A Deep Learning-Based Model That Reduces Speed of Sound Aberrations for Improved In Vivo Photoacoustic Imaging.

Seungwan Jeon,Wonseok Choi,Chulhong Kim,Byullee Park

doi:10.1109/tip.2021.3120053

Seungwan Jeon, Wonseok Choi + Show 2 more

Open Access

https://doi.org/10.1109/tip.2021.3120053

Copy DOI

Abstract

Photoacoustic imaging (PAI) has attracted great attention as a medical imaging method. Typically, photoacoustic (PA) images are reconstructed via beamforming, but many factors still hinder the beamforming techniques in reconstructing optimal images in terms of image resolution, imaging depth, or processing speed. Here, we demonstrate a novel deep learning PAI that uses multiple speed of sound (SoS) inputs. With this novel method, we achieved SoS aberration mitigation, streak artifact removal, and temporal resolution improvement all at once in structural and functional in vivo PA images of healthy human limbs and melanoma patients. The presented method produces high-contrast PA images in vivo with reduced distortion, even in adverse conditions where the medium is heterogeneous and/or the data sampling is sparse. Thus, we believe that this new method can achieve high image quality with fast data acquisition and can contribute to the advance of clinical PAI.

Highlights

Photoacoustic imaging (PAI) has become a trending medical imaging technique
We introduced a metric to quantify the difference between two functional PA images reconstructed from 128-/64-ch RF data to confirm that the proposed deep learning processing had little effect on the sO2 values, and to show that this deep neural networks (DNNs) method is compatible with PA functional imaging
With the 600 datasets (270 training datasets and 30 validation datasets reconstructed with the 128- and 64-ch RF data each), we trained three DNNs (e.g., Segnet, U-net, and SegU-net)

Summary

Introduction

Photoacoustic imaging (PAI) has become a trending medical imaging technique. This imaging modality detects ultrasound (US) signals generated through transient thermal expansion after optical absorbers are illuminated by pulsed light. Because two types of hemoglobin in living subjects mainly absorb visible and near-infrared light, high-resolution structural (e.g., total hemoglobin) and functional (e.g., hemoglobin oxygen saturation and blood flow) vascular images can be photoacoustically formed [1,2,3]. Poudel used a joint reconstruction method to simultaneously estimate initial pressure and SoS maps in PACT images [22, 23].

Methods

Results

Discussion

Conclusion