Research by Prof. Paul P. Sotiriadis's Group received the Best Student Paper Award in the IEEE 14th International Conference on Modern Circuits and Systems Technologies (MOCAST) on Electronics and Communications
We are pleased to announce that the paper entitled "A Low-Power Analog Integrated Decision Tree for Diabetic Retinopathy Detection" received Best Student Paper Award in the IEEE 14th International Conference on Modern Circuits and Systems Technologies (MOCAST) on Electronics and Communications, which took place in Dresden, Germany on 11-13 June, 2025.
The Award-Winning paper was co-authored by Vasileios Moustakas (PhD Student, ECE- NTUA and Researcher Archimedes/Athena RC), Vassilis Alimisis (Postdoctoral Researcher, ECE-NTUA and Senior Researcher Archimedes/Athena RC), Zisis Foufas (Graduate, ECE-NTUA), Konstantinos Cheliotis (Diploma Student, ECE- NTUA and Researcher Archimedes/Athena RC), Anna Mylona (Diploma Student, ECE- NTUA and Researcher Archimedes/Athena RC) and Paul P. Sotiriadis (Professor, ECE-NTUA; Lead Researcher, Archimedes/Athena RC; IEEE Fellow).
ABSTRACT: A power-efficient analog integrated decision tree classifier is presented in this work. Circuits that function in the sub-threshold region, including the argmax operator circuit and the Gaussian function circuit, constitute up the high-level architecture. A diabetic retinopathy detection classification task is used to evaluate the proposed classifier. It achieves a mean accuracy of 89.21% and consumes 1645nW. The Cadence IC Suite is used for both schematic and layout design in the implementation and testing of the proposed design, utilizing the TSMC 90nm CMOS process. Both Monte Carlo and corner analysis are used to verify the robustness of the proposed solution. In terms of power and efficiency, post-layout results are also compared with software implementations and existing analog hardware solutions.
