Novel multimodal sensing and machine learning strategies to classify cognitive workload in laparoscopic surgery

Area of Science:

• Medical technology
• Machine learning in healthcare
• Surgical simulation

Background:

• Surgeons face high cognitive workload (CWL) during operations, impacting patient outcomes and well-being.
• Objective CWL measurement is crucial but challenging with isolated physiological measures.
• This study addresses CWL classification and missing pupil diameter data imputation.

Purpose of the Study:

• To develop and propose a multimodal machine learning (ML) approach for classifying CWL levels.
• To create an ML approach for imputing missing pupil diameter (PD) data.
• To enhance surgeon well-being and surgical training through accurate CWL assessment.

Main Methods:

• Ten surgical trainees performed simulated laparoscopic cholecystectomies with increasing cognitive load.
• A novel sensor platform (MAESTRO) recorded physiological data: EEG, fNIRS, EMG, ECG, and PD.
• A multimodal ML model (CNN-BiLSTM, BMRFO, OCS) classified CWL, with OCS used for PD imputation.

Main Results:

• The multimodal ML approach achieved 97% accuracy in classifying CWL levels.
• The CNN-BiLSTM and BMRFO model outperformed other classification techniques.
• The OCS method showed superior performance in imputing missing PD data (9.15% RMSE).

Conclusions:

• A multimodal ML approach accurately classifies perceived cognitive workload levels in surgeons.
• This method holds potential for surgical training, assessment, and developing in-operating room cognitive support systems.