Multitask Deep Learning Reconstruction and Localization of Lesions in Limited Angle Diffuse Optical Tomography

Abstract

Diffuse optical tomography (DOT) leverages near-infrared light propagation through tissue to assess its optical properties and identify abnormalities. DOT image reconstruction is an ill-posed problem due to the highly scattered photons in the medium and the smaller number of measurements compared to the number of unknowns. Limited-angle DOT reduces probe complexity at the cost of increased reconstruction complexity. Reconstructions are thus commonly marred by artifacts and, as a result, it is difficult to obtain an accurate reconstruction of target objects, e.g., malignant lesions. Reconstruction does not always ensure good localization of small lesions. Furthermore, conventional optimization-based reconstruction methods are computationally expensive, rendering them too slow for real-time imaging applications. Our goal is to develop a fast and accurate image reconstruction method using deep learning, where multitask learning ensures accurate lesion localization in addition to improved reconstruction. We apply spatial-wise attention and a distance transform based loss function in a novel multitask learning formulation to improve localization and reconstruction compared to single-task optimized methods. Given the scarcity of realworld sensor-image pairs required for training supervised deep learning models, we leverage physics-based simulation to generate synthetic datasets and use a transfer learning module to align the sensor domain distribution between in silico and real-world data, while taking advantage of cross-domain learning. Applying our method, we find that we can reconstruct and localize lesions faithfully while allowing real-time reconstruction. We also demonstrate that the present algorithm can reconstruct multiple cancer lesions. The results demonstrate that multitask learning provides sharper and more accurate reconstruction.

A self-supervised learning approach for high-resolution diffuse optical tomography using neural fields

“…We compared NeuDOT to two state-of-the-art DOT reconstruction methods: 1) a physics-based method with modified Levenberg-Marquardt algorithm (LM) [13] and 2) a recently proposed CNN-based supervised learning method [14]. In Experiment 1, we assessed NeuDOT's reconstruction performance in the lateral direction using a scattering phantom with two-dimensional planar patterns.…”

Highly robust reconstruction framework for three-dimensional optical imaging based on physical model constrained neural networks

“…Here, the regularization parameters need to be adjusted empirically, leading to artificial subjectivity and non-robustness of the reconstruction results. In recent years, deep learning technology has been applied to the field of 3D OI and exhibits great potential (Gao et al 2018, Li et al 2020, Meng et al 2020, Zhang et al 2021b, Mozumder et al 2022, Yedder et al 2022, Li et al 2023, such as the inverse problem simulation (IPS) (Gao et al 2018), graph convolution networks (GNN) (Li et al 2020), 3D fusion dual-sampling deep neural networks (Li et al 2023), etc. Such datadriven based deep learning methods which have been obtained spread applications in the field of photonics (Li et al 2021, Yun et al 2022, Zhao et al 2022 could learn the relationship between the surface light distribution and internal targets directly from a training dataset, which has the advantage of being fast compared to iterative methods.…”

Periodic-net: an end-to-end data driven framework for diffuse optical imaging of breast cancer from noisy boundary data

“…[5][6][7][8] Applications of ML in optical fields ranging from biotechnology 9 to cancer diagnosis 10 have shown great potential for localization, classifying, and segmentation of tumors for diverse samples in biomedical applications. [11][12][13] Despite the excellent performance of convolution neural network (CNN)-based deep learning reconstruction methods, the locality of the convolution operator makes it difficult to learn global and long-range image information, i.e., reconstructing absorption and scattering coefficient in the same network well. 14 Previously, numerous data-driven algorithms for sources and detectors have been used to reconstruct DOI optical-property images by generating simulation datasets for training.…”