隱馬可夫模型對手術切骨操作意圖之初步研究

Abstract

人機互動是手術機器人相當重要的功能之一，其根據醫生的手術操作意圖，使手術機器人產生相應的輔助控制策略，協助手術進行。因此本論文利用隱馬可夫模型（Hidden Markov Model），描述手術過程中醫師的操作意圖。觀察人類的訊息傳遞過程是藉由不同的感覺受器，接受外在環境的刺激，再經由大腦的處理而產生知覺，知覺產生後人類會根據自身的經驗或是潛在的想法，進而調整所要執行的動作，動作最後對環境造成改變，大腦再次接受環境的刺激，修正動作行為，最後以達到目的。因此在切骨手術中，醫生會依據不同感官所觀察到不同的資訊，包括了觸覺及視覺，理解手術進行的狀況，並根據醫生本身所累積的經驗，產生不同的手術操作意圖，最後再轉換為手術的操作動作，本論文即藉由隱馬可夫模型，利用感測器來描述可以被觀察的醫生手術操作動作，並藉此來推測最有可能的手術操作意圖變化。而初步的實驗結果顯示所建構的隱馬可夫模型可適時地標定出操作意圖的改變，作為輔助切骨策略的重要依據。

This dissertation is to study the human-machine interaction in an orthopedics surgical robot. The human intention in bone cutting procedure is described using Hidden Markov Model (HMM). From observing the signal transduction of human, environmental stimuli are received by different receptors, passing to the brain where generating the cognition, and the motion are generated according to experience. Similarly an orthopedic surgeon determines how he/she performs bone cutting based on the information which could be observed including tactile and vision interfaces. In this dissertation force/vision sensors were used to measure the cutting forces, human exerting forces and cutting movement. Experiments were setup to collect the training data for the proposed HMM. The preliminary experimental results show that an HMM is of high potential to describe the human intention in bone cutting procedure. This result can be used in designing an assistive control strategy for surgical bone cutting.