Search

Ultrasound computed tomography (USCT) is a noninvasive biomedical imaging modality that can reconstruct the speed of sound (SOS) distributions of biological tissues for disease diagnosis. Image reconstruction methods are crucial for USCT and have been receiving increasing attention. However, despite good accuracy and fast reconstruction speed, state-of-the-art methods fail to quantify the uncertainty of reconstructed SOS distributions. Herein, we propose an adaptive edge-enhanced Markov chain Monte Carlo (AEE-MCMC) method for rendering the SOS distribution with uncertainty estimation in USCT. The performance of the proposed method is thoroughly demonstrated with numerical simulations, phantom experiments, and ex vivo experiments. Results show that the proposed method provides SOS distributions with sharper edges compared to the conventional nonstatistical and maximum a posteriori (MAP)-based statistical methods and successfully quantifies the uncertainty of reconstructed SOS distributions. Moreover, the proposed method precisely distinguishes heterogeneous structures of ex vivo organs and can differentiate abnormal tissues from normal tissues. Overall, the proposed AEE-MCMC method offers the opportunity to provide high-resolution reconstruction of SOS distributions and quantify its uncertainty estimation in USCT, which allows for reliable analysis of achieved results and is critical for lesion examination in the clinic.

A steel wire rope (SWR) is widely used in industrial scenarios because of its high strength and toughness. To avoid accidents that local flaws (LFs) can cause, SWRs should be inspected regularly. Magnetic flux leakage (MFL) image detection, which is a common SWR inspection method, is inevitably influenced by shaking noise and strand noise. Although noise-reduction methods for these two types of noise have been proposed, the phenomenon of noise distortion has not received enough attention. In this article, we investigate the noise distortion phenomenon on the MFL image and finds that the distorted noise severely affects the performance of existing noise-feature-oriented denoising methods. To solve this problem, we adopt a target-feature-oriented (TFO) denoising method. Specifically, the removal of noise is avoided; instead, an LF-enhancement-based denoising process is proposed. Moreover, to localize LFs in denoised images, a three-stage adaptive localization method mainly based on disjoint region analysis is proposed. The experimental results show that the proposed TFO denoising method significantly improves the denoising performance for images that distorted noise influences. In addition, the proposed adaptive localization method improves the intelligence of the localization process and has better localization performance for images with distorted noise.

Purpose: Ki67 is a protein associated with tumor proliferation and metastasis in breast cancer and acts as an essential prognostic factor. Clinical work requires recognizing tumor regions on Ki67-stained whole-slide images (WSIs) before quantitation. Deep learning has the potential to provide assistance but largely relies on massive annotations and consumes a huge amount of time and energy. Hence, a novel tumor region recognition approach is proposed for more precise Ki67 quantification. Methods: An unsupervised domain adaptive method is proposed, which combines adversarial and self-training. The model trained on labeled hematoxylin and eosin (H&E) data and unlabeled Ki67 data can recognize tumor regions in Ki67 WSIs. Based on the UDA method, a Ki67 automated assisted quantification system is developed, which contains foreground segmentation, tumor region recognition, cell counting, and WSI-level score calculation. Results: The proposed UDA method achieves high performance in tumor region recognition and Ki67 quantification. The AUC reached 0.9915, 0.9352, and 0.9689 on the validation set and internal and external test sets, respectively, substantially exceeding baseline (0.9334, 0.9167, 0.9408) and rivaling the fully supervised method (0.9950, 0.9284, 0.9652). The evaluation of automated quantification on 148 WSIs illustrated statistical agreement with pathological reports. Conclusion: The model trained by the proposed method is capable of accurately recognizing Ki67 tumor regions. The proposed UDA method can be readily extended to other types of immunohistochemical staining images. The results of automated assisted quantification are accurate and interpretable to provide assistance to both junior and senior pathologists in their interpretation.

Word-embedding acts as one of the backbones of modern natural language processing (NLP). Recently, with the need for deploying NLP models to low-resource devices, there has been a surge of interest to compress word embeddings into hash codes or binary vectors so as to save the storage and memory consumption. Typically, existing work learns to encode an embedding into a compressed representation from which the original embedding can be reconstructed. Although these methods aim to preserve most information of every individual word, they often fail to retain the relation between words, thus can yield large loss on certain tasks. To this end, this paper presents Relation Reconstructive Binarization (R2B) to transform word embeddings into binary codes that can preserve the relation between words. At its heart, R2B trains an auto-encoder to generate binary codes that allow reconstructing the word-by-word relations in the original embedding space. Experiments showed that our method achieved significant improvements over previous methods on a number of tasks along with a space-saving of up to 98.4%. Specifically, our method reached even better results on word similarity evaluation than the uncompressed pre-trained embeddings, and was significantly better than previous compression methods that do not consider word relations.