Neuromorphic Sensing and Computing by Memristive Devices at the Interface with Biology

The concept of in-Memory Computing is nowadays a well-established principle. Instead of separating the processing and the memory units, as usually done in von Neumann architectures, it proposes new neuromorphic architectures based on memristive devices providing simultaneous storing and computing. Indeed, this invited presentation proposes the new concept of in-Memory Sensing-and-Computing as a disruptive novel approach in neuromorphic computing. This novel concept is based on the fusion of three neuromorphic functions: sensing, computing, and memory. This novel concept is demonstrated by using simple architectures based on a new kind of single-devices simultaneously providing these three functions: the memristive sensors. Moreover, the demonstration of the new concept is provided at the interface with biological systems. In particular, the proposed novel approach is shown by a breakthrough- demonstration in cancer diagnostics: the estimation of the risk of prostate cancer based on simultaneous measures of several cancer biomarkers. So, this invited presentation discusses the first-ever-reported direct computations of cancer risks on cancer biomarkers simultaneously detected with memristive biosensors, meanwhile providing data storing too. The Memristive biosensors are quite a novelty emerged in literature with first publications starting from 2011 only. These new kind of memristive devices are of interest for many reasons but especially for their excellent Limit-of-Detection, which is down to attomolar-ranges in cancer markers detection. They are typically fabricated as silicon nanowires, and manufactured with standard lithographic processes. This means that the whole architecture may be fabricated with the CMOS technology (Complementary Metal–Oxide–Semiconductor) and the biosensors may be directly microfabricated onto the same CMOS dye. Therefore, this talk opens to an important new area of research and development: the possibility to provide simultaneous sensing, computing, and storing with innovative neuromorphic CMOS architectures at the interface with biology.