Full-view in vivo skin and blood vessels profile segmentation in photoacoustic imaging based on deep learning

Abstract

Photoacoustic (PA) microscopy allows imaging of the soft biological tissue based on optical absorption contrast and spatial ultrasound resolution. One of the major applications of PA imaging is its characterization of microvasculature. However, the strong PA signal from skin layer overshadowed the subcutaneous blood vessels leading to indirectly reconstruct the PA images in human study. Addressing the present situation, we examined a deep learning (DL) automatic algorithm to achieve high-resolution and high-contrast segmentation for widening PA imaging applications. In this research, we propose a DL model based on modified U-Net for extracting the relationship features between amplitudes of the generated PA signal from skin and underlying vessels. This study illustrates the broader potential of hybrid complex network as an automatic segmentation tool for the in vivo PA imaging. With DL-infused solution, our result outperforms the previous studies with achieved real-time semantic segmentation on large-size high-resolution PA images.

Deep-Learning-Based Algorithm for the Removal of Electromagnetic Interference Noise in Photoacoustic Endoscopic Image Processing

“…We evaluated their performance and effect in terms of reconstruction quality assessments such as RMSE, SSIM and MAE. Although these CNNs have already been applied to previous PA image segmentation and regression-related research [ 40 , 41 , 42 , 43 , 44 ], our work is the first to apply the architectures to remove EMI noise from PAE images, and more specifically from OR-PAE images. Although CNN-based algorithms have shown outstanding performance in white Gaussian noise removal [ 76 ] or in impulse noise removal [ 77 ], it was not known whether they have the ability to remove other types of noise, especially a type of noise as peculiar as EMI noise.…”

“…In this study, we propose a deep-learning-based EMI noise removal algorithm for use on PA images acquired by a newly constructed PAE system [ 29 ]. Although multiple studies have applied artificial intelligence (AI) techniques to PAT, all of them were related to other topics, such as PA image classification [ 30 , 31 ], reverberation removal [ 32 ], missing data restoration [ 33 , 34 ], artifact removal [ 35 , 36 , 37 ], reconstruction assistance [ 38 , 39 ], image segmentation [ 40 , 41 , 42 ], and resolution enhancement [ 43 , 44 ]; more details on the previous works in this area are provided in Supplementary Table S1 . To the best of our knowledge, no previous studies have addressed the problem of EMI noise removal as our work does.…”

“…Recent studies in the biomedical imaging community have shown the applicability of deep-learning techniques to solve such tedious and complex image-to-image regression problems in super-resolution [ 59 , 60 ], segmentation [ 40 , 41 , 42 , 61 , 62 ], and denoising [ 63 , 64 ] from a variety of imaging modalities. The type of deep neural network commonly used for handling image-to-image regression is a convolutional neural network (CNN) with convolutional layers end-to-end.…”

Semantic segmentation of multispectral photoacoustic images using deep learning

“…Several groups have already worked on methods for automatic image segmentation in PAI, for example for the automatic identification of structures in small animal images [7] , [8] , [9] , the segmentation of breast cancer [10] , or for vessel segmentation both in simulation studies [11] and experimental settings [12] , [13] . Furthermore, work has been conducted towards the annotation of different skin layers in raster-scanned images [14] , [15] , [16] . However, to our knowledge, no work has been published to date on the automatic multi-label semantic annotation of multispectral PA images in humans.…”

Design of Metaheuristic Optimization‐Based Vascular Segmentation Techniques for Photoacoustic Images

“…Photoacoustic tomography (PAT) is a kind of hybrid imaging approach which integrates the advantages of both ultrasonic and optical imaging. In PAT, ultrasonic waves are stimulated by the pulsed laser that has embodied both ultrasonic deep penetration and optical absorption contrast [ 1 ]. Various real-time applications are being studied to demonstrate their great opportunities in both clinical and preclinical imaging, such as breast cancer diagnostics and small animal entire body imaging [ 2 ].…”