Narrow Results

In the field of cytopathology, the accurate identification and counting of white blood cells (WBCs) in blood smears is crucial for diagnosing various types of cancer. The process of manually detecting and segmenting these structures, however, can be challenging due to their variable morphologies and the presence of overlapping objects in the images. This makes manual detection time-consuming, labor-intensive, and prone to error, particularly for individuals without extensive experience in cytopathology. In this paper, a deep learning algorithm is developed based on a Mask R-CNN model and driven by a sub-algorithm called KOWN (Keep Only White Blood Cells with Nuclei) for WBC segmentation and counting. The proposed algorithm improves the accuracy of measurements compared to other rapidly growing deep learning works, providing maximum precision in detecting and counting WBCs in both low- and high-blood-cell-density images.

The traditional information extraction technology of dashboard is easily affected by external factors, and the robustness is poor. To improve the safety of the pilot’s performance on the dashboard, this paper proposes a way for extracting the dashboard feature information in eye tracking, which acquires the line of sight point in simulated dashboard. It then uses the Mask R-CNN method to detect the gaze area and then extracts the target feature information. Finally, it fuses two sets of data to get the result of the pilot who extracts the target gaze area in the scene. Experiment results show that the method of new dashboard information extraction proposed in this paper has a better accuracy.

Recent developments in satellite image processing tend to eliminate the need for intensive on-site surveys of urban or rural areas for infrastructure allocation planning. In particular, the detection of buildings in satellite images can significantly aid in rural or urban planning. However, detecting individual buildings in low-resolution satellite images is challenging due to a lack of visual clarity. In order to address this problem, we propose a computer vision-based hybrid framework to detect densely constructed building regions in low resolution satellite images, which can also serve as an assistive framework for automated geo-spatial survey of various sites. Our hybrid framework is comprised of three modules, namely the Mask R-CNN module, an ANN-based refinement module, and a DBSCAN based refinement module. The Mask R-CNN is employed to predict probable clustered building regions in the satellite image, whereas the ANN-based refinement module is applied to remove homogeneous regions (e.g. vegetation area) from the Mask R-CNN-detected clustered building region. Finally, a DBSCAN-based refinement module removes non-congested built-up regions so that densely constructed built-up areas can be identified with better precision. Our experimentation using publicly accessible satellite image datasets (AIRS dataset-Kaggle) establishes the effectiveness of the proposed hybrid framework in identifying densely populated building areas in low-resolution satellite images. The outcomes of the proposed framework demonstrated an approximately 5–10% improvement over various Mask R-CNN and U-Net-based approaches in terms of achieving better precision value of the detection.

Deep learning artificial intelligence (AI) is a booming area in the research field. It allows the development of end-to-end models to predict outcomes based on input data without the need for manual extraction of features. This paper aims for evaluating the automatic crack detection process that is used in identifying the cracks in building structures such as bridges, foundations or other large structures using images. A hybrid approach involving image processing and deep learning algorithms is proposed to detect automatic cracks in structures. As cracks are detected in the images they are segmented using a segmentation process. The proposed deep learning models include a hybrid architecture combining Mask R-CNN with single layer CNN, 3-layer CNN, and8-layer CNN. These models utilizes depth wise convolution with varying dilation rates for efficiently extracting diversified features from the crack images. Further, performance evaluation shows that Mask R-CNN with a single layer CNN achieves an accuracy of 97.5% on a normal dataset and 97.8% on a segmented dataset. The Mask R-CNN with 2-layer convolution resulted in an accuracy of 98.32% on a normal dataset and 98.39% on a segmented dataset. The Mask R-CNN with 8-layers convolution achieves an accuracy of 98.4% on a normal dataset and 98.75% on a segmented dataset. The proposed Mask R-CNN have proved its feasibility in detecting cracks in huge building and structures.

In present clinic practice of otolaryngology, otolaryngologists utilized laryngoscopy to diagnose the larynx lesion of patients preliminarily. Nevertheless, it was challenging for otolaryngologists to interpret the detailed information from laryngoscopy videos comprehensively. In this paper, we proposed Mask R-CNN deep learning algorithm to segment the regions of the vocal folds and glottal from laryngoscopy videos, and self-built algorithm to calculate measured indicators including the length and curvature of vocal folds, the angle of glottal, the area of vocal folds and glottal, and the triangle type composed of vocal folds and glottal. Moreover, in order to provide otolaryngologists critical and immediate medical information during diagnosis, we also provided visualized information, which is labeled on the laryngoscopy images to meet all the needs in clinical practice. From the result of this research, the precision of segmentation has reached a high rate of 90.4% on average. It shows that the model not only achieves great performance in segmentation, but also further proved the indicators are accurate enough to be considered in practical diagnosis. In the future, it is possible for the proposed model to be applied in more kinds of laryngoscopy analyses for more comprehensive diagnosis, which would make a positive influence toward the clinical practice of otolaryngology.

Today, radiologists observe a mammogram to determine whether breast tissue is normal. However, calcifications on the mammogram are so small that sometimes radiologists cannot locate them without a magnified observation to make a judgment. If clusters formed by malignant calcifications are found, the patient should undergo a needle localization surgical biopsy to determine whether the calcification cluster is benign or malignant. However, a needle localization surgical biopsy is an invasive examination. This invasive examination leaves scars, causes pain, and makes the patient feel uncomfortable and unwilling to receive an immediate biopsy, resulting in a delay in treatment time. The researcher cooperated with a medical radiologist to analyze calcification clusters and lesions, employing a mammogram using a multi-architecture deep learning algorithm to solve these problems. The features of the location of the cluster and its benign or malignant status are collected from the needle localization surgical biopsy images and medical order and are used as the target training data in this study. This study adopts the steps of a radiologist examination. First, VGG16 is used to locate calcification clusters on the mammogram, and then the Mask R-CNN model is used to find micro-calcifications in the cluster to remove background interference. Finally, an Inception V3 model is used to analyze whether the calcification cluster is benign or malignant. The prediction precision rates of VGG16, Mask R-CNN, and Inception V3 in this study are 93.63%, 99.76%, and 88.89%, respectively, proving that they can effectively assist radiologists and help patients avoid undergoing a needle localization surgical biopsy.