Computer aided anesthesia monitoring: Anesthesia depth prediction and respiratory sound analysis

Anesthesia is a very important part of a surgery. Anesthesiologists monitor and analyze meaningful changes in patient's vital signs with their experience to adjust the anesthetic dosage and deal with critical incidents. Computer aided anesthesia management can reduce the workload of the anesthesiologists routine tasks and help anesthesiologist make more accurate and reliable decision.;This paper introduces a method of analyzing and characterizing patterns in lung sounds for critical condition care during surgery in the operating room. The method combines a new noise cancellation and signal separation technique with a stochastic classification approach for lung sound pattern analysis in real time. The noise cancellation and signal separation technique employs the unique features of breath and cardiac cycles to achieve sound channel identification and heart/lung extraction consecutively and iteratively, without any assumptions on signal/noise frequency separation or stochastic independence. Recursive optimized stochastic analysis is performed to extract sound patterns. This method highly improve the accuracy and reliability of lung sound pattern recognition and diagnosis under the impact of noise artifacts in OR.;We also develop a real time patient model for anesthesia drug infusion response. The modeling methodology captures the unique features encountered in developing a computer aided control strategy for anesthesia drug infusion. Rather than using models of high complexity, we follow the insights of anesthesiologists in representing the basic features of a patient response to drug infusion that are essential for computer-aided infusion control. For identification, the model parameters are initiated by expert knowledge and improved upon in real time using two step stochastic approximation when clinical measurement data become available.