Transformer models have demonstrated remarkable success in various fields, but face notable challenges in blood cell classification. These challenges arise primarily from two factors: (1) high computational demands due to the substantial number of parameters, and (2) insufficient labeled data in blood cell datasets, leading to overfitting risks. To address these issues, we introduce BccT, an innovative Transformer-based model specifically designed to optimize the classification of blood cell images. Central to our approach is the Token Fusion module, which intelligently merges similar tokens, thereby enhancing computational efficiency and reducing training overhead without compromising performance. This novel mechanism contrasts with traditional Transformer methods that process each token independently, resulting in significant redundancy. Additionally, we propose the Fixed-Random Classifier to mitigate overfitting risks, particularly in limited-data environments. Unlike conventional classifiers that necessitate constant parameter updates across all layers, our Fixed-Random Classifier comprises only three layers with predominantly static, randomly-initialized parameters, thus reducing dependency on extensive labeled data. Extensive experiments validate the effectiveness of our model. Using ViT_base as the backbone, our model achieves similar accuracy in half the training time compared to the ViT_base model. In experiments on four public blood cell datasets, our model improves accuracy by approximately 4.5% compared to other state-of-the-art methods.

Review A Short Survey on Computer-Aided Diagnosis of Alzheimer’s Disease: Unsupervised Learning, Transfer Learning, and Other Machine Learning Methods Si-Yuan Lu School of Communications and Information Engineering, Nanjing University of Posts and Telecommunications, Nanjing 210003, China Received: 15 April 2024; Revised: 7 May 2024; Accepted: 14 May 2024; Published: 31 May 2024 Alzheimer’s Disease (AD) is a neurodegenerative disorder, which is irreversible and incurable. Early diagnosis plays a significant role in controlling the progression of AD and improving the patient’s quality of life. Computer-aided diagnosis (CAD) methods have shown great potential to assist doctors in analyzing medical data, such as magnetic resonance images, positron emission tomography, and mini-mental state examination. Contributed by the advanced deep learning models, predictions of CAD methods for AD are becoming more and more accurate, which can provide a reference and verification for manual screening. In this paper, a short survey on the application of recent CAD methods in AD detection is presented. The advantages and drawbacks of these methods are discussed in detail, especially the methods based on convolutional neural networks, and the future research directions are summarized subsequently. With this survey, we hope to promote the development of CAD for early detection of AD.

Chest X-rays (CXRs) are widely utilized for screening pulmonary diseases. However, distinguishing between tuberculosis (TB) and pneumonia remains a challenging task, as their radiological features often exhibit significant overlap. This challenge is particularly acute in resource-constrained settings, where rapid and low-cost automated analysis techniques would be especially beneficial. To address this issue, we propose a lightweight deep learning framework with Triple Permuter and Split Multi-Head Self-Attention (LTS-Net), specifically designed for the classification of TB and pneumonia based on CXR images. The model integrates three core components: a multi-level knowledge distillation strategy aimed at transferring useful knowledge from a large teacher model to a compact encoder; a Triple Permuter (TP) module, which enhances spatial feature interactions without significantly increasing the parameter count; and a Split Multi-Head Self-Attention (SMHSA) module, which effectively reduces the computational complexity of the attention mechanism while preserving global contextual information. We evaluated the proposed method on four publicly available datasets: PX, K-4, TB, and SC. Experimental results demonstrate that LTS-Net achieves classification accuracies ranging from 96.47% to 99.95% across these datasets, while excelling at maintaining relatively low model size and computational cost. Collectively, these findings indicate that LTS-Net strikes a practical balance between diagnostic performance and operational efficiency, positioning it as a highly promising technical solution for automated chest X-ray (CXR) analysis in clinical environments with limited computational resources.