Lung cancer is one of the fatal diseases whose early diagnosis is essential to mitigate the death rate. Computed Tomography (CT) scans are widely used for lung cancer diagnosis, but manual interpretation by health professionals can lead to inconsistent results. To address this, we propose Lung-AttNet, a novel lightweight convolutional neural network (CNN) model enhanced with an attention mechanism. Lung-AttNet incorporates a convolutional block with a Lightweight Global Attention Module (LGAM) to effectively distinguish between lung cancer types. The convolutional block extracts both low- and high-dimensional features, while LGAM captures feature dependencies across channel and spatial dimensions. The model is evaluated using the Kaggle CT scan dataset, which includes four classes: adenocarcinoma, large cell carcinoma, squamous cell carcinoma, and normal. Extensive experiments, including ablation studies, 5-fold cross-validation, and explainable AI (XAI) techniques such as Grad-CAM and LIME, demonstrate that Lung-AttNet achieves an average accuracy of 91.5%. Furthermore, to address medical data sensitivity and privacy concerns, the model is deployed in a Federated Learning (FL) framework, where the global model is trained using weights from local models rather than sharing raw data. In the FL environment, Lung-AttNet achieves an accuracy of 92% with 2 and 3 clients, underscoring its robustness and adaptability for real-world applications.

The electrocardiogram (ECG) has significant clinical importance for analyzing most cardiovascular diseases. ECGs beat morphologies, beat durations, and amplitudes vary from subject to subject and diseases to diseases. Therefore, ECG morphology-based modeling has long-standing research interests. This work aims to develop a simplified ECG model based on a minimum number of parameters that could correctly represent ECG morphology in different cardiac dysrhythmias. A simple mathematical model based on the sum of two Gaussian functions is proposed. However, fitting more than one Gaussian function in a deterministic way has accuracy and localization problems. To solve these fitting problems, two hybrid optimization methods have been developed to select the optimal ECG model parameters. The first method is the combination of an approximation and global search technique (ApproxiGlo), and the second method is the combination of an approximation and multi-start search technique (ApproxiMul). The proposed model and optimization methods have been applied to real ECGs in different cardiac dysrhythmias, and the effectiveness of the model performance was measured in time, frequency, and the time-frequency domain. The model fit different types of ECG beats representing different cardiac dysrhythmias with high correlation coefficients (>0.98). Compared to the nonlinear fitting method, ApproxiGlo and ApproxiMul are 3.32 and 7.88 times better in terms of root mean square error (RMSE), respectively. Regarding optimization, the ApproxiMul performs better than the ApproxiGlo method in many metrics. Different uses of this model are possible, such as a syntactic ECG generator using a graphical user interface has been developed and tested. In addition, the model can be used as a lossy compression with a variable compression rate. A compression ratio of 20:1 can be achieved with 1 kHz sampling frequency and 75 beats per minute. These optimization methods can be used in different engineering fields where the sum of Gaussians is used.

Electroencephalographic signal is a representative signal that contains information about brain activity, which is used for the detection of epilepsy since epileptic seizures are caused by a disturbance in the electrophysiological activity of the brain. The prediction of epileptic seizure usually requires a detailed and experienced analysis of EEG. In this paper, we have introduced a statistical analysis of EEG signal that is capable of recognizing epileptic seizure with a high degree of accuracy and helps to provide automatic detection of epileptic seizure for different ages of epilepsy. To accomplish the target research, we extract various epileptic features namely approximate entropy (ApEn), standard deviation (SD), standard error (SE), modified mean absolute value (MMAV), roll-off (R), and zero crossing (ZC) from the epileptic signal. The k-nearest neighbours (k-NN) algorithm is used for the classification of epilepsy then regression analysis is used for the prediction of the epilepsy level at different ages of the patients. Using the statistical parameters and regression analysis, a prototype mathematical model is proposed which helps to find the epileptic randomness with respect to the age of different subjects. The accuracy of this prototype equation depends on proper analysis of the dynamic information from the epileptic EEG.

Diabetic retinopathy is one of the most significant retinal diseases that can lead to blindness. As a result, it is critical to receive a prompt diagnosis of the disease. Manual screening can result in misdiagnosis due to human error and limited human capability. In such cases, using a deep learning-based automated diagnosis of the disease could aid in early detection and treatment. In deep learning-based analysis, the original and segmented blood vessels are typically used for diagnosis. However, it is still unclear which approach is superior. In this study, a comparison of two deep learning approaches (Inception v3 and DenseNet-121) was performed on two different datasets of colored images and segmented images. The study’s findings revealed that the accuracy for original images on both Inception v3 and DenseNet-121 equaled 0.8 or higher, whereas the segmented retinal blood vessels under both approaches provided an accuracy of just greater than 0.6, demonstrating that the segmented vessels do not add much utility to the deep learning-based analysis. The study’s findings show that the original-colored images are more significant in diagnosing retinopathy than the extracted retinal blood vessels.