Dale Robin, Ross Nicholas, Howard Scott, O'Sullivan Thomas D, Dehghani Hamid
University of Birmingham, Medical Imaging Lab, School of Computer Science, University Rd W, Birmingham, B15 2TT, UK.
University of Notre Dame, Department of Electrical Engineering and Bioengineering Program, 275 Fitzpatrick Hall, Notre Dame, Indiana, 46556, USA.
Biomed Opt Express. 2025 Mar 26;16(4):1582-1601. doi: 10.1364/BOE.549880. eCollection 2025 Apr 1.
Diffuse optical tomography (DOT) performed using deep-learning allows high-speed reconstruction of tissue optical properties and could thereby enable image-guided scanning, e.g., to enhance clinical breast imaging. Previously published models are geometry-specific and, therefore, require extensive data generation and training for each use case, restricting the scanning protocol at the point of use. A transformer-based architecture is proposed to overcome these obstacles that encode spatially unstructured DOT measurements, enabling a single trained model to handle arbitrary scanning pathways and measurement density. The model is demonstrated with breast tissue-emulating simulated and phantom data, yielding - for 24 mm-deep absorptions ( ) and reduced scattering ( ') images, respectively - average RMSEs of 0.0095±0.0023 cm and 1.95±0.78 cm, Sørensen-Dice coefficients of 0.55±0.12 and 0.67±0.1, and anomaly contrast of 79±10% and 93.3±4.6% of the ground-truth contrast, with an effective imaging speed of 14 Hz. The average absolute and ' values of homogeneous simulated examples were within 10% of the true values.
使用深度学习进行的扩散光学层析成像（DOT）能够高速重建组织光学特性，从而实现图像引导扫描，例如增强临床乳腺成像。先前发表的模型是特定于几何形状的，因此，针对每个用例都需要大量的数据生成和训练，这在使用时限制了扫描协议。本文提出了一种基于Transformer的架构来克服这些障碍，该架构对空间非结构化的DOT测量进行编码，使单个训练模型能够处理任意扫描路径和测量密度。该模型通过乳腺组织模拟数据和体模数据进行了验证，对于深度为24mm的吸收（ ）和减少散射（ '）图像，平均均方根误差（RMSE）分别为0.0095±0.0023cm和1.95±0.78cm，索伦森-迪赛系数（Sørensen-Dice coefficient）为0.55±0.12和0.67±0.1，异常对比度为真实对比度的79±10%和93.3±4.6%，有效成像速度为14Hz。均匀模拟示例的平均绝对 和 '值在真实值的10%以内。
