Construction of annotated data for analysis of recorded cortical mapping videos

There is a need of surgery workflow analysis to increase an efficiency of advanced medical care. Surgical Operations have been recorded by several sensors for such as postoperative analysis and incidents detection. In particular, surgical video recording is commonly used, so there are some audio-visual recorded data, and they are useful to obtain a better understandings and description of advanced surgical operations. However, the recorded videos are not usually annotated, so it is not simple to conduct computational analysis, and data annotation is necessary to handle by computer. We target videos of awake craniotomy which is a special neurosurgery in this work. The cortical mapping process is the most important for brain tumor resection in awake craniotomy. Therefore, we aim to annotate this process to analyze medical staff's knowledge. We assume that the factor that affects the surgical procedures is below: positions of direct electric stimulations, duration of the stimulus, current intensity, tasks presented for patients. In this paper, we constructed annotated data from clinical recorded awake craniotomy videos. Data collection is performed manually by graphical user interface because several terms of annotation are hard to annotate completely automatically. After that, we visualized the several of annotated data and discussed the effect.