As a clinical research fellow at Penn Plastic Surgery, the biggest lesson I learned was how artificial intelligence can unlock the wealth of unstructured data in our environment. In healthcare, that means unearthing information from things that could not traditionally be put in a risk model: radiology, pathology, clinical notes, pain drawings.
Our nascent efforts in preoperative abdominopelvic image analysis for incisional hernia (IH) prediction have yielded the below result! There are key pathophysiologic drivers of IH which can be appreciated on preoperative CT imaging. For example, these two colorectal surgery patients have been matched on many characteristics, but the patient on the left does not develop IH whereas the patient on the right does! One difference is the patient on the left has less visceral adipose than the patient on the right. Body habitus, which can be quantified and appreciated with image analysis, may play a fundamental role in the pathogenesis of IH!
Comments