Student Loss: Towards the Probability Assumption in Inaccurate Supervision.

Noisy labels are often encountered in datasets, but learning with them is challenging. Although natural discrepancies between clean and mislabeled samples in a noisy category exist, most techniques in this field still gather them indiscriminately, which leads to their performances being partially robust. In this paper, we reveal both empirically and theoretically that the learning robustness can be improved by assuming deep features with the same labels follow a student distribution, resulting in a more intuitive method called student loss. By embedding the student distribution and exploiting the sharpness of its curve, our method is naturally data-selective and can offer extra strength to resist mislabeled samples. This ability makes clean samples aggregate tightly in the center, while mislabeled samples scatter, even if they share the same label. Additionally, we employ the metric learning strategy and develop a large-margin student (LT) loss for better capability. It should be noted that our approach is the first work that adopts the prior probability assumption in feature representation to decrease the contributions of mislabeled samples. This strategy can enhance various losses to join the student loss family, even if they have been robust losses. Experiments demonstrate that our approach is more effective in inaccurate supervision. Enhanced LT losses significantly outperform various state-of-the-art methods in most cases. Even huge improvements of over 50% can be obtained under some conditions.

Classification of health deterioration by geometric invariants.

BACKGROUND AND OBJECTIVES: Prediction of patient deterioration is essential in medical care, and its automation may reduce the risk of patient death. The precise monitoring of a patient's medical state requires devices placed on the body, which may cause discomfort. Our approach is based on the processing of long-term ballistocardiography data, which were measured using a sensory pad placed under the patient's mattress. METHODS: The investigated dataset was obtained via long-term measurements in retirement homes and intensive care units (ICU). Data were measured unobtrusively using a measuring pad equipped with piezoceramic sensors. The proposed approach focused on the processing methods of the measured ballistocardiographic signals, Cartan curvature (CC), and Euclidean arc length (EAL). RESULTS: For analysis, 218,979 normal and 216,259 aberrant 2-second samples were collected and classified using a convolutional neural network. Experiments using cross-validation with expert threshold and data length revealed the accuracy, sensitivity, and specificity of the proposed method to be 86.51 CONCLUSIONS: The proposed method provides a unique approach for an early detection of health concerns in an unobtrusive manner. In addition, the suitability of EAL over the CC was determined.

Automated Urine Cell Image Classification Model Using Chaotic Mixer Deep Feature Extraction.

Microscopic examination of urinary sediments is a common laboratory procedure. Automated image-based classification of urinary sediments can reduce analysis time and costs. Inspired by cryptographic mixing protocols and computer vision, we developed an image classification model that combines a novel Arnold Cat Map (ACM)- and fixed-size patch-based mixer algorithm with transfer learning for deep feature extraction. Our study dataset comprised 6,687 urinary sediment images belonging to seven classes: Cast, Crystal, Epithelia, Epithelial nuclei, Erythrocyte, Leukocyte, and Mycete. The developed model consists of four layers: (1) an ACM-based mixer to generate mixed images from resized 224 × 224 input images using fixed-size 16 × 16 patches; (2) DenseNet201 pre-trained on ImageNet1K to extract 1,920 features from each raw input image, and its six corresponding mixed images were concatenated to form a final feature vector of length 13,440; (3) iterative neighborhood component analysis to select the most discriminative feature vector of optimal length 342, determined using a k-nearest neighbor (kNN)-based loss function calculator; and (4) shallow kNN-based classification with ten-fold cross-validation. Our model achieved 98.52% overall accuracy for seven-class classification, outperforming published models for urinary cell and sediment analysis. We demonstrated the feasibility and accuracy of deep feature engineering using an ACM-based mixer algorithm for image preprocessing combined with pre-trained DenseNet201 for feature extraction. The classification model was both demonstrably accurate and computationally lightweight, making it ready for implementation in real-world image-based urine sediment analysis applications.

Automatic seizure detection by convolutional neural networks with computational complexity analysis.

BACKGROUND AND OBJECTIVES: Nowadays, an automated computer-aided diagnosis (CAD) is an approach that plays an important role in the detection of health issues. The main advantages should be in early diagnosis, including high accuracy and low computational complexity without loss of the model performance. One of these systems type is concerned with Electroencephalogram (EEG) signals and seizure detection. We designed a CAD system approach for seizure detection that optimizes the complexity of the required solution while also being reusable on different problems. METHODS: The methodology is built-in deep data analysis for normalization. In comparison to previous research, the system does not necessitate a feature extraction process that optimizes and reduces system complexity. The data classification is provided by a designed 8-layer deep convolutional neural network. RESULTS: Depending on used data, we have achieved the accuracy, specificity, and sensitivity of 98%, 98%, and 98.5% on the short-term Bonn EEG dataset, and 96.99%, 96.89%, and 97.06% on the long-term CHB-MIT EEG dataset. CONCLUSIONS: Through the approach to detection, the system offers an optimized solution for seizure diagnosis health problems. The proposed solution should be implemented in all clinical or home environments for decision support.

Novel favipiravir pattern-based learning model for automated detection of specific language impairment disorder using vowels.

Specific language impairment (SLI) is one of the most common diseases in children, and early diagnosis can help to obtain better timely therapy economically. It is difficult and time-consuming for clinicians to accurately detect SLI through standard clinical assessments. Hence, machine learning algorithms have been developed to assist in the accurate diagnosis of SLI. This work aims to investigate the graph of the favipiravir molecule-based feature extraction function and propose an accurate SLI detection model using vowels. We proposed a novel handcrafted machine learning framework. This architecture comprises the favipiravir molecular structure pattern, statistical feature extractor, wavelet packet decomposition (WPD), iterative neighborhood component analysis (INCA), and support vector machine (SVM) classifier. Two feature extraction models, statistical and textural, are employed in the handcrafted feature generation methodology. A new nature-inspired graph-based feature extractor that uses the chemical depiction of the favipiravir (favipiravir became popular with the COVID-19 pandemic) is employed for feature extraction. Finally, the proposed favipiravir pattern, statistical feature extractor, and wavelet packet decomposition are used to create a feature vector. Moreover, a statistical feature extractor is used in this work. The WPD generates multilevel features, and the most meaningful features are selected using the NCA feature selector. Finally, these chosen features are fed to SVM classifier for automated classification. Two validation methods, (i) leave one subject out (LOSO) and (ii) tenfold cross-validations (CV), are used to obtain robust classification results. Our proposed favipiravir pattern-based model developed using a vowel dataset can detect SLI children with an accuracy of 99.87% and 98.86% using tenfold and LOSO CV strategies, respectively. These results demonstrated the high vowel classification ability of the proposed favipiravir pattern-based model.

Are Graph Convolutional Networks With Random Weights Feasible?

Graph Convolutional Networks (GCNs), as a prominent example of graph neural networks, are receiving extensive attention for their powerful capability in learning node representations on graphs. There are various extensions, either in sampling and/or node feature aggregation, to further improve GCNs' performance, scalability and applicability in various domains. Still, there is room for further improvements on learning efficiency because performing batch gradient descent using the full dataset for every training iteration, as unavoidable for training (vanilla) GCNs, is not a viable option for large graphs. The good potential of random features in speeding up the training phase in large-scale problems motivates us to consider carefully whether GCNs with random weights are feasible. To investigate theoretically and empirically this issue, we propose a novel model termed Graph Convolutional Networks with Random Weights (GCN-RW) by revising the convolutional layer with random filters and simultaneously adjusting the learning objective with regularized least squares loss. Theoretical analyses on the model's approximation upper bound, structure complexity, stability and generalization, are provided with rigorous mathematical proofs. The effectiveness and efficiency of GCN-RW are verified on semi-supervised node classification task with several benchmark datasets. Experimental results demonstrate that, in comparison with some state-of-the-art approaches, GCN-RW can achieve better or matched accuracies with less training time cost.

Unsupervised Monocular Visual Odometry for Fast-Moving Scenes Based on Optical Flow Network with Feature Point Matching Constraint.

Robust and accurate visual feature tracking is essential for good pose estimation in visual odometry. However, in fast-moving scenes, feature point extraction and matching are unstable because of blurred images and large image disparity. In this paper, we propose an unsupervised monocular visual odometry framework based on a fusion of features extracted from two sources, that is, the optical flow network and the traditional point feature extractor. In the training process, point features are generated for scene images and the outliers of matched point pairs are filtered by FlannMatch. Meanwhile, the optical flow network constrained by the principle of forward-backward flow consistency is used to select another group of corresponding point pairs. The Euclidean distance between the matching points found by FlannMatch and the corresponding point pairs by the flow network is added to the loss function of the flow network. Compared with SURF, the trained flow network shows more robust performance in complicated fast-motion scenarios. Furthermore, we propose the AvgFlow estimation module, which selects one group of the matched point pairs generated by the two methods according to the scene motion. The camera pose is then recovered by Perspective-n-Point (PnP) or the epipolar geometry. Experiments conducted on the KITTI Odometry dataset verify the effectiveness of the trajectory estimation of our approach, especially in fast-moving scenarios.

Automated detection of pain levels using deep feature extraction from shutter blinds-based dynamic-sized horizontal patches with facial images.

Pain intensity classification using facial images is a challenging problem in computer vision research. This work proposed a patch and transfer learning-based model to classify various pain intensities using facial images. The input facial images were segmented into dynamic-sized horizontal patches or "shutter blinds". A lightweight deep network DarkNet19 pre-trained on ImageNet1K was used to generate deep features from the shutter blinds and the undivided resized segmented input facial image. The most discriminative features were selected from these deep features using iterative neighborhood component analysis, which were then fed to a standard shallow fine k-nearest neighbor classifier for classification using tenfold cross-validation. The proposed shutter blinds-based model was trained and tested on datasets derived from two public databases-University of Northern British Columbia-McMaster Shoulder Pain Expression Archive Database and Denver Intensity of Spontaneous Facial Action Database-which both comprised four pain intensity classes that had been labeled by human experts using validated facial action coding system methodology. Our shutter blinds-based classification model attained more than 95% overall accuracy rates on both datasets. The excellent performance suggests that the automated pain intensity classification model can be deployed to assist doctors in the non-verbal detection of pain using facial images in various situations (e.g., non-communicative patients or during surgery). This system can facilitate timely detection and management of pain.

Novel nested patch-based feature extraction model for automated Parkinson's Disease symptom classification using MRI images.

OBJECTIVE: Parkinson's disease (PD) is a common neurological disorder with variable clinical manifestations and magnetic resonance imaging (MRI) findings. We propose a handcrafted image classification model that can accurately (i) classify different PD stages, (ii) detect comorbid dementia, and (iii) discriminate PD-related motor symptoms. METHODS: Selected image datasets from three PD studies were used to develop the classification model. Our proposed novel automated system was developed in four phases: (i) texture features are extracted from the non-fixed size patches. In the feature extraction phase, a pyramid histogram-oriented gradient (PHOG) image descriptor is used. (ii) In the feature selection phase, four feature selectors: neighborhood component analysis (NCA), Chi2, minimum redundancy maximum relevancy (mRMR), and ReliefF are used to generate four feature vectors. (iii) Two classifiers: k-nearest neighbor (kNN) and support vector machine (SVM) are used in the classification step. A ten-fold cross-validation technique is used to validate the results. (iv) Eight predicted vectors are generated using four selected feature vectors and two classifiers. Finally, iterative majority voting (IMV) is used to attain general classification results. Therefore, this model is named nested patch-PHOG-multiple feature selectors and multiple classifiers-IMV (NP-PHOG-MFSMCIMV). RESULTS: Our presented NP-PHOG-MFSMCIMV model achieved 99.22, 98.70, and 99.53% accuracies for the collected PD stages, PD dementia, and PD symptoms classification datasets, respectively. SIGNIFICANCE: The obtained accuracies (over 98% for all states) demonstrated the performance of developed NP-PHOG-MFSMCIMV model in automated PD state classification.

A conversational agent for querying Italian Patient Information Leaflets and improving health literacy.

In the last years, the rise of digital technologies has enormously augmented the possibility for people to access health information and consult online versions of Patient Information Leaflets (PILs), enabling them to improve their knowledge about medication and adherence to therapies. However, health information may often be difficult to consult and comprehend due to an excessively lengthy and undersized text, coupled with the presence of many incomprehensible medical terms. To face these issues, this paper proposes a conversational agent as a valuable solution to simplify health information retrieval and improve health literacy in Italian by codifying PILs and making them query-able in natural language. In particular, the system has been devised to: i) comprehend natural language questions on medicines of interest; ii) proactively ask the user or automatically infer from the dialog state all the missing information necessary to generate an answer; iii) extract the answer from a structured knowledge base built from PILs of registered drugs. An experimental study has been carried out to evaluate both the performance and usability of the proposed system. Results showed an adequate ability of the system to handle most of the dialogues started by participants correctly, good users satisfaction, and, thus, proved its feasibility and usefulness.

A Hybrid Lightweight System for Early Attack Detection in the IoMT Fog.

Cyber-attack detection via on-gadget embedded models and cloud systems are widely used for the Internet of Medical Things (IoMT). The former has a limited computation ability, whereas the latter has a long detection time. Fog-based attack detection is alternatively used to overcome these problems. However, the current fog-based systems cannot handle the ever-increasing IoMT's big data. Moreover, they are not lightweight and are designed for network attack detection only. In this work, a hybrid (for host and network) lightweight system is proposed for early attack detection in the IoMT fog. In an adaptive online setting, six different incremental classifiers were implemented, namely a novel Weighted Hoeffding Tree Ensemble (WHTE), Incremental K-Nearest Neighbors (IKNN), Incremental Naïve Bayes (INB), Hoeffding Tree Majority Class (HTMC), Hoeffding Tree Naïve Bayes (HTNB), and Hoeffding Tree Naïve Bayes Adaptive (HTNBA). The system was benchmarked with seven heterogeneous sensors and a NetFlow data infected with nine types of recent attack. The results showed that the proposed system worked well on the lightweight fog devices with ~100% accuracy, a low detection time, and a low memory usage of less than 6 MiB. The single-criteria comparative analysis showed that the WHTE ensemble was more accurate and was less sensitive to the concept drift.

A Novel COVID-19 Data Set and an Effective Deep Learning Approach for the De-Identification of Italian Medical Records.

In the last years, the need to de-identify privacy-sensitive information within Electronic Health Records (EHRs) has become increasingly felt and extremely relevant to encourage the sharing and publication of their content in accordance with the restrictions imposed by both national and supranational privacy authorities. In the field of Natural Language Processing (NLP), several deep learning techniques for Named Entity Recognition (NER) have been applied to face this issue, significantly improving the effectiveness in identifying sensitive information in EHRs written in English. However, the lack of data sets in other languages has strongly limited their applicability and performance evaluation. To this aim, a new de-identification data set in Italian has been developed in this work, starting from the 115 COVID-19 EHRs provided by the Italian Society of Radiology (SIRM): 65 were used for training and development, the remaining 50 were used for testing. The data set was labelled following the guidelines of the i2b2 2014 de-identification track. As additional contribution, combined with the best performing Bi-LSTM + CRF sequence labeling architecture, a stacked word representation form, not yet experimented for the Italian clinical de-identification scenario, has been tested, based both on a contextualized linguistic model to manage word polysemy and its morpho-syntactic variations and on sub-word embeddings to better capture latent syntactic and semantic similarities. Finally, other cutting-edge approaches were compared with the proposed model, which achieved the best performance highlighting the goodness of the promoted approach.

Using artificial intelligence techniques for COVID-19 genome analysis.

The genome of the novel coronavirus (COVID-19) disease was first sequenced in January 2020, approximately a month after its emergence in Wuhan, capital of Hubei province, China. COVID-19 genome sequencing is critical to understanding the virus behavior, its origin, how fast it mutates, and for the development of drugs/vaccines and effective preventive strategies. This paper investigates the use of artificial intelligence techniques to learn interesting information from COVID-19 genome sequences. Sequential pattern mining (SPM) is first applied on a computer-understandable corpus of COVID-19 genome sequences to see if interesting hidden patterns can be found, which reveal frequent patterns of nucleotide bases and their relationships with each other. Second, sequence prediction models are applied to the corpus to evaluate if nucleotide base(s) can be predicted from previous ones. Third, for mutation analysis in genome sequences, an algorithm is designed to find the locations in the genome sequences where the nucleotide bases are changed and to calculate the mutation rate. Obtained results suggest that SPM and mutation analysis techniques can reveal interesting information and patterns in COVID-19 genome sequences to examine the evolution and variations in COVID-19 strains respectively.

Deep learning network for medical volume data segmentation based on multi axial plane fusion.

BACKGROUND AND OBJECTIVE: High-dimensional data generally contains more accurate information for medical image, e.g., computerized tomography (CT) data can depict the three dimensional structure of organs more precisely. However, the data in high-dimension often needs enormous computation and has high memory requirements in the deep learning convolution networks, while dimensional reduction usually leads to performance degradation. METHODS: In this paper, a two-dimensional deep learning segmentation network was proposed for medical volume data based on multi-pinacoidal plane fusion to cover more information under the control of computation.This approach has conducive compatibility while using the model proposed to extract the global information between different inputs layers. RESULTS: Our approach has worked in different backbone network. Using the approach, DeepUnet's Dice coefficient (Dice) and Positive Predictive Value (PPV) are 0.883 and 0.982 showing the satisfied progress. Various backbones can enjoy the profit of the method. CONCLUSIONS: Through the comparison of different backbones, it can be found that the proposed network with multi-pinacoidal plane fusion can achieve better results both quantitively and qualitatively.

Application of mechanical trigger for unobtrusive detection of respiratory disorders from body recoil micro-movements.

Background and Objectives Automatic detection of breathing disorders plays an important role in the early signalization of respiratory diseases. Measuring methods can be based on electrocardiogram (ECG), sound, oximetry, or respiratory analysis. However, these approaches require devices placed on the human body or they are prone to disturbance by environmental influences. To solve these problems, we proposed a heart contraction mechanical trigger for unobtrusive detection of respiratory disorders from the mechanical measurement of cardiac contractions. We designed a novel method to calculate this mechanical trigger purely from measured mechanical signals without the use of ECG. Methods The approach is a built-on calculation of the so-called euclidean arc length from the signals. In comparison to previous researches, this system does not require any equipment attached to a person. This is achieved by locating the tensometers on the bed. Data from sensors are fused by the Cartan curvatures method to beat-to-beat vector input for the Convolutional neural network (CNN) classifier. Results In sum, 2281 disordered and 5130 normal breathing samples was collected for analysis. The experiments with use of 10-fold cross validation show that accuracy, sensitivity, and specificity reach values of 96.37%, 92.46%, and 98.11% respectively. Conclusions By the approach for detection, the system offers a novel way for a completely unobtrusive diagnosis of breathing-related health problems. The proposed solution can effectively be deployed in all clinical or home environments.

CMC: A consensus multi-view clustering model for predicting Alzheimer's disease progression.

Machine learning has been used in the past for the auxiliary diagnosis of Alzheimer's Disease (AD). However, most existing technologies only explore single-view data, require manual parameter setting and focus on two-class (i.e., dementia or not) classification problems. Unlike single-view data, multi-view data provide more powerful feature representation capability. Learning with multi-view data is referred to as multi-view learning, which has received certain attention in recent years. In this paper, we propose a new multi-view clustering model called Consensus Multi-view Clustering (CMC) based on nonnegative matrix factorization for predicting the multiple stages of AD progression. The proposed CMC performs multi-view learning idea to fully capture data features with limited medical images, approaches similarity relations between different entities, addresses the shortcoming from multi-view fusion that requires manual setting parameters, and further acquires a consensus representation containing shared features and complementary knowledge of multiple view data. It not only can improve the predication performance of AD, but also can screen and classify the symptoms of different AD's phases. Experimental results using data with twelve views constructed by brain Magnetic Resonance Imaging (MRI) database from Alzheimer's Disease Neuroimaging Initiative expound and prove the effectiveness of the proposed model.

Crosslingual named entity recognition for clinical de-identification applied to a COVID-19 Italian data set.

The COrona VIrus Disease 19 (COVID-19) pandemic required the work of all global experts to tackle it. Despite the abundance of new studies, privacy laws prevent their dissemination for medical investigations: through clinical de-identification, the Protected Health Information (PHI) contained therein can be anonymized so that medical records can be shared and published. The automation of clinical de-identification through deep learning techniques has proven to be less effective for languages other than English due to the scarcity of data sets. Hence a new Italian de-identification data set has been created from the COVID-19 clinical records made available by the Italian Society of Radiology (SIRM). Therefore, two multi-lingual deep learning systems have been developed for this low-resource language scenario: the objective is to investigate their ability to transfer knowledge between different languages while maintaining the necessary features to correctly perform the Named Entity Recognition task for de-identification. The systems were trained using four different strategies, using both the English Informatics for Integrating Biology & the Bedside (i2b2) 2014 and the new Italian SIRM COVID-19 data sets, then evaluated on the latter. These approaches have demonstrated the effectiveness of cross-lingual transfer learning to de-identify medical records written in a low resource language such as Italian, using one with high resources such as English.

Data-Driven Models for Objective Grading Improvement of Parkinson's Disease.

Parkinson's disease (PD) is a progressive disorder of the central nervous system that causes motor dysfunctions in affected patients. Objective assessment of symptoms can support neurologists in fine evaluations, improving patients' quality of care. Herein, this study aimed to develop data-driven models based on regression algorithms to investigate the potential of kinematic features to predict PD severity levels. Sixty-four patients with PD (PwPD) and 50 healthy subjects of control (HC) were asked to perform 13 motor tasks from the MDS-UPDRS III while wearing wearable inertial sensors. Simultaneously, the clinician provided the evaluation of the tasks based on the MDS-UPDRS scores. One hundred-ninety kinematic features were extracted from the inertial motor data. Data processing and statistical analysis identified a set of parameters able to distinguish between HC and PwPD. Then, multiple feature selection methods allowed selecting the best subset of parameters for obtaining the greatest accuracy when used as input for several predicting regression algorithms. The maximum correlation coefficient, equal to 0.814, was obtained with the adaptive neuro-fuzzy inference system (ANFIS). Therefore, this predictive model could be useful as a decision support system for a reliable objective assessment of PD severity levels based on motion performance, improving patients monitoring over time.