SCH: Edge AI for a Bioimpedance System-on-Chip

Prostate cancer is the second most common cancer diagnosis, which can potentially be cured by surgically removing the prostate. Unfortunately, in one out of five cases, removing the prostate still leaves behind some cancerous cells, requiring additional treatment that often has serious side effects. This project will develop an innovative, intelligent surgical probe that allows the surgeon to check that all cancerous cells are removed during the prostate removal surgery.

Creating this surgical probe will require fundamental research on how to sense and analyze the properties of the prostate tissue with sufficiently low power and small size to fit into the probe. The same underlying technology will be useful beyond prostate cancer treatment, including in (1) preventing injury to nearby nerves during tooth implant surgery; (2) continuously monitoring the brain after it has experienced physical injury; and (3) measuring blood pressure from a smartwatch.

The novel technology that this project will build is an edge artificial intelligence (AI) system-on-chip for bioimpedance analysis that will check the periphery of the tissue being removed, known as surgical margins, for cancerous cells during the surgery. This chip is based on two concepts for applying analog signal processing to bioimpedance analysis: (1) an adaptive filter that removes the baseline from bioimpedance signals; (2) a new type of gated recurrent neural network that is implementable as a small, analog circuit but performs with similar accuracy as a large, discrete-time system.

To evaluate the chip, it will be integrated into a surgical margin assessment probe and tested on how accurately it can distinguish cancerous from healthy tissue in resected prostates (retrieved from men that have undergone prostatectomy as part of their standard of care). The project will make contributions to edge AI by developing a new analog gated recurrent neural network paradigm that is several times more power- and area-efficient than the state-of-the-art.

The project will also make contributions to the field of bioimpedance instrumentation by enabling high frequency bioimpedance analysis in a millimeter-scale form factor, thus providing new capabilities in surgical, implantable, and lab-on-a-chip applications. This project is jointly funded by the Smart Health program (SCH) and the Established Program to Stimulate Competitive Research (EPSCoR).

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.