Surgical Procedure Recognition Using Quantum Machine Learning

Surgical procedure recognition is the process of identifying tasks and gestures done during a surgical process and is a field that has been widely researched due to its use in robot assisted surgeries to improve surgical performances and training. This work investigates the use of Quantum Machine Learning (QML) algorithms, in particular Quantum Support Vector Classifier (QSVC), for the identification of patterns in kinematic data collected from the JIGSAWS dataset which includes 76 kinematic features related to suturing, knot tying and needle passing. In order to evaluate QSVC performance, we compared its performance measures such as accuracy, precision, recall and F1-score with that of a classical Support Vector Classifier (SVC). Quantum kernel-based methods like QSVM embed classical data into high-dimensional Hilbert spaces via quantum feature maps, offering the potential to capture complex data relationships more efficiently. Using ZFeatureMap and quantum circuits implemented in Qiskit, we demonstrate that QSVM shows slight performance advantages over its classical counterpart in specific tasks. These findings lay the groundwork for a context-aware surgical system to support medical practitioners in real time and help advance surgical practice and educational approaches for enhancing patients quality of life.