A 65 nm Multi-Modal Bayesian Inference Engine with 16.3 fJ/Sample Calibration-Free GRNG for Risk-Aware At-Home Skin Lesion Screening

2026-06-05

AI Summary

This work addresses the challenges of privacy-preserving and on-device skin lesion screening in uncontrolled home environments by proposing an in-memory computing-based multimodal Bayesian inference engine. It presents the first calibration-free multimodal Bayesian neural network deployed on edge devices, leveraging intra-word Gaussian mixture sampling and complementary process-induced variations to generate Gaussian random numbers. Implemented in 65 nm CMOS technology, the system significantly enhances robustness and uncertainty quantification. Compared to existing unimodal approaches, it achieves a 1.4× improvement in equal-risk operating coverage, over 1.5× greater resilience to data perturbations, a 5.5× increase in tolerance to process variations, a 1.8% gain in balanced accuracy, and an energy efficiency of 16.3 fJ per sample.

Abstract

We present a 65-nm risk-aware multimodal Bayesian inference engine for privacy-preserving, fully on-device skin lesion screening under uncontrolled at-home conditions. The proposed compute-in-memory architecture performs in-word Mixture-of-Gaussian sampling, improving uncertainty modeling beyond conventional unimodal Bayesian neural networks. This added probabilistic expressiveness increases equal-risk operating coverage by 1.4x, improves robustness to user-data perturbations by >1.5x, enhances process-variation resilience by 5.5x, and improves balanced accuracy by 1.8% over state-of-the-art unimodal Bayesian neural networks. Hardware robustness is further supported by calibration-free Gaussian random-number generation using complementary process variation, achieving 16.3 fJ/sample and 168.6 GSa/s/mm^2 efficiency. These results demonstrate a practical, energy-efficient, and risk-aware edge-AI solution for privacy-conscious medical screening.

Problem

• risk-aware
• skin lesion screening
• Bayesian inference
• on-device
• uncertainty modeling

Innovation

• Multi-Modal Bayesian Inference
• Compute-in-Memory
• Calibration-Free GRNG
• Risk-Aware AI
• On-Device Medical Screening