Real-time Detection of Surgeon Stress Level During Laparoscopic Training

Yi Zheng, Grey Leonard, Ann Majewicz Fey

Introduction

Laparoscopic surgery is a new surgical technique which has the advantages of less large open wounds and less pain or discomfort. It helps decrease the patient’s recovery time. Surgical stress during the operation is commonly experienced by the surgeons and has a negative effect on their surgical performance and patient safety. The purpose of this study is to find a framework for a real-time and objective detection of the stress levels in conjunction with an assessment of performance using the kinematic data.

Method

30 medical students were recruited for the study and were randomized into control and stressed group. Each subject was asked to complete a six-minute peg transfer drill on a simulation hardware (FLS trainer). The control group proceed while hearing normal vital signs. The stress group performed under a period of progressively deteriorating vital signs with a particular increase in intensity beginning at the three-minute mark. The moderator also simultaneously provided feedback to the illusory anesthesiologist and nurse circulator of the increased danger of the dummy patient. The movement of hands were recorded and streamed using electromagnetic trackers mounted on the handles of tools (3D positional data).

Results

The data of control group was annotated as stress level 0. The data of the first and second halves of stressed group was annotated as stress level 1 and 2, respectively. A LSTM Classifier was trained by using the recorded movement data to detect the stress levels. The data was framed into 60-sample windows with a 50% overlap. We used the Leave-One-Participant-Out cross-validation method (LOPO) to test the classifier and got an overall F1 score of 0.8630(0.011).

Conclusion

This study showed the feasibility of using kinematic data to objectively detect the stress level experienced by the surgeons during laparoscopic training. And the proposed method can serve as a groundwork for the stress detection in robotic surgical training (da Vinci System) in which the kinematic data can be streamed directly.