Search

Patch-level histological tissue classification is an effective pre-processing method for histological slide analysis. However, the classification of tissue with deep learning requires expensive annotation costs. To alleviate the limitations of annotation budgets, the application of active learning (AL) to histological tissue classification is a promising solution. Nevertheless, there is a large imbalance in performance between categories during application, and the tissue corresponding to the categories with relatively insufficient performance are equally important for cancer diagnosis. In this paper, we propose an active learning framework called ICAL, which contains Incorrectness Negative Pre-training (INP) and Category-wise Curriculum Querying (CCQ) to address the above problem from the perspective of category-to-category and from the perspective of categories themselves, respectively. In particular, INP incorporates the unique mechanism of active learning to treat the incorrect prediction results that obtained from CCQ as complementary labels for negative pre-training, in order to better distinguish similar categories during the training process. CCQ adjusts the query weights based on the learning status on each category by the model trained by INP, and utilizes uncertainty to evaluate and compensate for query bias caused by inadequate category performance. Experimental results on two histological tissue classification datasets demonstrate that ICAL achieves performance approaching that of fully supervised learning with less than 16% of the labeled data. In comparison to the state-of-the-art active learning algorithms, ICAL achieved better and more balanced performance in all categories and maintained robustness with extremely low annotation budgets. The source code will be released at https://github.com/LactorHwt/ICAL.

Respiratory diseases pose a massive threat to human health; thus, early diagnosis and treatment are essential. Although electronic stethoscopes have shown effectiveness in enhancing auscultation, the diagnosis still necessitates the expertise of a specialist. In this article, we propose a Swin Transformer utilizing overlap fusion-based generalized S-transform (OFGST-Swin) for respiratory cycle classification. The proposed OFGST-Swin demonstrates the capability to categorize respiratory sounds captured by electronic stethoscopes and detect adventitious respiratory cycles within these recordings, and it consists of two novel modules: the sliding window-based augmentation (SWA) for respiratory cycle data enhancement, and the overlap fusion-based generalized S-transform (OFGST) for respiratory cycle feature extraction. The SWA addresses data imbalance in medical datasets by generating adventitious respiratory cycles through a sliding window. The OFGST incorporates the innovative triangular window-based overlap fusion (TWOF) into the enhanced generalized S-transform (EGST), for extracting respiratory cycle features. The proposed OFGST-Swin has been evaluated on two datasets: the ICBHI 2017 dataset and the SPRsound respiratory sound dataset. The experimental results indicate that the proposed OFGST-Swin achieves a better accuracy score of 0.5605 on four-category classification tasks in the ICBHI 2017 dataset, and 0.8018 on seven-category classification tasks in the SPRsound dataset. The proposed method, serving as a signal processing backend for electronic stethoscopes, offers highly effective diagnostic advice to physicians.

Recently, the Deep Neural Networks (DNNs) have had a large impact on imaging process including medical image segmentation, and the real-valued convolution of DNN has been extensively utilized in multi-modal medical image segmentation to accurately segment lesions via learning data information. However, the weighted summation operation in such convolution limits the ability to maintain spatial dependence that is crucial for identifying different lesion distributions. In this paper, we propose a novel Quaternion Cross-modality Spatial Learning (Q-CSL) which explores the spatial information while considering the linkage between multi-modal images. Specifically, we introduce to quaternion to represent data and coordinates that contain spatial information. Additionally, we propose Quaternion Spatial-association Convolution to learn the spatial information. Subsequently, the proposed De-level Quaternion Cross-modality Fusion (De-QCF) module excavates inner space features and fuses cross-modality spatial dependency. Our experimental results demonstrate that our approach compared to the competitive methods perform well with only 0.01061 M parameters and 9.95G FLOPs.

Pedestrian detection has become an important topic in applications such as automatic driver assistance systems for automobiles and pedestrian tracking in surveillance systems, and many powerful object detectors have been widely used in smart sensing instruments. In realistic scenarios, pedestrians in image data are prone to overlap, and detection of fully bracketed boxes may still tend to be false positives in crowded scenes. In addition, low-level parameters shared among features during detection can cause mutual cancellation, resulting in a pair or set of head-enveloping boxes or body-enveloping boxes returning incorrect results. To address the above problems, we propose a triangular chain closed-loop detection network to improve detection in the case of body overlap. We propose a shared parameter elimination module to eliminate the interaction of shared low-level parameters, which has the advantage of improving the feature representation of occluded pedestrians and increasing feature utilization. Because the head bounding box detection encounters fewer occlusions in the occlusion case, the detection capability is better. Therefore, we propose a bidirectional matching module and a chain linking module to enhance the detection capability of the full bounding box using the head bounding box. These modules can better distinguish pedestrians in our network by focusing on individual region features on the pedestrian body, and then learn more representative pedestrian features by minimizing the vector similarity of the whole body, visible region, and head features in space. Our model has been extensively experimented on two challenging dense pedestrian datasets, CrowdHuman and Citypersons. Compared with the experimental results, our method achieves the best performance, especially on heavily occluded subsets, compared with o other popular existing technical methods. This method achieved good results on the CrowdHuman dataset, with an averaged precision (AP) improvement of 1.19% compared with our baseline CrowdDetection method. Especially, using Faster R-CNN as the framework and incorporating our proposed modules, $\text {MR}^{-2}$ of the reasonable set in the CityPersons dataset was reduced by 0.91.

Response: Pixels with location affinity, which can be also called “pixels of affinity,” have similar semantic information. Group convolution and dilated convolution can utilize them to improve the capability of the model. However, for group convolution, it does not utilize pixels of affinity between layers. For dilated convolution, after multiple convolutions with the same dilated rate, the pixels utilized within each layer do not possess location affinity with each other. To solve the problem of group convolution, our proposed quaternion group convolution uses the quaternion convolution, which promotes the communication between to promote utilizing pixels of affinity between channels. In quaternion group convolution, the feature layers are divided into 4 layers per group, ensuring the quaternion convolution can be performed. To solve the problem of dilated convolution, we propose the quaternion sawtooth wave-like dilated convolutions module (QS module). QS module utilizes quaternion convolution with sawtooth wave-like dilated rates to effectively leverage the pixels that share the location affinity both between and within layers. This allows for an expanded receptive field, ultimately enhancing the performance of the model. In particular, we perform our quaternion group convolution in QS module to design the quaternion group dilated neutral network (QGD-Net). Extensive experiments on Dermoscopic Lesion Segmentation based on ISIC 2016 and ISIC 2017 indicate that our method has significantly reduced the model parameters and highly promoted the precision of the model in Dermoscopic Lesion Segmentation. And our method also shows generalizability in retinal vessel segmentation.