MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073937 A) filed by N J R Muniraj; S. Saravanasundaram; Deeparani M; T. Sivakumar; S. Shobha Christila; R. Vishwajith; T. Murugeswari; K. Kiruthika; R Gayathri Priyadarshini; S. Afreena; Prashanth S; and M. Sreelakshmi on June 15, 2026, for Breathomics Ai Soc Hierarchical Cnn Accelerator And Feature Extraction Engine On Gpdk 90nm For Voc Based Cancer Sereening.
Inventors include N J R Muniraj; S. Saravana Sundaram; Deeparani M; T Sivakumar; S. Shobha Christila; R. Vishwajith; T. Murugesw Ari; K Kiruthika; R Gayathri Priyadarshini; S Afreena; Prashanth S; and Im. Sreelakshmi.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: Abstract: This invention presents a novel Breathomics AI System-on-Chip (Breathomics-AlSoC) - a full-scale custom Application-Specific Integrated Circuit integrating multi-channel sensor acquisition, statistical feature extraction, and on-chip lD Convolutional Neural Network (CNN) inference for real-time Volatile Organic Compound (VOC) analysis in non-invasive lung and oral cancer screening. The SoC comprises three hierarchical tiers: ( 1) an 8-channell2-bit ADC breath sensor interface with power gating and SPI output, (2) a statistical feature extraction engine computing mean, variance, inter-channel correlation, and temporal slope, and (3) a pipelined lD-CNN accelerator with MAC arrays for direct inference. The complete RTL-to-synthesis flow was executed using Cadence EDA tools targeting GPDK 90nm technology. Post-synthesis results show 1,699 standard cells, synthesized area of 17,687.996 J.liD2 , total power consmnption of 1.098 mW at 100 MHz, and excellent timing closure with +6.53 ns positive slack. Functional verification in Cadence SimVision confirms correct 8-channel ADC acquisition, feature computation, CNN inference, and risk score generation.
Disclaimer: Curated by HT Syndication.
