Brain tumors being ninth in terms of prevalence and one of the most frequently diagnosed malignant tumors, negatively impact millions of individuals. Identifying and classifying tumors from MRI used for health monitoring poses a challenge for radiologists, yet early detection could significantly enhance the chances of effective treatment. Researchers in the field of explainable AI are currently focused on developing sophisticated techniques to classify and diagnose brain diseases. This study presents a novel framework that enhances the interpretability of our proposed system for brain tumor detection by utilizing explainable AI techniques. To enhance interpretability, we integrate the optimized recursive feature elimination selection technique with support vector machines. This method effectively eliminates redundant features, identifies the most important ones, and enhances the efficiency of detecting tasks. Following that, the optimal recursive feature elimination (ORFE) features are combined using the supervised support vector machine (SVM) technique. While EfficientNet-CNN is a very useful feature extraction framework that extracts the most important features from a transparent model, we reduced the overall computational complexity through feature elimination and supervised models, and the Figshre dataset clearly demonstrated the efficacy of our model. This study achieved very exceptional results and reduced computational complexity as compared to a single CNN model. The experimental results indicate that the proposed SVM-RFE based technique accurately detects brain tumors with a 99.51% accuracy and a specificity score of 99.63%. The proposed approach obtained an accuracy of 98.93% with a standard deviation of 0.032 using 10-fold cross-validation. Additionally, it produced an optimal ROC_AUC of 100% for cases including meningiomas and pituitary tumors.

Innovation in learning algorithms has made retinal vessel segmentation and automatic grading techniques crucial for clinical diagnosis and prevention of diabetic retinopathy. The traditional methods struggle with accuracy and reliability due to multi-scale variations in retinal blood vessels and the complex pathological relationship in fundus images associated with diabetic retinopathy. While the single-modal diabetic retinopathy grading network addresses class imbalance challenges and lesion representation in fundus image data, dual-modal diabetic retinopathy grading methods offer superior performance. However, the scarcity of dual-modal data and the lack of effective feature fusion methods limit their potential due to multi-scale variations. This paper addresses these issues by focusing on multi-scale retinal vessel segmentation, dual feature fusion, data augmentation, and attention-based grading. The proposed model aims to improve comprehensive segmentation for retinal images with varying vessel thicknesses. It employs a dual-branch parallel architecture that integrates a transformer encoder with a convolutional neural network encoder to extract local and global information for synergistic saliency learning. Besides that, the model uses residual structures and attention modules to extract critical lesions, enhancing the accuracy and reliability of diabetic retinopathy grading. To evaluate the efficacy of the proposed approach, this study compared it with other pre-trained publicly open models, ResNet152V2, ConvNext, Efficient Net, DenseNet, and Swin Transform, with the same developmental parameters. All models achieved approximately 85% accuracy with the same image preparation method. However, the proposed approach outperforms and optimizes existing models by achieving an accuracy of 99.17%, 99.04%, and 99.24%, on Kaggle APTOS19, IDRiD, and EyePACS datasets, respectively. These results support the model’s utility in helping ophthalmologists diagnose diabetic retinopathy more rapidly and accurately.

The architecture of software-defined networking (SDN) involves the separation of the network control plane from the routing plane. If this initiative turns out well, it has the potential to reduce operating expenses and the duration required to provide new services in comparison to traditional networks. However, this architecture has additional security concerns, including a single point of failure that could potentially provide any user with unrestricted access to the entire network. Nevertheless, it is essential to reduce the probability of security breaches. The development of immediate intrusion detection systems (IDSs) that can quickly spot and stop malicious activities like distributed denial of service (DDoS), DoS, web-attacks, and Bot-NET is an important part of SDN architecture. Several researchers are using cutting-edge methods, such as machine learning, to investigate and elucidate the causes behind the sudden rise in attacks and abnormal behavior, but the majority of these methods are deficient in terms of flexibility and accuracy. This study proposed a lightweight method for detecting different SDN attacks from intrusion-defined networks. The lightweight long short-term memory (LSTM) network has the capability to capture temporal patterns and sequential interactions in the SDN data. It also learned important context that is efficient for feature extraction and then developed supervised random forest (SRF) for the attack prediction. The dataset consists of 207,146 rows and 84 features that were preprocessed, including separate features and target attacks. The experiments show that the proposed method achieved 99.93% accuracy for attack detection and 0.0090 loss, confirming its efficacy. We also tested the proposed method on another SDN dataset and achieved 99.43% accuracy for multi-class attack detection. Furthermore, the use of supervised random forest reduces the model’s complexity, resulting in increased overall efficiency.