Publications

Abstract : Multi-parameter patient monitors (MPM) using human vital parameters, heart rate, blood pressure, respiration rate and oxygen saturation (SpO2), are extremely valuable in enhancing the health care of the patients in the intensive care unit (ICU) and general wards. Linear support vector machine (SVM) based implementations are computationally more efficient over the non-linear Kernel based SVMs. This makes the linear SVM implementation of patient monitors as a step towards building low cost affordable healthcare solutions, using smart phones or FPGAs. However, it may be noted that performance of non-linear Kernel SVM based MPM outperform linear SVM based MPMs. Hence, it would be of great interest to explore on how to enhance the performance of MPM systems using linear SVMs. There exists an intrinsic relationship between vital parameters that is known very well in the medical community, but not so well in the engineering community. In this work, we propose to use correlation features to capture the intrinsic relationship between the vital parameters through the geometric mean of the vital parameters taken in pairs of two, in addition to four vital parameters. The new features are seen to enhance the performance of the linear SVM classifier. We then implement the proposed algorithm in FPGA, to make a low cost implementation of the multi-parameter patient monitor possible.