Evaluation of the efficiency of biofield diagnostic system in breast cancer detection using clinical study results and classifiers

Abstract

The division of breast cancer cells results in regions of electrical depolarisation within the breast. These regions extend to the skin surface from where diagnostic information can be obtained through measurements of the skin surface electropotentials using sensors. This technique is used by the Biofield Diagnostic System (BDS) to detect the presence of malignancy. This paper evaluates the efficiency of BDS in breast cancer detection and also evaluates the use of classifiers for improving the accuracy of BDS. 182 women scheduled for either mammography or ultrasound or both tests participated in the BDS clinical study conducted at Tan Tock Seng hospital, Singapore. Using the BDS index obtained from the BDS examination and the level of suspicion score obtained from mammography/ultrasound results, the final BDS result was deciphered. BDS demonstrated high values for sensitivity (96.23%), specificity (93.80%), and accuracy (94.51%). Also, we have studied the performance of five supervised learning based classifiers (back propagation network, probabilistic neural network, linear discriminant analysis, support vector machines, and a fuzzy classifier), by feeding selected features from the collected dataset. The clinical study results show that BDS can help physicians to differentiate benign and malignant breast lesions, and thereby, aid in making better biopsy recommendations.