Deep learning in macroscopic diffuse optical imaging.

Jason T Smith,Grant Haskins,Denzel Faulkner,Xavier Intes,Marien Ochoa

doi:10.1117/1.jbo.27.2.020901

Jason T Smith, Grant Haskins + Show 3 more

Open Access

https://doi.org/10.1117/1.jbo.27.2.020901

Copy DOI

Abstract

.SignificanceBiomedical optics system design, image formation, and image analysis have primarily been guided by classical physical modeling and signal processing methodologies. Recently, however, deep learning (DL) has become a major paradigm in computational modeling and has demonstrated utility in numerous scientific domains and various forms of data analysis.AimWe aim to comprehensively review the use of DL applied to macroscopic diffuse optical imaging (DOI).ApproachFirst, we provide a layman introduction to DL. Then, the review summarizes current DL work in some of the most active areas of this field, including optical properties retrieval, fluorescence lifetime imaging, and diffuse optical tomography.ResultsThe advantages of using DL for DOI versus conventional inverse solvers cited in the literature reviewed herein are numerous. These include, among others, a decrease in analysis time (often by many orders of magnitude), increased quantitative reconstruction quality, robustness to noise, and the unique capability to learn complex end-to-end relationships.ConclusionsThe heavily validated capability of DL’s use across a wide range of complex inverse solving methodologies has enormous potential to bring novel DOI modalities, otherwise deemed impractical for clinical translation, to the patient’s bedside.

Highlights

The scientific value of monitoring biological tissues with light was recognized many centuries ago, as reported in published works dating as early as the late 1800s for monitoring brain hemorrhage, 1 as well as the early 1900s for imaging breast cancer[2,3] and performing tissue oximetry.[4]
Smith et al.: Deep learning in macroscopic diffuse optical imaging deep tissues.[7–10]
Smith et al.: Deep learning in macroscopic diffuse optical imaging of reinforcement learning (RL) would be uniquely advantageous relative to the approaches previously discussed.[55]

Summary

Introduction

The scientific value of monitoring biological tissues with light was recognized many centuries ago, as reported in published works dating as early as the late 1800s for monitoring brain hemorrhage, 1 as well as the early 1900s for imaging breast cancer[2,3] and performing tissue oximetry.[4]. Optical imaging techniques have greatly benefited numerous biomedical fields. A wide range of optical techniques provide unique means to probe the functional, physiological, metabolic, and molecular states of deep tissue noninvasively with high sensitivity. As scattering is the predominant phenomenon ruling light propagation in intact biological tissues, the photons harnessed to probe the tissue have typically experienced multiple scattering events (or diffusion); this field can be broadly classified as diffuse optical imaging (DOI). Applications of DOI range from macroscopic extraction of optical properties (OPs), such as absorption and scattering, for further tissue classification and 2D representations,[5,6] to 3D tomographic renderings of the functional chromophores or fluorophore within. Despite the numerous benefits of DOI, its diverse implementations can still be challenging due to the necessity of computational methods that model light propagation and/or the unique contrast mechanism leveraged to be quantitative. Numerous implementations in DOI require a certain level of expertise while being dependent on the optimization of intrinsic parameters of these computational models—limiting their potential for dissemination and, translational impact

Objectives

Results

Conclusion