Bringing In-Sensor Intelligence in Radiation Detectors: a Short Review

Technological trends, challenges and solutions recently proposed in the literature to embed intelligence, in particular machine learning, into solid-state sensors and radiation detectors are discussed. Starting from the paradigms adopted in low-power wearable devices for local, real-time processing of medical signals such as System-on-Chip and System-in- Package, we will move to the case of solid-state radiation detectors readout by multichannel application specific integrated circuits (ASIC). Here, special focus is devoted to analog solutions which hold potential for radical reduction of power dissipation and interconnection lines (and data throughput) between the detector head and the acquisition back-end. We present the concept of a CMOS analog implementation of an artificial neural network, fusible with the detector analog front-end circuit, for instance for gamma cameras and we compare it with novel ADC-less approaches emerging in the field.