An Analog Classifier for Liver Disorders based on Centroid Machine Learning Model

This work presents a pure analog integrated implementation of a centroid-based classifier with application in liver disorders. Liver disorder refers to any condition affecting the normal functioning of the liver, encompassing a wide range of diseases and conditions that impact liver health. The high-level architecture consists of Euclidean distance circuits and a current comparator circuit. To achieve low-power operation all transistors operate in the sub-threshold region. Additionally, it achieves 95.65% classification accuracy and consumes 121nW. Also, it is robust and provides high accuracy over Corners and Monte Carlo analysis. Moreover, a comparison study is provided both with the software-equivalent and with other classifiers. The architecture was designed and simulated in the TSMC 90nm CMOS process using the Cadence IC Suite.