Neven Saleh

Skin cancer (SC) is one of the most prevalent forms of cancer worldwide. Both melanoma and non-melanoma types pose major challenges for early detection, accurate diagnosis, and proper treatment. Conventional diagnostic approaches, such as biopsy and visual examination, are often time-consuming, subjective, and prone to human error. Recent advances in artificial intelligence (AI) and deep learning (DL) have greatly improved the accuracy of SC diagnosis. This systematic review explores the applications of DL techniques in the segmentation and classification of skin lesions between 2014 and 2024. Following the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines and applying predefined inclusion and exclusion criteria, a total of 77 experimental studies out of 540 were analyzed from major databases, including Scopus, IEEE, PubMed, and MDPI. Convolutional neural networks (CNNs) were identified as the most widely used for classification, while U-Net and its variants dominated segmentation tasks. Hybrid and ensemble frameworks demonstrated superior performance on benchmark datasets such as the International Skin Imaging Collaboration (ISIC) archive and HAM10000. Moreover, this work incorporates a formal risk-of-bias analysis, revealing critical concerns about class imbalance and data leakage. Almost all reviewed studies for the classification task achieved an average accuracy of 96% for the ISIC dataset, while the HAM10000 dataset attained an average accuracy of 93%. Despite these advances, challenges such as class imbalance, limited dataset diversity, and insufficient clinical validation persist. Addressing these issues through data augmentation, explainable AI, and federated learning could further enhance the generalizability and clinical applicability of AI-driven diagnosis systems. Additionally, this study identifies a clear paradigm shift from standalone CNNs to hybrid frameworks and multi-source feature fusion strategies, aiming to improve SC diagnosis.

Malaria remains a critical global health challenge, particularly in tropical and subtropical regions. While traditional methods for diagnosis are effective, they face some limitations related to accuracy, time consumption, and manual effort. This study proposes an advanced, automated diagnostic framework for malaria detection using a multi-model architecture integrating deep learning and machine learning techniques. The framework employs a transfer learning approach that incorporates ResNet 50, VGG16, and DenseNet-201 for feature extraction. This is followed by feature fusion and dimensionality reduction via principal component analysis. A hybrid scheme that combines support vector machine and long short-term memory networks is used for classification. A majority voting mechanism aggregates outputs from all models to enhance prediction robustness. The approach was validated on a publicly available dataset comprising 27,558 microscopic thin blood smear images. The results demonstrated superior performance, achieving an accuracy of 96.47%, sensitivity of 96.03%, specificity of 96.90%, precision of 96.88%, and F1-score of 96.45% using the majority voting ensemble. Comparative analysis highlights the framework’s advancements over existing methods in diagnostic reliability and computational efficiency. This work underscores the potential of AI-driven solutions in advancing malaria diagnostics and lays the foundation for applications in other blood-borne diseases.

Objectives: One of the worldwide public health issues mostly affecting children and expectant mothers is Anemia. Recently, non-invasive hemoglobin (Hb) measurements, such as machine learning (ML) algorithms, can diagnose Anemia more quickly and efficiently. Methods: To diagnose Anemia using photoplethysmography (PPG), two tracks are investigated in this paper, based on clinical data and PPG signals. We use state-of-the-art data for Hb levels, extracted from PPG signals. This first track’s methodology is divided into three stages: the labelling of the data as normal and abnormal; the data pre-processing; and applying ML algorithms based on four given features. We extracted nineteen features for red and infrared measurements in the second track. The second track’s methodology is broken down into five stages: labelling of the data; data processing; signal augmentation; feature extraction; and applying ML algorithms. A five-fold cross-validation technique was applied for both tracks. Results: We succeeded in classifying the anemic condition with 100 % classification accuracy. Our accurate detection of anemic status will promote preventive healthcare. Conclusions: Ultimately, this proposed ML model in this paper validated the effectiveness of the ML algorithms as non-invasive techniques for identifying Anemia.

Neurological disorders affect the nervous system and can impair physical, cognitive, or emotional functions. They often result in challenges such as movement difficulties and the inability to perform daily activities. Common conditions include stroke, traumatic brain injury, and cerebral palsy. Physical therapy is a common approach to managing these disorders. Recently, virtual reality (VR), a technology that creates interactive, simulated environments, has been used in rehabilitation. This study presents a rehabilitative game-based system to improve patients’ movements and cognitive abilities. Six games were designed using the Unity platform, namely, “Piano,” “Connect,” “Drag & Drop,” “Little Intelligent,” “Memory,” and “Hack & Slash.” The Oculus Quest 2 VR headset was used to simulate the virtual environment for gaming. A mobile application called “Recover Me” was created to facilitate communication between patients and physiotherapists. A score index was generated for each patient, indicating the performance. It enables monitoring and assessment of the patients, leading to customizing the treatment plan based on progress. The study proposed simulating monitoring and evaluation of the patients by training an artificial neural network model to predict scores for the developed games and consequently indicate the patient’s actual status. A dataset of 50 patients with different injuries was used. Results indicate patient satisfaction with gaming and enjoyment. Moreover, a regression analysis was performed to detect the progress level of each patient, indicating that 60% of the tested patients had improved. A low-cost VR game-based system has proven effective in rehabilitating neurological disorders.

Skin cancer is a type of cancer disease in which abnormal alterations in skin characteristics can be detected. It can be treated if it is detected early. Many artificial intelligence-based models have been developed for skin cancer detection and classification. Considering the development of numerous models according to various scenarios and selecting the optimum model was rarely considered in previous works. This study aimed to develop various models for skin cancer classification and select the optimum model. Convolutional neural networks (CNNs) in the form of AlexNet, Inception V3, MobileNet V2, and ResNet 50 were used for feature extraction. Feature reduction was carried out using two algorithms of the grey wolf optimizer (GWO) in addition to using the original features. Skin cancer images were classified into four classes based on six machine learning (ML) classifiers. As a result, 51 models were developed with different combinations of CNN algorithms, without GWO algorithms, with two GWO algorithms, and with six ML classifiers. To select the optimum model with the best results, the multicriteria decision-making approach was utilized to rank the alternatives by perimeter similarity (RAPS). Model training and testing were conducted using the International Skin Imaging Collaboration (ISIC) 2017 dataset. Based on nine evaluation metrics and according to the RAPS method, the AlexNet algorithm with a classical GWO yielded the optimum model, achieving a classification accuracy of 94.5%. This work presents the first study on benchmarking skin cancer classification with many models. Feature reduction not only reduces the time spent on training but also improves classification accuracy. The RAPS method has proven its robustness in the problem of selecting the best model for skin cancer classification.

Medical Laboratory Equipment (MLE) is one of the most influential means for diagnosing a patient in healthcare facilities. The accuracy and dependability of clinical laboratory testing is essential for making disease diagnosis. A risk-reduction plan for managing MLE is presented in the study. The methodology was initially based on the Failure Mode and Effects Analysis (FMEA) method. Because of the drawbacks of standard FMEA implementation, a Technique for Ordering Preference by Similarity to the Ideal Solution (TOPSIS) was adopted in addition to the Simple Additive Weighting (SAW) method. Each piece of MLE under investigation was given a risk priority number (RPN), which in turn assigned its risk level. The equipment performance can be improved, and maintenance work can be prioritized using the generated RPN values. Moreover, five machine learning classifiers were employed to classify TOPSIS results for appropriate decision-making. The current study was conducted on 15 various hospitals in Egypt, utilizing a 150 MLE set of data from an actual laboratory, considering three different types of MLE. By applying the TOPSIS and SAW methods, new RPN values were obtained to rank the MLE risk. Because of its stability in ranking the MLE risk value compared to the conventional FMEA and SAW methods, the TOPSIS approach has been accepted. Thus, a prioritized list of MLEs was identified to make decisions related to appropriate incoming maintenance and scrapping strategies according to the guidance of machine learning classifiers.

Health Technology Assessment (HTA) approaches are increasingly widely used in healthcare decision-making. HTA in medical devices remains a serious difficulty today, despite its extensive usage in medications and surgery. Because of scarce resources and rapidly rising healthcare costs, it is now necessary to assess the efficacy, appropriateness, and costs of health technologies such as drugs, biologics, devices, equipment and supplies, medical and surgical procedures, support systems, and organizational and managerial systems. The main goal of studying HTA devices is not to improve the cost-effectiveness ratio but to make decisions about purchasing and/or incorporating the device. Another significant goal is to extend the lifetime of the device. This study proposes a program to assist a clinical engineer in assessing a large variety of radiological devices (Ultrasound, x-Ray, Mammography, Digital Radiography (DR), and Computed Radiology (CR) using various Multi-Criteria Decision Making (MCDM) Methods. MCDM is now one of the best tools for evaluating medical equipment. This paper employed multiple MCDM methods, including the Evaluation Based on Distance from Average Solution (EDAS) method, Multi-objective Optimization based on Ratio Analysis (MOORA) method, The Technique for Order of Preference by Similarity Ideal Solution (TOPSIS) method, and Simple Additive Weighting (SAW) method. After experimenting with them, we discovered that the EDAS technique yields the greatest results in terms of device effectiveness. Using the various MCDM approaches, we were able to obtain the effectiveness and cost-effectiveness values needed to conduct HTA, which means we could assess 250 different devices from the four modalities in various locations.

The incidence of vision impairment is rapidly increasing. Diagnosis and classifying retinal abnormalities in ophthalmological applications is a significant challenge. Using Optical Coherence Tomography (OCT), the study aims to develop a computer aided diagnosis system for detecting and classifying retinal disorders. Choroidal neovascularization, diabetic macular edema, drusen, and normal cases are the investigated groups. Both deep learning and machine learning are combined to build the system. The SqueezeNet neural network was modified to extract features. The Support Vector Machine (SVM), K-Nearest Neighbor (K-NN), Decision Tree (DT), and Ensemble Model (EM) algorithms were used for disorder classification. The Bayesian optimization technique was also used to determine the best hyperparameters for each model. The model’ performance was evaluated through nine criteria using 12,000 OCT images. The results have demonstrated accuracies of 97.39, 97.47, 96.98, and 95.25% for the SVM, K-NN, DT, and EM, respectively. When results are compared to relevant studies in terms of accuracy and tested samples, they show superior performance. As a result, a novel computer-aided diagnosis system for detecting and classifying retinal diseases has been developed, reducing human error while also saving time.