PROJECT SUMMARY / ABSTRACT Mingrui Yang, PhD, is a biomedical imaging research scientist with quantitative background whose overarching goal is to conduct clinically oriented multidisciplinary research to improve the understanding, diagnosis, and treatment of musculoskeletal and other disorders through non-invasive medical imaging and quantitative methods such as machine learning. The study he proposes entitled Automated Arthroscopic Partial Meniscectomy using Deep Learning aims to develop automated systems for knee articular cartilage and meniscus segmentation and abnormality detection based on cutting-edge deep-learning models, as well as one- year patient outcome prediction after arthroscopic partial meniscectomy (APM) when physical therapy fails, utilizing MR imaging biomarkers identified by the deep-learning models. Candidate: Dr. Yang is a junior faculty member in the Program of Advanced Musculoskeletal Imaging (PAMI) of the Department of Biomedical Engineering at Cleveland Clinic. His training has been focused on quantitative sciences including mathematics, statistics, computer science, and technical development of magnetic resonance (MR) imaging. He aims to transition to clinically research in osteoarthritis (OA) and medical imaging. The proposed career development plan consists of four training goals to compel him toward an independent investigator: 1) Gain knowledge in knee OA and the role of APM in knee OA; 2) Gain an understanding of orthopaedic surgery, patient cohort, and clinical patient outcomes; 3) Gain an understanding of MR imaging biomarkers for knee OA and APM outcomes; 4) Gain expertise in clinical research. Environment: Dr. Yang and his primary mentor, Xiaojuan Li, PhD, have assembled a prominent team to guide Dr. Yang’s training and research plans. As a member of PAMI, he will work closely with researchers and clinicians in departments of Biomedical Engineering, Orthopaedics, and Diagnosis Imaging for his career development. Research: The proposed study will be carried out in three phases corresponding to the three aims: Aim 1 builds a deep learning system for automatic high-grade articular cartilage lesion detection on heterogeneous clinical knee MR images; Aim 2 develops an automated deep learning system to detect the presence of meniscal root tears on heterogeneous clinical knee MR images; Aim 3 utilizes the imaging finding(s) from the deep learning system(s) and patient demographics to predict the clinical outcomes after APM. Summary: The proposal will provide a novel, automated, and consistent tool for cartilage and meniscus segmentation and lesion detection on heterogeneous knee MR images collected from clinical routine practice. A prediction model using these imaging findings with patient demographics can help predict clinical outcomes for patients undergoing APM surgery. This proposal will also advance Dr. Yang’s career development toward an independent investigator in OA and biomedical imaging research.

Project Narrative Arthroscopic partial meniscectomy provides no clinically meaningful benefits for certain patient groups even after physical therapy fails. Preoperative identification of this population can help substantially improve clinical treatment and management plans of these patients by reducing unnecessary surgeries and cost to the healthcare system. The proposed study will provide a novel outcome prediction model to achieve this goal by utilizing imaging findings from a deep learning based automated and consistent system for cartilage and meniscus segmentation and lesion detection on heterogeneous knee MR images collected from clinical routine practice.