Medical image segmentation is crucial for medical image processing and the development of computer-aided diagnostics. In recent years, deep Convolutional Neural Networks (CNNs) have been widely adopted for medical image segmentation and have achieved significant success. UNet, which is based on CNNs, is the mainstream method used for medical image segmentation. However, its performance suffers owing to its inability to capture long-range dependencies. Transformers were initially designed for Natural Language Processing (NLP), and sequence-to-sequence applications have demonstrated the ability to capture long-range dependencies. However, their abilities to acquire local information are limited. Hybrid architectures of CNNs and Transformer, such as TransUNet, have been proposed to benefit from Transformer’s long-range dependencies and CNNs’ low-level details. Nevertheless, automatic medical image segmentation remains a challenging task due to factors such as blurred boundaries, the low-contrast tissue environment, and in the context of ultrasound, issues like speckle noise and attenuation. In this paper, we propose a new model that combines the strengths of both CNNs and Transformer, with network architectural improvements designed to enrich the feature representation captured by the skip connections and the decoder. To this end, we devised a new attention module called Three-Level Attention (TLA). This module is composed of an Attention Gate (AG), channel attention, and spatial normalization mechanism. The AG preserves structural information, whereas channel attention helps to model the interdependencies between channels. Spatial normalization employs the spatial coefficient of the Transformer to improve spatial attention akin to TransNorm. To further improve the skip connection and reduce the semantic gap, skip connections between the encoder and decoder were redesigned in a manner similar to that of the UNet++ dense connection. Moreover, deep supervision using a side-output channel was introduced, analogous to BASNet, which was originally used for saliency predictions. Two datasets from different modalities, a CT scan dataset and an ultrasound dataset, were used to evaluate the proposed UNet architecture. The experimental results showed that our model consistently improved the prediction performance of the UNet across different datasets.

Therapeutic ultrasound waves are the main instruments used in many noninvasive clinical procedures. They are continuously transforming medical treatments through mechanical and thermal effects. To allow for effective and safe delivery of ultrasound waves, numerical modeling methods such as the Finite Difference Method (FDM) and the Finite Element Method (FEM) are used. However, modeling the acoustic wave equation can result in several computational complications. In this work, we study the accuracy of using Physics-Informed Neural Networks (PINNs) to solve the wave equation when applying different combinations of initial and boundary conditions (ICs and BCs) constraints. By exploiting the mesh-free nature of PINNs and their prediction speed, we specifically model the wave equation with a continuous time-dependent point source function. Four main models are designed and studied to monitor the effects of soft or hard constraints on the prediction accuracy and performance. The predicted solutions in all the models were compared to an FDM solution for prediction error estimation. The trials of this work reveal that the wave equation modeled by a PINN with soft IC and BC (soft–soft) constraints reflects the lowest prediction error among the four combinations of constraints.

Deep learning is heavily being borrowed to solve problems in medical imaging applications, and Siamese neural networks are the front runners of motion tracking. In this article, we propose to upgrade one such Siamese architecture-based neural network for robust and accurate landmark tracking in ultrasound images to improve the quality of image-guided radiation therapy. Although several researchers have improved the Siamese architecture-based networks with sophisticated detection modules and by incorporating transfer learning, the inherent assumptions of the constant position model and the missing motion model remain unaddressed limitations. In our proposed model, we overcome these limitations by introducing two modules into the original architecture. We employ a reference template update to resolve the constant position model and a linear Kalman filter (LKF) to address the missing motion model. Moreover, we demonstrate that the proposed architecture provides promising results without transfer learning. The proposed model was submitted to an open challenge organized by MICCAI and was evaluated exhaustively on the Liver US Tracking (CLUST) 2D dataset. Experimental results proved that the proposed model tracked the landmarks with promising accuracy. Furthermore, we also induced synthetic occlusions to perform a qualitative analysis of the proposed approach. The evaluations were performed on the training set of the CLUST 2D dataset. The proposed method outperformed the original Siamese architecture by a significant margin.

Noninvasive imaging of microvascular structures in deep tissues provides morphological and functional information for clinical diagnosis and monitoring. Ultrasound Localization Microscopy (ULM) is an emerging imaging technique that can generate microvascular structures with subwavelength diffraction resolution. However, the clinical utility of ULM is hindered by technical limitations, such as long data acquisition time, high microbubble concentration, and inaccurate localization. In this paper, we propose a Swin transformer-based neural network to perform end-to-end mapping to implement microbubble localization. The performance of the proposed method was validated using synthetic and in vivo data using different quantitative metrics. The results indicate that our proposed network can achieve higher precision and better imaging capability than previously used methods. Furthermore, the computational cost of processing per frame is three to four times faster than traditional methods, which makes the real-time application of this technique feasible in the future.

Block matching techniques have been studied exhaustively for motion estimation in Ultrasound (US) images. Exhaustive Search (ES) is the most commonly used search algorithm for block matching in US images. However, ES can be computationally expensive and slow. In this paper, a faster search algorithm called the Adaptive Rood Pattern Search (ARPS) is adopted to US images along with subpixel matching to reduce the computational cost and enhance block matching. Both ES and ARPS were applied in the context of block matching based 2D speckle tracking and were compared using Number of Computations per Frame (NCF), Computational Time per Frame (CTF) and Root Mean Squared Error (RMSE) as metrics. Our simulations and experimental results proved that ARPS outperformed ES by a substantial margin. Adaptation of this technique could help improve the performance of real-time motion estimation drastically.