Background: Accurate classification of brain tumors in medical images is vital for effective diagnosis and treatment planning, which improves the patient’s survival rate. In this paper, we investigate the application of Convolutional Neural Networks (CNN) as a powerful tool for enhancing diagnostic accuracy using a Magnetic Resonance Imaging (MRI) dataset. Method: This study investigates the application of CNNs for brain tumor classification using a dataset of Magnetic Resonance Imaging (MRI) with a resolution of 200 × 200 × 1. The dataset is pre-processed and categorized into three types of tumors: Glioma, Meningioma, and Pituitary. The CNN models, including the Classic layer architecture and the ResNet50 architecture, are trained and evaluated using an 80:20 training-testing split. Results: The results reveal that both architectures accurately classify brain tumors. Classic layer architecture achieves an accuracy of 94.55%, while the ResNet50 architecture surpasses it with an accuracy of 99.88%. Compared to previous studies and 99.34%, our approach offers higher precision and reliability, demonstrating the effectiveness of ResNet50 in capturing complex features. Conclusions: The study concludes that CNNs, particularly the ResNet50 architecture, exhibit effectiveness in classifying brain tumors and hold significant potential in aiding medical professionals in accurate diagnosis and treatment planning. These advancements aim to further enhance the performance and practicality of CNN-based brain tumor classification systems, ultimately benefiting healthcare professionals and patients. For future research, exploring transfer learning techniques could be beneficial. By leveraging pre-trained models on large-scale datasets, researchers can utilize knowledge from other domains to improve brain tumor classification tasks, particularly in scenarios with limited annotated data.

Human facial feature extraction plays a critical role in various applications, including biorobotics, polygraph testing, and driver fatigue monitoring. However, many existing algorithms rely on end-to-end models that construct complex classifiers directly from face images, leading to poor interpretability. Additionally, these models often fail to capture dynamic information effectively due to insufficient consideration of respondents' personal characteristics. To address these limitations, this paper evaluates two prominent approaches: the constrained local model (CLM), which accurately extracts facial features depending on patch experts, and the active shape model (ASM), designed to simultaneously extract the appearance and shape of an object. We assess the performance of these models on the MORPH dataset using point to point error as evaluation metrics. Our experimental results demonstrate that the CLM achieves higher accuracy, while the ASM exhibits better efficiency. These findings provide valuable insights for selecting the appropriate model based on specific application requirements.

The heterogeneity of breast tissue and subtle morphological variations in ultrasound images make breast cancer detection a challenging task. This study proposes a hybrid deep learning framework that integrates EfficientNetB4 and ConvNeXt within a dual-stream architecture enhanced by advanced attention mechanisms. The model combines multi-scale texture representation with spatial feature extraction to improve classification performance. A two-stage preprocessing pipeline, consisting of adaptive median filtering and bilateral filtering, is applied to reduce speckle noise while preserving important structural details. The proposed method is evaluated on BUSI and UDAIT datasets, achieving 87.82% accuracy, 87.33% precision, and 85.33% recall on BUSI, and 85.69% accuracy, 84.00% precision, and 78.00% recall on UDAIT. These results outperform several baseline models, including ResNet-50, DenseNet-121, and vision transformers. Error analysis shows limitations in detecting small lesions and cross-modal generalization, with reduced performance on mammography images. Attention visualization demonstrates strong agreement with radiologist annotations, supporting model interpretability. The findings highlight the effectiveness of hybrid architectures for ultrasound-based breast cancer detection while emphasizing the need for modality-specific optimization.

American scientific publishing group