Introduction: Nowadays, Artificial intelligence and machine learning have emerged as a powerful tool for the analysis of medical images such as MRI scans. This technology holds significant potential to improve diagnostic services and accelerate medical advances by facilitating clinical decision-making. Method: In this work, we developed a Convolutional Neural Network (CNN) model specifically designed for the classification of medical images. Using a selected database, the model achieved a classification accuracy of 92%. To further improve the performance, we leveraged the pre-trained VGG16 model, which increased the classification accuracy to 100%. Additionally, we preprocessed the MRI images using the Roboflow platform and then developed YOLOv5 models for the detection of tumors, infections, and cancerous lesions. Result: The results demonstrate a localization accuracy of 50.41% for these medical conditions. Conclusion: This research highlights the value of AI-driven approaches in enhancing medical image analysis and their potential to support more accurate diagnoses and accelerate advancements in healthcare.

Automatic perception of potential safety hazards: A cross-modal multi-task framework for feature alignment, image classification and captioning

Reduced storage direct tensor ring decomposition for convolutional neural networks compression.

Classification of color images of leathers tanned with different vegetable tannins by convolution neural network

ShallowMRI: A novel lightweight CNN with novel attention mechanism for Multi brain tumor classification in MRI images

SR-PAN-EffNet: A collaborative optimization approach for low-quality image classification and restoration

MixerCA: An efficient and accurate model for high-performance hyperspectral image classification

Dual adversarial attacks on Explainable Deep Learning in medical image classification.

Gradual machine learning for medical image classification via evolutionary feature fusion

Limitations of principal component analysis in COVID-19 CT image classification

Reinforced graph aggregation cross-domain few-shot learning for hyperspectral remote sensing image classification

Purpose Digital image automatic classification has become a critical field in machine learning. With the exponential increase in the availability of digital data and the growing complexity of applications, the need to develop accurate and efficient data automatic classification models has become urgent across multiple sectors, as they contribute to enhancing operational efficiency. This scenario underscores the necessity of exploring and developing new approaches capable of overcoming these challenges, further improving the accuracy and efficiency of classification techniques. This paper aims to present the ongoing research for development and testing an automatic image classification model for digital libraries, based on complex neural networks (CNNs). Design/methodology/approach Despite the significant advancements achieved with the advent of deep learning approaches, the challenges of automatic classification of digital resources remain in terms of generalization, model interpretability and reducing dependence on large training data sets. After outlining the state-of-the-art digital resources’ automatic classification, the paper describes the model research and design and the pilot of implemented application workflow. Finally, preliminary research results are assessed, considering that experimentation is still ongoing to evaluate the potential integration of AI tools to enhance model performance. Findings The research addresses the challenge of developing a model effective for classifying digital resources referring to the huge and various contexts of digital libraries including resources representative of manuscript, early printed and modern books. The process to develop the first pilot of the automatic classification system the researchers designed and developed has been clearly outlined. The workflow and the generation of a specific CNN for classifying digital libraries are detailed by examples, figures and tables that show each step of the process describing the methods, techniques and technologies used. Research limitations/implications There are no research limitations/implications. Practical implications There are no practical implications. Social implications There are no social implications. Originality/value The experimentation results provide an encouraging overall picture of the developed models’ performance, highlighting their potential for analyzing and classifying the structures of the considered materials. An extensive series of tests were conducted on a diverse data set to assess their effectiveness, accompanied by a rigorous validation procedure on an even larger sample. The pilot model shows remarkable performance in accuracy, achieving an average correct classification rate of 78% for the three analyzed types and over the full validation data set. The learning curve displayed good convergence, suggesting that further optimizations could improve the overall precision, particularly fine-tuning the hyperparameters regulating the training process and refining the machine learning model topology. Such improvements would increase accuracy and reduce uncertainty for classes that showed greater variability.

Convolutional neural networks are constructed with massive operations with different types and are highly computationally intensive. Among these operations, multiplication operation is higher in computational complexity and usually requires more energy consumption with longer inference time than other operations, which hinders the deployment of convolutional neural networks on mobile devices. In many resource-limited edge devices, complicated operations can be calculated via lookup tables to reduce computational cost. Motivated by this, in this paper, we introduce a generic and efficient lookup operation which can be used as a basic operation for the construction of neural networks. Instead of calculating the multiplication of weights and activation values, simple yet efficient lookup operations are adopted to compute their responses. To enable end-to-end optimization of the lookup operation, we construct the lookup tables in a differentiable manner and propose several training strategies to promote their convergence. By replacing computationally expensive multiplication operations with our lookup operations, we develop lookup networks for the image classification, image super-resolution, and point cloud classification tasks. It is demonstrated that our lookup networks can benefit from the lookup operations to achieve higher efficiency in terms of energy consumption and inference speed while maintaining competitive performance to vanilla convolutional networks. Extensive experiments show that our lookup networks produce state-of-the-art performance on different tasks (both classification and regression tasks) and different data types (both images and point clouds).

Dual-student co-training network using Mamba and unreliable sample learning with class-adaptation for hyperspectral image classification

Cross-scene hyperspectral image classification based on cross-domain feature extraction and category decision collaborative optimization

Deeply understanding features to achieve efficient remote sensing image classification

Knowledge distillation and teacher-student learning in medical imaging: Comprehensive overview, pivotal role, and future directions.

FairREAD: Re-fusing demographic attributes after disentanglement for fair medical image classification.

Gaze-guided vision transformer for chest X-ray image classification

Hybrid quantum sparse coding and dynamic convolution capsule network for enhanced image classification