Thyroid nodules are a common endocrine disorder, and accurate imaging diagnosis plays a crucial role in differentiating benign from malignant lesions. However, conventional image analysis often depends on manual annotations and physicians’ subjective experience, leading to inconsistency and variability in diagnostic outcomes. To address these challenges, this study proposes residual and spatial attention U-net (RSA-UNet), an automated segmentation model for thyroid nodule ultrasound images. The proposed architecture builds upon the U-Net framework. It integrates three key components: a ResNet50 encoder to enhance feature extraction through residual learning, a spatial channel block attention module (SCBAM) to strengthen the network’s focus on salient features, and an atrous spatial pyramid pooling (ASPP) module to capture multi-scale global contextual information. Experimental results demonstrate that RSA-UNet achieves superior segmentation performance, with an accuracy of 96.45%, a mean intersection over union (mIoU) of 85.55%, and a Dice similarity coefficient (DSC) of 91.87%, outperforming various U-Net variants. These results highlight the model’s effectiveness and its potential for clinical application in the automated assessment of thyroid nodules.

Chest X-ray is one of the most common tests in radiology and plays a vital role in helping physicians spot different chest conditions. This paper proposes a new model called CAT-UNet to segment lung masses in chest X-ray images. The CAT-UNet uses TransUNet as the leading architecture and mixes CNN and transformer as an encoder. The CNN uses ResNet50 as the backbone, embedding the coordinate attention (CA) block and four skip connections of different scales to improve the accuracy of finding shallow features. Vision Transformer (ViT), which applied the transformer structure, was used in our method to enhance the feature representation ability of images, and Atrous Spatial Pyramid Pooling (ASPP) was used to adjust the filter’s field-of-view and control the resolution of features. In the decoder, the Convolutional Block Attention Modules (CBAM) are embedded for upsampling so that the segmentation details can be better optimized. To evaluate the performance and generalizability of the proposed method, we conducted a 3-fold cross-validation experiment using 1914 chest X-ray images labeled by radiologists collected from the Department of Radiology at Dalin Tzu Chi Hospital, Taiwan. Experimental results show that the proposed CAT-UNet achieves 89.06% on Dice, 91.62% on sensitivity, 98.64% on specificity, and 95.15% on accuracy, outperforming U-Net, TransUNet, and Swin-UNet encoders.

Cracks on road surfaces pose serious risks to both pedestrians and drivers. Traditional manual crack detection methods are not only slow but also pose safety risks. Automating this process has the potential to greatly enhance detection efficiency and consequently improve driving safety. Although previous methods have shown promise in road crack detection, they often neglect interactions between multiple scales, causing smaller cracks to be overlooked in later stages of detection. This paper introduces the Cross-scale Overlapping Patch-based attention Network (COP-Net), which incorporates two critical components: the Scale-aware Channel Attention (SCA) module and the Patch-based Cross-scale Attention (PCA) module for crack detection. These innovations enable dynamic inference on multiple scales, resulting in a significant improvement in crack detection and segmentation. Notably, our approach excels at detecting both small and large cracks simultaneously. To validate the effectiveness of our approach, we conducted evaluations on three open datasets: CRACK500, CFD, and AEL. These evaluation results demonstrate that COP-Net surpasses eleven comparison methods, including HED, DeepCrack, UHDN, SSGNet, MFANet, FPHBN, DeepCrack, PBNet, PAFNet, CarNet, and SegFormer. Our model achieves new State-of-The-Art (SoTA) performance levels in terms of segmentation metrics such as AIU, ODS, and OIS.

Space requirement for storing indexes and performance for query processing are two critical issues in music information retrieval (MIR) system. To overcome difficulties in variable length of queries and enhance efficiency of music retrieval, we propose an effective and efficient numeric indexing structure. It differs greatly from preexisting researches in textual indexing techniques. We show how the development of this framework has been motivated and demonstrate how the technique may be naturally applied to solve this two fundamental MIR issues. Experiments are performed to compare our method with previous solutions. The results show that our method is more scalable and economical than previous methods. The method we proposed can achieve dramatically and significantly improvement in saving time and storing space for retrieving and indexing

We develop an algorithm for chord decision for monophonic music, and it can be applied to music indexing and automatic composition systems. The basic ideas of the proposed algorithm can be divided into two phases: the local recognition phase and the global decision phase. The algorithm analyzes the music and assigns a set of chord candidates in the local recognition phase, while deciding the best appropriate chord according to the chord progression rules in the global decision phase. In order to improve the perceptual accuracy, we consider the beat of corresponding notes in a measure and eliminate the effect of the passing notes and ornaments.

To extract music features from the raw data of music object and organize them as the music indices is important for music retrieval. A repeating pattern is a series of notes which appear more than once. Most of the repeating patterns are themes or tones for people to remember easily. A figure is defined as a melody contour of a musical segment. A sequence pattern is a melody segment that has the same figure with other melody segments. In this paper, we use the idea of interval between two adjacent notes to form quantized melody contour for representing music objects. This representation differs from existing methods that use notes to form a melody string. We also propose a method to find significant repeating figures. These figures could cover most of the repeating patterns that are in a noted melody string. In addition, the number of repeating figures found in samples of music is less than that of repeating patterns. As a result, the total execution time and memory space can be dramatically decreased.