Background and Motivation: COVID-19 has resulted in a massive loss of life during the last two years. The current imaging-based diagnostic methods for COVID-19 detection in multiclass pneumonia-type chest X-rays are not so successful in clinical practice due to high error rates. Our hypothesis states that if we can have a segmentation-based classification error rate <5%, typically adopted for 510 (K) regulatory purposes, the diagnostic system can be adapted in clinical settings. Method: This study proposes 16 types of segmentation-based classification deep learning-based systems for automatic, rapid, and precise detection of COVID-19. The two deep learning-based segmentation networks, namely UNet and UNet+, along with eight classification models, namely VGG16, VGG19, Xception, InceptionV3, Densenet201, NASNetMobile, Resnet50, and MobileNet, were applied to select the best-suited combination of networks. Using the cross-entropy loss function, the system performance was evaluated by Dice, Jaccard, area-under-the-curve (AUC), and receiver operating characteristics (ROC) and validated using Grad-CAM in explainable AI framework. Results: The best performing segmentation model was UNet, which exhibited the accuracy, loss, Dice, Jaccard, and AUC of 96.35%, 0.15%, 94.88%, 90.38%, and 0.99 (p-value <0.0001), respectively. The best performing segmentation-based classification model was UNet+Xception, which exhibited the accuracy, precision, recall, F1-score, and AUC of 97.45%, 97.46%, 97.45%, 97.43%, and 0.998 (p-value <0.0001), respectively. Our system outperformed existing methods for segmentation-based classification models. The mean improvement of the UNet+Xception system over all the remaining studies was 8.27%. Conclusion: The segmentation-based classification is a viable option as the hypothesis (error rate <5%) holds true and is thus adaptable in clinical practice.

/pdf/segmentation-based-classification-deep-learning-model-thx6k6su.pdf

Segmentation-Based Classification Deep Learning Model Embedded with Explainable AI for COVID-19 Detection in Chest X-ray Scans

Biometrics is a powerful tool for identifying and authenticating persons based on their unique characteristics. Finger vein (FV) seems to be an emerging biometric of all types of hand-based biometrics, which have garnered considerable interest because of the extensive information and ease of implementation. As the FV system has grown in popularity, there have been numerous attempts to compromise it. Recent research reveals that the finger-vein recognition (FVR) system is vulnerable to presentation attacks, in which the sensory device accepts a fake printed FV image and grants access as if it were a genuine attempt. Few deep learning (DL) studies were developed in the past to detect spoof attacks in FV images. Existing works rely on performance improvement by neglecting the problems of the limited dataset, high computational complexity, and unavailability of lightweight and efficient feature descriptors. Therefore, we have proposed a novel depthwise separable convolution neural network (DSC) with residual connection and a linear support vector machine (LSVM) for the automatic detection of FV presentation attacks. At first, we apply the data augmentation method to raw images to increase the size of datasets. Afterward, this DSC technique is applied to extract robust features from FV images. Finally, a LSVM classifier is used to classify the images into bonafide and fake images. The proposed DSC-LSVM method is evaluated on two publicly available datasets such as Idiap and SCUT-FVD. The experimental results show that the DSC-SVM model attained a low error rate of 0.00% for presentation attack detection (PAD) on both datasets compared to state-of-the-art approaches. Several experimental results show that the DSC-LSVM model required less computation time and fewer parameters to detect spoof finger vein images. It concludes that the DSC-LSVM system is outperformed for PAD on finger vein images compared to existing methods. 

Finger-vein presentation attack detection using depthwise separable convolution neural network

Machine Learning Methods for Diagnosis of Eye-Related Diseases: A Systematic Review Study Based on Ophthalmic Imaging Modalities

Fabric defect detection based on separate convolutional UNet

In today’s information age, how to accurately identify a person’s identity and protect information security has become a hot topic of people from all walks of life. At present, a more convenient and secure solution to identity identification is undoubtedly biometric identification, but a single biometric identification cannot support increasingly complex and diversified authentication scenarios. Using multimodal biometric technology can improve the accuracy and safety of identification. This paper proposes a biometric method based on finger vein and face bimodal feature layer fusion, which uses a convolutional neural network (CNN), and the fusion occurs in the feature layer. The self-attention mechanism is used to obtain the weights of the two biometrics, and combined with the RESNET residual structure, the self-attention weight feature is cascaded with the bimodal fusion feature channel Concat. To prove the high efficiency of bimodal feature layer fusion, AlexNet and VGG-19 network models were selected in the experimental part for extracting finger vein and face image features as inputs to the feature fusion module. The extensive experiments show that the recognition accuracy of both models exceeds 98.4%, demonstrating the high efficiency of the bimodal feature fusion.

https://www.mdpi.com/1424-8220/22/16/6039/pdf?version=1660534625

Convolutional Neural Network Approach Based on Multimodal Biometric System with Fusion of Face and Finger Vein Features

Finger vein recognition received special attention among all other biometric traits due to its high security. Adequate recognition and classification accuracy ensure the security of personal authentication. Many convolutional neural networks (CNNs) have been proposed with a promising performance in biometric finger vein recognition. However, their architectures have several problems, such as high complexity, extraction of robust features, degraded performance, etc. Considering the issues of CNNs, the authors present a pre-trained CNN network named Xception model based on depth-wise separable CNNs with residual connection, which is considered to be a more effective, less complex neural network to extract robust features. Our work can be seen as a three-stage process. Initially, the concept of data pre-processing is applied to convert the raw input samples into the standard format. Afterward, data augmentation using different geometrical techniques is incorporated to overcome the lack of training samples required for training the deep learning model. Finally, the feature extraction and classification task is performed through the pre-trained Xception architecture to verify the person's identity. SDUMLA and THU-FVFDT2 datasets are utilized to test and evaluate the proposed multi-layered CNN model performance with existing arts. Our proposed method for the SDUMLA database achieved an accuracy of 99% with an F1-score of 98%. While on THU-FVFDT2, the proposed method obtained an accuracy of 90% with an F1-score of 88%. Experimental results conclude that the proposed work obtained excellent performance compared to existing methods.

DS-CNN: A Pre-trained Xception Model based on Depth-wise Separable Convolutional Neural Network for Finger Vein Recognition

Recent advancements in finger vein recognition technology: Methodology, challenges and opportunities

Biometric deals with the verification and identification of a person based on behavioural and physiological traits. This article presents recent advances in physiological-based biometric multimodalities, where we focused on finger vein, palm vein, fingerprint, face, lips, iris, and retina-based processing methods. The authors also evaluated the architecture, operational mode, and performance metrics of biometric technology. In this article, the authors summarize and study various traditional and deep learning-based physiological-based biometric modalities. An extensive review of biometric steps of multiple modalities by using different levels such as preprocessing, feature extraction, and classification, are presented in detail. Challenges and future trends of existing conventional and deep learning approaches are explained in detail to help the researcher. Moreover, traditional and deep learning methods of various physiological-based biometric systems are roughly analyzed to evaluate them. The comparison result and discussion section of this article indicate that there is still a need to develop a robust physiological-based method to advance and improve the performance of the biometric system.

Xingming Zhang

Papers

DS-CNN: A Pre-trained Xception Model based on Depth-wise Separable Convolutional Neural Network for Finger Vein Recognition

Recent advancements in finger vein recognition technology: Methodology, challenges and opportunities

A Systematic Review on Physiological-Based Biometric Recognition Systems: Current and Future Trends