Grants We Funded

Abstract

Impact Statement: Our aim is to establish image-based artificial intelligence as a modality to predict and address key surgical problems, across surgical disciplines. Pre-operative surgical imaging contains unrecognized and unmeasured anatomic, structural, tissual, and relational features that can help surgeons quantify risk of surgical complications. Unlocking such data can power new domains of risk prediction to transform surgical care. The results of this project will be key for a future federally-funded proposal to evaluate using image-based feature analysis to optimize prediction for a wide variety of post-operative complications. These results will directly minimize complications, reduce surgical costs, and improve the patient experience.

Project Summary: Surgery is common and increasingly able to treat a greater scope of healthcare conditions. Surgical complications are prevalent, and affect between 3-16% of all patients undergoing inpatient procedures. They impact all stakeholders due to worsened health for patients, challenges for surgeons, and increased costs for the healthcare system. Predicting complications before surgery, therefore, is paramount. Current risk prediction tools are based exclusively on structured clinical data. However, 80% of healthcare data is unstructured and not easily extracted from Electronic Health Records. Our central hypothesis is that pre-operative surgical imaging in particular contains unrecognized and unmeasured features indicative of the pathophysiologic mechanisms that drive post-operative complications. Incisional hernia (IH), a common long-term complication of abdominal surgery, presents an ideal opportunity to test image-based risk prediction due to: 1) manifestation across surgical disciplines; 2) high costs and morbidity; 3) challenging prediction; and 4) simple clinical detectability. Pilot analysis of CT images for 100 abdominal surgery patients, matched 1:1 for IH development, has confirmed there are common anatomic phenotypes which are associated with IH formation. To further refine these phenotypes, our group aims to build upon pilot work by analyzing preoperative CT images for 100 additional matched patients. Following manual segmentation, these images will be assessed for morphometric, intensity-based, and texture-based features. Using our total 200 patient studies, we will employ the validated Optimal Biomarker (OBM) methodology to distill the most salient features that portend hernia risk. The 5-10 most potent OBMs will be incorporated into a model to predict IH formation. The final model will be assessed for accuracy of prediction and area under the curve. A multi-fold cross validation analysis will be performed using balanced and unbalanced techniques. Our trans-disciplinary approach to image-based risk analysis will ultimately lead to a more vigorous model for IH formation. This line of research serves as a paradigm for future work using routine pre-operative imaging to predict risk of any post-operative complication.

Biography
Dr. John P. Fischer is an Assistant Professor and the Director of Clinical Research in the Department of Surgery at the Perelman School of Medicine at the University of Pennsylvania. Before joining the University of Pennsylvania Health System, Dr. Fischer completed his undergraduate studies at Hamilton College and medical school at SUNY Upstate Medical University followed by the completion of his plastic surgery training at the University of Pennsylvania. After becoming a faculty member, Dr. Fischer continued his education and received his MPH from the Bloomberg School at Johns Hopkins University. Dr. Fischer’s practice has a focus on aesthetic and reconstructive surgery of the trunk, specifically abdominal incisional hernia repair. Correspondingly, his research efforts focus on prevention, treatment and, clinical outcomes following abdominal hernia. He believes that risk prediction research will provide a pathway to prevention of incisional hernia as well as improve surgical outcomes. Dr. Fischer’s synergistic balance between active clinical practice and independent clinical research fosters a deeper appreciation for surgical and health problems which he proposes to use to maximize research impact and improve health of all patients.