Anne van Dalen

A review on the current applications of artificial intelligence in the operating room I 223 8 Three techniques were researched, namely, adaptive neuro-fuzzy inference systems (ANFISs), multiple linear regression analysis (MLRA), and ANNs. However, ANFIS is a fusion between the adaptive learning capability of ANNs and the intuitive logic of human reasoning, formulated as a feed-forward neural network. The results of procedure duration prediction were compared between the three techniques, and the ANFIS model came out to be performing better than the other 2 as portrayed in Table 3. 24 Gesture Recognition To decrease the risk of contamination during surgical procedures, Cho et al. 23 researched a noncontact interface based on ML models in order to enhance the accuracy of gesture recognition. Support vector machines (SVMs) and naive Bayes classifiers, ML models with associated algorithms used for classification, were used in the study. 30 Cho et al. used 30 features, including hand and finger data, as input for these ML models to predict and train 5 types of gestures. The overall accuracy of the 5 gestures was 99.58% ± 0.06 and 98.74% ± 3.64, respectively, for SVM and naive Bayes classifiers. Self-training methods of SVMs and naive Bayes classifiers improved accuracies by about 5–10%. 23 Intraoperative Cancer Detection During brain tumor removal it is important yet very difficult to detect and remove all cancer cells. As a consequence, when not completely removed, the patient is at risk for recurrence of cancer. With certain types of brain cancer in vivo, Raman spectroscopy can detect these invasive cancer cells. A downside to this technique is the fact that the Raman signal is weakened by spectral artifacts generated by the regular lights in the OR. Table 3. Comparison of techniques to estimate procedure duration. 24