A noise robust decoder

About

Background: Stanford researchers have invented a decoder for multiplexed readouts of imaging arrays that optimizes the signal-to-noise ratio (SNR) of the decoded detector pixel signals. It uses maximum likelihood estimation, which is referred to as “maximum likelihood CS” (ML-CS) decoding. For noisy imaging applications, it can improve the signal-to-noise ratio (SNR) performance of multiplexed readouts for imaging arrays with a practical, affordable implementation. This invention can reduce the cost of an imaging sensor by reducing the number of readout channels. It is applicable to a wide range of imaging applications, ranging from medical imaging to digital cameras. Stage of Research: - For positron emission tomography (PET), simulations showed that the invention can improve the SNR of the decoded signal by 3-4 times over compressed sensing techniques on compressed sensing multiplexing topologies.  - The decoder can also be applied to conventional multiplexing topologies to provide a 50% decoded SNR improvement over conventional multiplexing decoders. Applications: Medical imaging applications such as planar imaging by X-rays or nuclear medicine, X-ray CT, MRI, PET, and SPECT High-speed optical imaging with digital cameras Low-light optical imaging with digital cameras Time-of-flight imaging for medical and non-medical applications Advantages: More robust to noise that previous methods in practice Improves the signal-to-noise ratio (SNR) performance Computationally feasible to implement Lowers cost of imaging by enabling multiplexing under noisy conditions Multiplexing reduces the number of readout channels and can improve yield by allowing an imaging sensor to be used with a few "bad" pixels