Photoacoustic digital brain and its image reconstruction

Photoacoustic tomography (PAT) is an emerging medical imaging modality which is used in human brain imaging researches in recent years. However, there will be strong acoustic attenuation and aberration in acoustic waves while passing through human skull. In this work, we create 180 numerical human brain phantoms from T1 weighted magnetic resonance imaging (MRI) human brain volumes along with the magnetic resonance angiography (MRA) vascular data. Monte-Carlo based optical simulation and the k-wave based acoustic simulation are conducted on each numerical phantom for acquiring the photoacoustic sinograms, which are used to train a 2D U-net model to correct the acoustic aberration. The evaluation results show that U-net trained with numerical human brain phantoms can effectively alleviate the acoustic aberration and improve the image quality.