ANP-OT: A 0.17nW/Synapse 0.46pJ/SOP Neuromorphic Olfactory Processor with On-Chip Transfer Learning for Non-Invasive Cross-Hospital Cross-Pulmonary-Disease Diagnosis

📋 论文概要

该论文提出了一种用于非侵入式跨医院肺病诊断的神经形态嗅觉处理器，通过片上迁移学习实现了超过94%的检测准确率，并达到了0.46pJ/SOP的能量效率和0.17nW/突触的功率密度。

💡 主要创新点

• 首次在神经形态嗅觉处理器中集成片上迁移学习，支持跨医院和跨肺病类型的诊断。
• 实现了极低的功耗（0.46pJ/SOP和0.17nW/突触），适用于可穿戴或便携式医疗设备。

🏷 关键词

📄 原文摘要

Shijiazhuang, China National Tsing Hua University, Hsinchu, Taiwan *Equally Credited Authors (ECAs) 1 4 Abstract This paper reports an olfactory processor with on-chip transfer learning for cross-hospital and cross-pulmonary-disease diagnosis, which achieves >94% pulmonary disease detection accuracy at 3 different hospitals. This work achieves a 0.46pJ/SOP energy efficiency, and a power density of 0.17nW/synapse. Many pulmonary diseases have subtle or easily missed symptoms in their early stages [1]. However, the breath composition of early-stage pulmonary disease patients differs from that of healthy individuals [1]. Such differences lead to variations in the response of a multichannel gas sensor array, generating measurable data deviations. These deviations can be

👥 作者与机构

Dexuan Huo*1, Ziyi Cheng*1, Jilin Zhang1, Yumeng Jiang2, Lushuo Zhang3, Hui Wang3, Na Ma2, Zebin Huang2, Minggui Lin2, Yunxia Zhao3, Zhihua Wang1,

Kea-Tiong Tang4, Hong Chen1 Tsinghua University, Beijing, China, 2Beijing Tsinghua Changgung Hospital, Beijing, China, 3Hebei Medical University Third Hospital,