Adaptive enhancement of acoustic resolution photoacoustic microscopy imaging via deep CNN prior

Abstract

Acoustic resolution photoacoustic microscopy (AR-PAM) is a promising medical imaging modality that can be employed for deep bio-tissue imaging. However, its relatively low imaging resolution has greatly hindered its wide applications. Previous model-based or learning-based PAM enhancement algorithms either require design of complex handcrafted prior to achieve good performance or lack the interpretability and flexibility that can adapt to different degradation models. However, the degradation model of AR-PAM imaging is subject to both imaging depth and center frequency of ultrasound transducer, which varies in different imaging conditions and cannot be handled by a single neural network model. To address this limitation, an algorithm integrating both learning-based and model-based method is proposed here so that a single framework can deal with various distortion functions adaptively. The vasculature image statistics is implicitly learned by a deep convolutional neural network, which served as plug and play (PnP) prior. The trained network can be directly plugged into the model-based optimization framework for iterative AR-PAM image enhancement, which fitted for different degradation mechanisms. Based on physical model, the point spread function (PSF) kernels for various AR-PAM imaging situations are derived and used for the enhancement of simulation and in vivo AR-PAM images, which collectively proved the effectiveness of proposed method. Quantitatively, the PSNR and SSIM values have all achieve best performance with the proposed algorithm in all three simulation scenarios; The SNR and CNR values have also significantly raised from 6.34 and 5.79 to 35.37 and 29.66 respectively in an in vivo testing result with the proposed algorithm.

Keywords: Acoustic resolution; Deep CNN prior; Deep learning; Image enhancement; Model-based method; Photoacoustic microscopy; Plug and play prior.

Fig. 1

(a) Illustration of plane circular transducer and its field pattern; (b) Illustration of spherical focused transducer and its field pattern at focal plane.

Fig. 2

FFDNet architecture for adaptive image denoising.

Fig. 3

Schematic of workflow to the proposed algorithm.

Fig. 4

Schematic of AR-PAM imaging set up.

Fig. 5

Denoising results of simulated noise images selected in testing dataset. Columns represent Ground truth , simulated noise images, denoised results of FFDNet respectively in terms of noise level $\sigma$ = 10 (a–c), noise level $\sigma$ = 25.5 (d–f), and noise level $\sigma$ = 50 (g i). Scale bar: 2 mm.

Fig. 6

Restoration results of simulated AR-PAM images selected in testing dataset. Columns represent Ground truth , simulated AR images, enhanced results with FDU-Net, total variation, and proposed algorithm, respectively in terms of different lateral resolutions. Scale bar: 2 mm.

Fig. 7

The enhancement of in vivo results from different AR-PAM imaging system: (a) Mouse leg vasculature imaged with the first AR-PAM system; (b) Enhancement result of (a) with FDU-Net; (c) Enhancement result of (a) with total variation; (d) Enhancement result of (a) with proposed algorithm; (e) Mouse ear vasculature image with the second AR-PAM system; (f) Enhancement result of (e) with FDU-Net; (g) Enhancement result of (e) with total variation; (h) Enhancement result of (e) with proposed algorithm. (Scale bar: 1 mm).

Fig. 8

The enhancement of in vivo results from different AR-PAM imaging depths: (a) Mouse vasculature image obtained in near focus region; (b) Enhancement result of (a) with FDU-Net; (c) Enhancement result of (a) with total variation; (d) Enhancement result of (a) with proposed algorithm; (e) Mouse vasculature image obtained in out of focus region; (f) Enhancement result of (e) with FDU-Net; (g) Enhancement result of (e) with total variation; (h) Enhancement result of (e) with proposed algorithm (Scale bar: 2 mm).

Fig. 9

Example AR-PAM image enhancement in different iterations by model based equation (upper row) and FFDNet (bottom row).

Fig. 10

The enhancement of in vivo AR-PAM imaging results with corresponding OR-PAM imaging result as ground truth: (a) AR-PAM imaging result; (b) OR-PAM imaging result; (c) Enhancement result of (a) with FDU-Net; (d) Enhancement result of (a) with total variation algorithm; (e) Enhancement result of (a) with proposed algorithm; (f) signal intensity profiles along the vertical dashed line. (Scale bar: 1 mm).