Precise determination of HER2 subtype is essential for selecting appropriate targeted therapies in breast cancer. However, current HER2 assessment methods remain dependent on invasive tissue biopsies, which are limited by tumor heterogeneity and sampling bias. To address these challenges, this paper proposes a knowledge-guided multi-modal neural network (KMNet) for non-invasive HER2 subtyping by integrating clinical data and ultrasound images. KMNet introduces a Graph-based Clinical Feature encoder (GCF), which constructs a causal graph among clinical indicators based on medical knowledge and extracts high-order feature relationships via the Graph Convolutional Network (GCN). Meanwhile, the Convolutional Neural Network (CNN) and Vision Transformer (ViT)-based hybrid image encoder (CVUIF) captures both local details (calcifications and blood flow) and global dependencies between intra- and peritumoral regions. In addition, the Reduced Dimensional Fusion (RDF) module integrates key information from clinical graph features, ultrasound image features, and structured clinical data to construct a unified multi-modal representation for downstream HER2 subtyping task. Experiments were conducted on the private datasets (HER2USC) and the public datasets (BCW, BCa and SIIM-ISIC). Experimental results demonstrate that KMNet outperformed other reported state-of-the-art multi- modal algorithms in HER2 subtyping task, offering strong potential for clinical decision support in breast cancer treatment.

A shape-based statistical inversion method is proposed for Electrical Impedance Tomography (EIT) and Ultrasound Reflection Tomography (URT) dual-modality imaging. It is promising to improve the imaging accuracy in inclusion detection problems. The proposed image reconstruction method is based on the statistical shape inversion framework. The likelihood function is derived from EIT and URT forward models. The prior distribution is constructed using the Markov random field (MRF) prior. The measurement uncertainty is modeled by conditional error model method. The statistical shape inversion problem is solved by the Maximum a posterior (MAP) method with conventional error model. A set of numerical and experimental tests are carried out to evaluate the performance of the proposed method. The results show that the proposed EIT/URT dual-modality imaging method has obvious improvement in imaging accuracy compared to the traditional single-modality EIT and URT methods.

A statistical shape-constrained reconstruction (SSCR) framework is presented to incorporate the statistical prior information of human lung shapes for lung electrical impedance tomography. The prior information is extracted from 8000 chest-computed tomography scans across 800 patients. The reconstruction framework is implemented with two approaches-a one-step SSCR and an iterative SSCR in lung imaging. The one-step SSCR provides fast and high accurate reconstructions of healthy lungs, whereas the iterative SSCR allows to simultaneously estimate the pre-injured lung and the injury lung part. The approaches are evaluated with the simulated examples of thorax imaging and also with the experimental data from a laboratory setting, with difference imaging considered in both the approaches. It is demonstrated that the accuracy of lung shape reconstruction is significantly improved. In addition, the proposed approaches are proved to be robust against measurement noise, modeling error caused by inaccurately known domain boundary, and the selection of the regularization parameters.