BA, University of Michigan
MD, Ohio State University
Neurosurgery residency, University of Michigan
The objective of my research program is to improve the surgical management of skull base and malignant brain tumors by developing methods that detect, diagnose, and interpret brain tumor tissue at the time of surgery. Surgical goals are divergent depending on the underlying brain tumor diagnosis; unfortunately, we do not know the tumor diagnosis prior to surgery in the majority of patients. A major problem with our current techniques for diagnosing brain tumors during surgery is that they are time-, resource-, and labor-intensive. To address this problem, our laboratory has developed a rapid optical imaging method, called stimulated Raman histology (SRH), that can acquire high-resolution images during surgery. Moreover, my group focuses on developing algorithms to detect tumor tissue and diagnose the type of tumor in an automated fashion. We use the latest advances in artificial intelligence (AI) and machine learning to improve the speed and accuracy of intraoperative brain tumor diagnosis. Our work culminated in a multicenter prospective clinical trial, where we demonstrated that an AI-based diagnosis of SRH images was equivalent to pathologist-based interpretation of conventional histologic images for the 10 most common brain tumor types (94.6% versus 93.9% accuracy, respectively). This research laid the foundation for our efforts to improve the surgical management of brain tumor patients using state-of-the-art biomedical imaging and AI algorithms.
The future direction of our research is in two related and parallel directions. Firstly, and as an extension of our previous results, we aim not only to identify specific tumor types, but also to identify specific gene mutations present in the tumor types at the time of surgery. The current brain tumor classification system from the World Health Organization uses both histologic features and genetic mutations to classify brain tumors into diagnostic groups. Our aim is to provide surgeons with the best diagnostic data possible so that they are able to provide optimal surgical care. Secondly, we hope to use our expertise in SRH, AI, and computer science in order to discover the complex relationships between gene mutations (i.e., genotype) and the growth pattern/clinical behavior of brain tumors (i.e., phenotype). These genotype-phenotype relationships have been challenging to elucidate because they are inherently multifactorial without simple one-to-one mappings.
Rapid intraoperative molecular classification of diffuse gliomas using stimulated Raman histology and deep neural networks
Data science / Analytics / AI | Deep neural networks | Neuro-oncology | medical imaging
- Rapid Intraoperative Diagnosis of Sellar Region Tumors Using Stimulated Raman Histology. Podium Presentation
- Abstract PR005: The dynamic tumor microenvironment: Oncostreams are self-organizing structures that modulate glioma progression and treatment
- Ventricular Volume Change as a Predictor of Shunt-Dependent Hydrocephalus in Aneurysmal Subarachnoid Hemorrhage
- Co-investigator of: The role of collagen and its signaling mechanisms in glioma progression and invasion.
- Principal investigator of: Molecular diagnosis of brain tumors using stimulated Raman histology
- Principal investigator of: Molecular Classification of Diffuse Gliomas using Deep Learning and Optical Imaging