Résumé: Background Cardiac image analysis plays an important role in detecting and categorizing cardiovascular diseases (CVDs), such as coronary artery disease (CAD), heart failure, congenital heart defects, arrhythmias (irregular heartbeat), and valvular heart disease. The automated identification of these diseases represents a significant advancement in achieving early diagnosis and mitigating disease exacerbations. While extant methodologies offer advanced means for the automatic segmentation and identification of cardiac structures and pathologies, recent strides in deep learning (DL) and modern imaging modalities within cardiology have introduced new opportunities for researchers. This has underscored the importance of deep model compression and optimization techniques. This review comprehensively surveys recent deep learning applications in interpreting cardiac images, encompassing common imaging modalities. Method Following the PRISMA methodology, we reviewed recent research on advanced and frequently used DL architectures in cardiac image analysis, along with the most employed cardiac imaging datasets. Additionally, we analyse recent contributions focused on deep model compression in cardiac image processing tasks. Results The application of DL techniques in cardiac image analysis has seen significant progress through the introduction of new techniques such as transformers, foundation models and compression techniques. Conclusions The paper concludes with a critical discussion addressing open challenges and proposing future research directions in this domain.

Résumé: Classifier subset selection becomes an important stage in multiple classifier systems (MCSs) design to reduce the number of classifiers by eliminating the identical and inaccurate members. Minimum redundancy maximum relevance (mRMR) is a feature selection method that compromises between relevance and redundancy by obliterating similar members and keeping the most pertinent ones. In the current work, a novel classifier subset selection method based on mRMR method and diversity measures is proposed for building an efficient classifier ensemble. The proposed selection model suggested the greedy search algorithm using diversity-accuracy criteria to determine the optimal classifier set. The disagreement and Q-statistic measures are calculated to estimate the diversity among the members. Furthermore, the relevance is used as a means to determine the accuracy of the ensemble and its members. The experimental results over 24 datasets from the UCI repository and Kuncheva collection for real datasets are tested. The results established the efficiency of the proposed selection method with superior performance compared to the popular ensembles and several selection methods.

Résumé: Urban drainage systems may derive advantages from the implementation of machine learning methods for decision-making and cleansing operations. Conventional decision support systems are rendered ineffective in tackling the intricate and indeterminate aspects of urban planning concerns. This paper provides an overview of machine learning methods applied in modeling urban drainage systems, which are classified into five distinct approaches: Supervised learning, unsupervised learning, deep learning, Reinforcement learning, and finally hybrid approaches combining two or more previous algorithms. This study explores also diverse datasets that researchers can utilize in their scientific investigations. The present related works exploration reveals that a majority of studies lean towards reinforcement learning and deep learning methods, predominantly utilizing local datasets. Consequently, we have compiled a table containing some reference datasets. These advancements result in enhanced accuracy and efficiency in predicting and managing urban drainage systems.

Résumé: In the last three years, The COVID-19 pandemic has placed immense strain on healthcare systems worldwide. Imaging techniques, such as Chest X-Ray (CXR) images, play an essential role in diagnosing COVID-19. Recently, Computer Aided Diagnosis systems based on Deep Learning have been widely used to identify COVID-19 from other upper respiratory diseases such as viral pneumonia. To enhance COVID-19 diagnosis using CXR images, this article proposes a methodology that combines three deep learning models renowned for their depth and feature extraction capabilities: ResNet, DenseNet, and VGG. The model’ strengths are harnessed using Aggregation techniques such as stacking and voting (majority, weighted, average), to improve prediction accuracy. Moreover, the used dataset was subjected to preprocessing, balancing, and augmentation techniques, including rotation and zooming, to ensure robust recognition of CXR images, even in challenging positions. Experimental results demonstrate that the combination of model ensembles and enhanced dataset leads to superior accuracy in diagnosing COVID-19, with an impressive 97.95% accuracy, 98.2% F1 score, and 99.9% AUC-ROC. This approach not only outperforms individual models but also offers a more reliable and efficient solution for COVID-19 detection, particularly in complex cases where symptoms overlap with other respiratory illnesses.

Résumé: Medical imaging technologies have drastically changed the way healthcare professionals diagnose and treat patients. However, medical imaging datasets are limited in terms of size and diversity, which can lead to inaccurate diagnosis and treatment plans. To overcome these limitations, Generative Adversarial Networks (GANs) have been used to generate new medical images based on existing datasets. GANs are a type of deep learning artificial intelligence that can use existing datasets to create new synthetic images that look almost indistinguishable from real images. GANs can be used to generate medical images of organs, tissues, and diseases that are not present in existing datasets, providing healthcare professionals with more accurate and diverse datasets to diagnose and treat patients. GANs for medical image synthesis is a powerful tool that can revolutionize the healthcare industry. In this paper an investigation of the most used Gans architecture are presented. a state of the art of the latest works applying Gans are also discussed based on the adopted modality.

Résumé: In recent years, the rapid growth of electronic data and information has gotten a lot of attention to find relevant knowledge such as textual information. The goal of automatic text classification is to automatically predict textual articles classes, especially in the medical domain. However, for some applications, the used data must inherently be described by more than one label. In this research, a new scheme of medical multi-label text classification is investigated which is based on intelligent engineering features using GloVe technique and LSTM classifier. The main particularity of GloVe permits the extraction of informative features to the word level automatically and capture the global and local textual semantics. The choice of the LSTM model is motivated by the success that has been achieved by taking into account the very long-term dependencies between words. The experiment of our approach named GL-LSTM based on Ohsumed cardiovascular text dataset has produced impressive results with an overall accuracy of 0.927 compared with related works existing in the literature

Résumé: Medical imaging technologies have drastically changed the way healthcare professionals diagnose and treat patients. However, medical imaging datasets are limited in terms of size and diversity, which can lead to inaccurate diagnosis and treatment plans. To overcome these limitations, Generative Adversarial Networks (GANs) have been used to generate new medical images based on existing datasets. GANs are a type of deep learning artificial intelligence that can use existing datasets to create new synthetic images that look almost indistinguishable from real images. GANs can be used to generate medical images of organs, tissues, and diseases that are not present in existing datasets, providing healthcare professionals with more accurate and diverse datasets to diagnose and treat patients. GANs for medical image synthesis is a powerful tool that can revolutionize the healthcare industry. In this paper an investigation of the most used Gans architecture are presented. a state of the art of the latest works applying Gans are also discussed based on the adopted modality.

Résumé: In recent years, the rapid growth of electronic data and information has gotten a lot of attention to find relevant knowledge such as textual information. The goal of automatic text classification is to automatically predict textual articles classes, especially in the medical domain. However, for some applications, the used data must inherently be described by more than one label. In this research, a new scheme of medical multi-label text classification is investigated which is based on intelligent engineering features using GloVe technique and LSTM classifier. The main particularity of GloVe permits the extraction of informative features to the word level automatically and capture the global and local textual semantics. The choice of the LSTM model is motivated by the success that has been achieved by taking into account the very long-term dependencies between words. The experiment of our approach named GL-LSTM based on Ohsumed cardiovascular text dataset has produced impressive results with an overall accuracy of 0.927 compared with related works existing in the literature