Nacer Eddine Benzebouchi

Reliable evaluation of artificial intelligence models for coronavirus disease 2019 (COVID-19) imaging remains challenging because patient-level dependencies, source-related biases, and differences between imaging modalities can lead to overly optimistic performance estimates. This study proposes a leakage-controlled, modality-aware deep ensemble framework for COVID-19 classification using chest X-ray (CXR) and computed tomography (CT) images from the COVID-19 Radiography Database and the SARS-CoV-2 CT-Scan Dataset. Separate evaluation protocols were established for CXR and CT, with duplicate control and available source information incorporated for CXR and patient-level partitioning applied to CT before data augmentation. Four ImageNet-pretrained convolutional neural networks (CNNs)—VGG16, ResNet50, DenseNet169, and InceptionV3—were evaluated, followed by temperature scaling for probability calibration and validation-optimized weighted soft voting for ensemble fusion. The original CT architectural comparison was conducted at the slice level, whereas the final CT evaluation was performed at the patient level. In the final modality-specific held-out evaluation, DenseNet169 achieved an area under the receiver operating characteristic curve (AUROC) of 0.985, an accuracy of 97.32%, and an F1-score of 98.1% on the 2870-image CXR test set. The calibrated weighted ensemble achieved an AUROC of 0.990, an accuracy of 98.43%, and an F1-score of 97.9%, with a Brier score of 0.030 and an expected calibration error (ECE) of 0.023. For CT, DenseNet169 achieved an AUROC of 0.941 and correctly classified 16 of 18 patients (88.89%), whereas the calibrated weighted ensemble achieved an AUROC of 0.969 and correctly classified 17 of 18 patients (94.44%), with a Brier score of 0.070 and an ECE of 0.053. An uncertainty-aware selective-prediction analysis identified 316 of 2870 CXR test images (11.01%) and 2 of 18 CT test patients (11.11%) as low-confidence cases for specialist review. Among the remaining CT cases, 15 of 16 patients were correctly classified (93.75%). Because the CT test cohort contained only 18 patients, the patient-level CT performance and selective-prediction results should be interpreted cautiously as descriptive estimates from this limited cohort. The original pooled evaluation, in which DenseNet169 achieved 98.51% accuracy, is retained only as an exploratory architectural comparison and should not be interpreted as directly comparable with the final modality-specific evaluation. The findings demonstrate that evaluation design, statistical unit, probability calibration, and ensemble fusion materially influence the interpretation of COVID-19 imaging-classification performance. The proposed framework provides leakage-aware and calibrated performance estimates under the evaluated datasets; however, independent multicenter external validation is required before broader claims of clinical generalizability or deployment readiness can be made.

Diabetes is characterized by an abnormally enhanced concentration of glucose in the blood serum. It has a damaging impact on several noble body systems. Today, the concept of unbalanced learning has developed considerably in the domain of medical diagnosis, which greatly reduces the generation of erroneous classification results. The paper takes a hybrid approach to imbalanced learning by proposing an enhanced multimodal meta-learning method called IRESAMPLE+St to distinguish between normal and diabetic patients. This approach relies on the Stacking paradigm by utilizing the complementarity that may exist between classifiers. In the same focus of this study, a modified RESAMPLE-based technique referred to as IRESAMPLE+ and the SMOTE method are integrated as a preliminary resampling step to overcome and resolve the problem of unbalanced data. The suggested IRESAMPLE+St provides a computerized diabetes diagnostic system with impressive results, comparing it to the principal related studies, reflecting the design and engineering successes achieved. Request access from your librarian to read this article's full text.

Breast cancer is the most feared disease in the female population. Early detection plays an important role to improve prognosis. Mammography is the best examination for the detection of breast cancer. However, in some cases, reading mammograms is difficult for radiologists. For this reason, several researches have been conducted to develop Computer Aided Diagnosis tools (CAD) for this disease which aims to interpret mammography images. This paper investigates a new CAD system based on Transductive scheme and Mutual Information for breast abnormalities diagnosis. In the proposed method, a feature vector contains a combination of two features extraction method: Grey Level Co-occurrence Matrix and local Binary Pattern. In the next step, a novel scheme combining Mutual Information and Correlation-based feature selection was applied for selecting the most relevant features. Finally, the classification was achieved using a Transductive Support Vector Machine classifier. The effectiveness of the proposed CAD is examined on the DDSM dataset using classification accuracy, recall and precision. Experimental results demonstrate that the proposed CAD system is clinically significant and can be used to classify the abnormalities of the breast.

Glaucoma is a major public health problem that can lead to an optic nerve lesion, requiring systematic screening in the population over 45 years of age. The diagnosis and classification of this disease have had a marked and excellent development in recent years, particularly in the machine learning domain. Multimodal data have been shown to be a significant aid to the machine learning domain, especially by its contribution to improving data driven decision-making. Solving classification problems by combinations of classifiers has made it possible to increase the robustness as well as the classification reliability by using the complementarity that may exist between the classifiers. Complementarity is considered a key property of multimodality. A Convolutional Neural Network (CNN) works very well in pattern recognition and has been shown to exhibit superior performance, especially for image classification which can learn by themselves useful features from raw data. This article proposes a multimodal classification approach based on deep Convolutional Neural Network and Support Vector Machine (SVM) classifiers using multimodal data and multimodal feature for glaucoma diagnosis from retinal fundus images from RIM-ONE dataset. We make use of handcrafted feature descriptors such as the Gray Level Co-Occurrence Matrix, Central Moments and Hu Moments to co-operate with features automatically generated by the CNN in order to properly detect the optic nerve and consequently obtain a better classification rate, allowing a more reliable diagnosis of glaucoma. The experimental results confirm that the combination of classifiers using a new hybrid fusion approach and the BWWV technique is better than learning classifiers separately. The proposed method provides a computerized diagnosis system for glaucoma disease with impressive results comparing them to the main related studies that allow us to continue in this research path.

Text mining is one of the main and typical tasks of machine learning (ML). Authorship identification (AI) is a standard research subject in text mining and natural language processing (NLP) that has undergone a remarkable evolution these last years. We need to identify/determine the actual author of anonymous texts given on the basis of a set of writing samples. Standard text classification often focuses on many handcrafted features such as dictionaries, knowledge bases, and different stylometric characteristics, which often leads to remarkable dimensionality. Unlike traditional approaches, this paper suggests an authorship identification approach based on automatic feature engineering using word2vec word embeddings, taking into account each author's writing style. This system includes two learning phases, the first stage aims to generate the semantic representation of each author by using word2vec to learn and extract the most relevant characteristics of the raw document. The second stage is to apply the multilayer perceptron (MLP) classifier to fix the classification rules using the backpropagation learning algorithm. Experiments show that MLP classifier with word2vec model earns an accuracy of 95.83% for an English corpus, suggesting that the word2vec word embedding model can evidently enhance the identification accuracy compared to other classical models such as n-gram frequencies and bag of words.

The number of people who suffer from tension problem is growing around the world, which is capable of causing other very serious illness like glaucoma. This last is a chronic and irreversible disease which can lead to vision loss and blindness, that's why doctors recommend continuous follow-up all the time. The development of computer aided diagnosis systems (CAD) in the medical field has been viewed with great interest by doctors to have a second diagnosis. In our previous study concerning classical methods, a Twin Support Vector Machine (TWSVM) classifier was proposed to glaucoma images classification by using handcrafted features. To overcome the main limits of classical approach, we explore a deep learning algorithm based on convolutional neural network (CNN) for the automatic generation of features depending on any used dataset. To highlight the impact of the CNN classifier in medical diagnosis, several empirical studies was performed based on convolution layers numbers and activation function. Obtained results will be compared and discussed with previous work based on TWSVM, by applying the RimOne dataset.

Authorship Verification is considered as a topic of growing interest in research, which has shown excellent development in recent years. We want to know if an unknown document belongs to the documents set known to an author or not. Classical text classifiers often focus on many human designed features, such as dictionaries, knowledge bases and special tree kernels. Other studies use the N-gram function that often leads to the curse of dimensionality. Contrary to traditional approaches, this article proposes a new scheme of Machine Learning model based on fusion of three different architectures namely, Convolutional Neural Networks, Recurrent-Convolutional Neural Networks and Support Vector Machine classifiers without human-designed features. Word2vec based Word Embeddings is proposed to learn the best word representations for automatic authorship verification. Word Embeddings provides semantic vectors and extracts the most relevant information about raw text with a relatively small dimension. As well as the classifiers generally make different errors on the same learning samples which results in a combination of several points of view to maintain relevant information contained in different classifiers. The final decision of our system is obtained by combining the results of the three models using the voting method.