Humans' decision making process often relies on utilizing visual information from different views or perspectives. However, in machine-learning-based image classification we typically infer an ...object's class from just a single image showing an object. Especially for challenging classification problems, the visual information conveyed by a single image may be insufficient for an accurate decision. We propose a classification scheme that relies on fusing visual information captured through images depicting the same object from multiple perspectives. Convolutional neural networks are used to extract and encode visual features from the multiple views and we propose strategies for fusing these information. More specifically, we investigate the following three strategies: (1) fusing convolutional feature maps at differing network depths; (2) fusion of bottleneck latent representations prior to classification; and (3) score fusion. We systematically evaluate these strategies on three datasets from different domains. Our findings emphasize the benefit of integrating information fusion into the network rather than performing it by post-processing of classification scores. Furthermore, we demonstrate through a case study that already trained networks can be easily extended by the best fusion strategy, outperforming other approaches by large margin.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Various types of atherosclerotic plaque and varying grades of stenosis could lead to different management of patients with a coronary artery disease. Therefore, it is crucial to detect and classify ...the type of coronary artery plaque, as well as to detect and determine the degree of coronary artery stenosis. This paper includes retrospectively collected clinically obtained coronary CT angiography (CCTA) scans of 163 patients. In these, the centerlines of the coronary arteries were extracted and used to reconstruct multi-planar reformatted (MPR) images for the coronary arteries. To define the reference standard, the presence and the type of plaque in the coronary arteries (no plaque, non-calcified, mixed, calcified), as well as the presence and the anatomical significance of coronary stenosis (no stenosis, non-significant, i.e., <50% luminal narrowing, and significant, i.e., ≥50% luminal narrowing) were manually annotated in the MPR images by identifying the start- and end-points of the segment of the artery affected by the plaque. To perform an automatic analysis, a multi-task recurrent convolutional neural network is applied on coronary artery MPR images. First, a 3D convolutional neural network is utilized to extract features along the coronary artery. Subsequently, the extracted features are aggregated by a recurrent neural network that performs two simultaneous multi-class classification tasks. In the first task, the network detects and characterizes the type of the coronary artery plaque. In the second task, the network detects and determines the anatomical significance of the coronary artery stenosis. The network was trained and tested using the CCTA images of 98 and 65 patients, respectively. For detection and characterization of coronary plaque, the method was achieved an accuracy of 0.77. For detection of stenosis and determination of its anatomical significance, the method was achieved an accuracy of 0.80. The results demonstrate that automatic detection and classification of coronary artery plaque and stenosis are feasible. This may enable automated triage of patients to those without coronary plaque and those with coronary plaque and stenosis in need for further cardiovascular workup.
Sleep apnea syndrome (SAS) is a disorder in which respiratory airflow frequently stops during sleep. Alterations in electroencephalogram (EEG) signal are one of the physiological changes that occur ...during apnea, and can be used to diagnose and monitor sleep apnea events. Herein, we proposed a method to automatically distinguish sleep apnea events using characteristics of EEG signals in order to categorize obstructive sleep apnea (OSA) events, central sleep apnea (CSA) events and normal breathing events. Through the use of an Infinite Impulse Response Butterworth Band pass filter, we divided the EEG signals of C3-A2 and C4-A1 into five sub-bands. Next, we extracted sample entropy and variance of each sub-band. The neighbor composition analysis (NCA) method was utilized for feature selection, and the results are used as input coefficients for classification using random forest, K-nearest neighbor, and support vector machine classifiers. After a 10-fold cross-validation, we found that the average accuracy rate was 88.99%. Specifically, the accuracy of each category, including OSA, CSA and normal breathing were 80.43%, 84.85%, and 95.24%, respectively. The proposed method has great potential in the automatic classification of patients' respiratory events during clinical examinations, and provides a novel idea for the development of an automatic classification system for sleep apnea and normal events without the need for expert intervention.
Objectives
An evaluation of the effectiveness of a new computational system proposed for automatic classification, developed based on a Siamese network combined with Convolutional Neural Networks ...(CNNs), is presented. It aims to identify endodontic technical errors using Cone Beam Computed Tomography (CBCT). The study also aims to compare the performance of the automatic classification system with that of dentists.
Methods
One thousand endodontically treated maxillary molars sagittal and coronal reconstructions were evaluated for the quality of the endodontic treatment and the presence of periapical hypodensities by three board-certified dentists and by an oral and maxillofacial radiologist. The proposed classification system was based on a Siamese network combined with EfficientNet B1 or EfficientNet B7 networks. Accuracy, sensivity, precision, specificity, and F1-score values were calculated for automated artificial systems and dentists. Chi-square tests were performed.
Results
The performances were obtained for EfficienteNet B1, EfficientNet B7 and dentists. Regarding accuracy, sensivity and specificity, the best results were obtained with EfficientNet B1. Concerning precision and F1-score, the best results were obtained with EfficientNet B7. The presence of periapical hypodensity lesions was associated with endodontic technical errors. In contrast, the absence of endodontic technical errors was associated with the absence of hypodensity.
Conclusions
Quality evaluation of the endodontic treatment performed by dentists and by Siamese Network combined with EfficientNet B7 or EfficientNet B1 networks was comparable with a slight superiority for the Siamese Network.
Clinical relevance
CNNs have the potential to be used as a support and standardization tool in assessing endodontic treatment quality in clinical practice.
•Cardiovascular disease is one of the major causes of mortality worldwide.•Traditionally auscultation is a noninvasive, cheap and simple method for diagnosing diseases.•Segmentation step was ...eliminated because a lot of its problems.•The information of PCG signal sequence is obtained by using the curve fitting method.•PCG signal is self-similar and crumpled in murmur’s location so considered as fractal.
Cardiovascular disease is one of the major causes of mortality worldwide. Audio signal produced by the mechanical activity of heart provides useful information about the heart valves operation. To increase discriminability between heart sound signals of different normal and abnormal persons, extraction of appropriate features is so important. An accurate segmentation of heart sound signal requires its corresponding ECG11Electrocardiogram signal. But, acquiring of ECG is generally expensive and time consuming. So, one of the main goals of this paper is to eliminate the segmentation step. In this paper, two feature extraction methods are proposed. In the first proposed method, curve fitting is used to achieve the information contained in the sequence of heart sound signal. In the second method, the powerful features extracted by MFCC22Mel frequency cepstrum coefficients are fused with the fractal features by stacking. The experiments are done on six popular datasets to assess the efficiency of different methods One of the data sets contains four classes and the rest of them include two classes (normal and pathologic). In the classification step, the nearest neighbor classifier with Euclidean distance is used. The proposed method has good performance compared to previous methods such as Filter banks and Wavelet transform. Particularly, the performance of the second method is significantly better than the first proposed method. For three data sets, the overall accuracy of 92%, 81% and 98% are achieved, respectively.
Null variants are prevalent within the human genome, and their accurate interpretation is critical for clinical management. In 2018, the ClinGen Sequence Variant Interpretation (SVI) Working Group ...refined the only criterion with a very strong pathogenicity rating (PVS1). To streamline PVS1 interpretation, we have developed an automatic classification tool with a graphical user interface called AutoPVS1. The performance of AutoPVS1 was assessed using 56 variants manually curated by the ClinGen's SVI Working Group; it achieved an interpretation concordance of 93% (52/56). A further analysis of 28,586 putative loss‐of‐function variants by AutoPVS1 demonstrated that at least 27.7% of them do not reach a very strong strength level, 17.5% because of variant‐specific issues and 10.2% due to disease mechanism considerations. Notably, 41.0% (1,936/4,717) of splicing variants were assigned a decreased preliminary PVS1 strength level, a significantly greater fraction than in frameshift variants (13.2%) and nonsense variants (10.8%). Our results reinforce the necessity of considering variant‐specific issues and disease mechanisms in variant interpretation and demonstrate that AutoPVS1 meets an urgent need by enabling biocurators to easily assign accurate, reliable and reproducible PVS1 strength levels in the process of variant interpretation. AutoPVS1 is publicly available at http://autopvs1.genetics.bgi.com/.
Automatic PVS1 interpretation (AutoPVS1) is an automatic classification tool developed by BGI Genomics to support PVS1 interpretation of null variants. AutoPVS1 is based on the recommendations outlined by ClinGen's SVI Working Group, which consists of two steps. The first step is to address variant specific issues; the second step is to consider disease mechanisms. AutoPVS1 defines the results of the first step as “preliminary PVS1 strength” and the results of the second step as “adjusted PVS1 strength”
Oral Squamous Cell Carcinoma (OSCC) is a common type of cancer of the oral epithelium. Despite their high impact on mortality, sufficient screening methods for early diagnosis of OSCC often lack ...accuracy and thus OSCCs are mostly diagnosed at a late stage. Early detection and accurate outline estimation of OSCCs would lead to a better curative outcome and a reduction in recurrence rates after surgical treatment. Confocal Laser Endomicroscopy (CLE) records sub-surface micro-anatomical images for in vivo cell structure analysis. Recent CLE studies showed great prospects for a reliable, real-time ultrastructural imaging of OSCC in situ. We present and evaluate a novel automatic approach for OSCC diagnosis using deep learning technologies on CLE images. The method is compared against textural feature-based machine learning approaches that represent the current state of the art. For this work, CLE image sequences (7894 images) from patients diagnosed with OSCC were obtained from 4 specific locations in the oral cavity, including the OSCC lesion. The present approach is found to outperform the state of the art in CLE image recognition with an area under the curve (AUC) of 0.96 and a mean accuracy of 88.3% (sensitivity 86.6%, specificity 90%).
The condition monitoring of concrete surface plays a significant role in civil infrastructure management system. Crack is the main threat to concrete surface of buildings, bridges, roads and ...pavements. This issue has been researched for several decades, however, it is still a challenge to classify crack since there are many inferior factors, e.g., intense inhomogeneity, structure complexity and background noise of concrete surface. In this paper, a novel deep-width network (DWN) architecture is used for binary and multi-label concrete surface crack classification without handcraft feature extraction. It intelligently learns cracking structures from input raw images by linear and nonlinear mapping process, flexible dynamically updates new weights and efficiently constructs the network by adding new incremental samples. The presented crack distress classification method is tested on two concrete surface crack image datasets and compared with many popular classification methods like sparse autoencoder (SAE), convolution neural network (CNN), and broad learn system (BLS). Experimental results demonstrate that it obviously outperforms those methods both in accuracy and efficiency.
Surveying threatened and invasive species to obtain accurate population estimates is an important but challenging task that requires a considerable investment in time and resources. Estimates using ...existing ground-based monitoring techniques, such as camera traps and surveys performed on foot, are known to be resource intensive, potentially inaccurate and imprecise, and difficult to validate. Recent developments in unmanned aerial vehicles (UAV), artificial intelligence and miniaturized thermal imaging systems represent a new opportunity for wildlife experts to inexpensively survey relatively large areas. The system presented in this paper includes thermal image acquisition as well as a video processing pipeline to perform object detection, classification and tracking of wildlife in forest or open areas. The system is tested on thermal video data from ground based and test flight footage, and is found to be able to detect all the target wildlife located in the surveyed area. The system is flexible in that the user can readily define the types of objects to classify and the object characteristics that should be considered during classification.