Blood flow recovery from subsampled data in photoacoustic microscopy

Photo acoustic microscopy (PAM) achieves high contrast, intravital imaging of the microvasculature by utilizing the specificity of endogenous optical absorption. PAM has been further augmented by using loss of correlation (LoC) methods to image blood flow. However, estimating blood flow using LoC methods necessitates dense spatial sampling which increases laser fluence. To address concerns over the increase in laser fluence, we develop a sparse modeling algorithm to reconstruct blood flow in PAM from downsampled data. The proposed method is superior to reconstruction by bicubic interpolation, exhibiting an error of 5.6 ± 3.4% in vivo compared to 33.4 ± 32.7% for bicubic interpolation, for data downsampled eight times.