Paper by ECE-NTUA Ph.D. Candidate V. Alimisis, Diploma Students N. P. Eleftheriou and Savvas Leventikidis and Prof. P. P. Sotiriadis received the Best Paper Award (3rd Place) in the IEEE 35th International Conference on Microelectronics 2023
We are pleased to announce that the paper entitled "An Analog Integrated, Low-Power, Area-Efficient, Gilbert, Modulo-based Classifier with Application to Lung-Cancer Classification" received the Best Paper Award (3rd Place) in the IEEE 35th International Conference on Microelectronics 2023 that was held on 17-20 December 2023 in Abu Dhabi, UAE.
The Award-Winning paper was co-authored by Vassilis Alimisis (Ph.D. Candidate, ECE NTUA), Nikolaos P. Eleftheriou (Diploma Student, ECE NTUA), Savvas Leventikidis (Diploma Student, ECE NTUA) and Paul P. Sotiriadis (Professor, ECE NTUA, Fellow IEEE).
Short Abstract: This study presents an alternative approach to develop low-power (744nW) analog classifiers capable of efficiently handling multiple input features while maintaining high levels of accuracy and minimizing power consumption. The proposed classifier relies on Voting and Bayes mathematical models, incorporating Gilbert two-signal four-quadrant multipliers and current comparators. The analog classifier is validated through testing with a real-world lung-cancer surgery dataset, achieving an accuracy of 75.45%. It predicts all testset samples of patients suffering from lung-cancer. Additionally, a comparison with related analog classifiers using the same dataset is conducted. The models are trained via a software-based implementation. The proposed architecture is realized using the TSMC 90nm CMOS process and simulated using the Cadence IC Suite.