ORKA Seminar: Deep Learning for Intelligent Surgical Tool Management

17 Aug 2021 2:10 PM - 3:00 PM
Presenter/Speaker: Mark Rodrigues
Location: S.1.03

We develop a multi-level Deep Learning prediction framework which provides a viable solution for intelligent management of surgical tools in a hospital. We  present a new hierarchical organised surgical tool dataset, HospiTools, for classification and retrieval of surgical tool images. In addition to 360 tool classes, we create a four level hierarchical structure for our dataset defined by 2 specialities, 12 packs and 35 sets. We employ different convolutional neural network training strategies to evaluate performance on this dataset, including the provision of prior information in the form of a taxonomic hierarchy tree structure. We demonstrate that the hierarchical nature of surgical tool classes could be used to make improved predictions. Our solution can potentially lead to greater patient safety, significant cost savings and increased efficiencies across New Zealand DHBs.