Experimental validation of learning-based compensation for skull-induced aberrations in transcranial photoacoustic computed tomography

Transcranial photoacoustic computed tomography (PACT) shows promise for neuroimaging, but skull-induced aberrations remain a challenge. Model-based reconstruction methods can compensate for aberrations but are computationally burdensome and require that the skull properties to be known accurately. This study employs ex-vivo experimental data to investigate a learning-based reconstruction method for 3D transcranial PACT that mitigates modeling errors that arise when the skull's acoustic and elastic parameters are not precisely known and reduces reconstruction times from 30 hours to 10 minutes as compared to model-based approaches. This is the first experimental demonstration of a learned image reconstruction method for 3D transcranial PACT.