Workflow mining for visualization and analysis of surgeries

Objective: Modeling the workflow of a surgery is a topic of growing interest. Workflow models can be used to analyze statistical properties of a surgery, for intuitive visualization, evaluation and other applications. In most cases, workflow models are created manually, which is a time consuming process that might suffer from a personal bias. In this work, an approach for automatic workflow mining is presented. Materials and methods: Ten process logs, each describing a single instance of a laparoscopic cholecystectomy, are used to build a Hidden Markov Model (HMM). Using a merging approach, models at different levels of detail are generated. These embody statistical information concerning aspects like duration of actions or tool usage during the surgery. Results: A Graphical User Interface (GUI) is presented, that uses a graph representation of the HMM to intuitively visualize surgical workflow. It allows changing the level of detail by expanding and merging nodes. The GUI can also be used to compare videos of surgeries which are synchronized to the model. Conclusions: The proposed method allows automatic generation and visualization of a statistical model describing the workflow of a surgery.