. I am trying to be more consistent with running and biking and enjoy being in the outdoors when I'm not at my computer coding!Project ideas
Classification of Eye Diseases in Fundus Imaging using a Multitask Vision Transformer
Many eye diseases can manifest with different abnormalities in Fundus imaging, a type of imaging that images the rear region of the eye (the fundus). Creating an automated method of classifying these diseases from a large dataset of fungus images would benefit physicians by reducing workload and automating current clinical software. This could be achieved by curating a dataset of various fundus images representing common eye diseases, and training a multitask vision transformer to both segment regions of diseased regions of the eye, and to classify the overall image as either a healthy or diseased eye. I've worked with a few types of eye imaging, as well as CT imaging for chest, MRI, DTI, and fluoroscopy for brain. This project is an example of the type of software development that I'm familiar with: automating a component of physician workload via a deep learning model (CNN or transformer based). Undertaking this type of project would have you learning coding skills (Python programming language), how to curate, organize, and maintain a dataset of medical images, communicate results with potential users of your software, and practice writing and presenting your research findings. The end goal of this project is usually either a poster or paper publication in a scientific journal.