An Analog Bayesian Classifier Implementation, for Thyroid Disease Detection, based on a Low-Power, Current-Mode Gaussian Function Circuit
Vassilis Alimisis, Georgios Gennis, Christos Dimas, Paul P. Sotiriadis
Abstract
The thyroid gland is a small organ that’s located in the front of the neck, wrapped around the windpipe. Τhyroid releases and controls hormones that help the metabolism work correctly. Metabolism plays a main role in many different systems throughout the human body. Thyroid disorder involves the abnormal production of thyroid hormones. In this regard, if a thyroid disease could be detected, patients could take a specific treatment and greatly reduce the symptoms. This work proposes a novel low power, low voltage (0.6V) analog architecture of a Bayesian classifier for thyroid disease detection. The architecture is based on a new Gaussian function circuit and the Lazzaro Winner-Take-All circuit. The proper operation of the analog classifier is verified using a real-world dataset. The proposed architecture is realized in TSMC 90nm CMOS process and was simulated using the Cadence IC Suite.