MHANet: A hybrid attention mechanism for retinal diseases classification

Abstract

With the increase of patients with retinopathy, retinopathy recognition has become a research hotspot. In this article, we describe the etiology and symptoms of three kinds of retinal diseases, including drusen(DRUSEN), choroidal neovascularization(CNV) and diabetic macular edema(DME). In addition, we also propose a hybrid attention mechanism to classify and recognize different types of retinopathy images. In particular, the hybrid attention mechanism proposed in this paper includes parallel spatial attention mechanism and channel attention mechanism. It can extract the key features in the channel dimension and spatial dimension of retinopathy images, and reduce the negative impact of background information on classification results. The experimental results show that the hybrid attention mechanism proposed in this paper can better assist the network to focus on extracting thr fetures of the retinopathy area and enhance the adaptability to the differences of different data sets. Finally, the hybrid attention mechanism achieved 96.5% and 99.76% classification accuracy on two public OCT data sets of retinopathy, respectively.

Effective automatic detection of anterior cruciate ligament injury using convolutional neural network with two attention mechanism modules

“…In medical imaging, the ATM can help neural networks focus on key information when processing large amounts of medical image data. As a result, the ATM has been applied to the medical field, including in assisting clinicians with the identification of melanoma [18], retinal lesions [19], the pathological sections of colorectal cancer [20] and breast cancer [21], and it has achieved excellent results. Thus, we believe that adding ATM to our CNN model can effectively improve In short, when the ATM was computing, the input medical image feature was defined as Q(queries), K(keys), and V(values).…”

Luminosity Normalized Symmetric Deep Convolute Tubular Classifier for Diabetic Retinopathy Detection

“…However, the negative impact of background information on classification results was not analyzed. To concentrate on this aspect, a hybrid attention mechanism was employed in [25], with both spatial and channel attention mechanisms, to minimize the negative impact. Over the last few years, wavelet-based DR detection has been found to have greater influence, and perhaps the Deep Learning models, like CNN's, have progressed in contributing the highest prediction accuracy.…”

Enhancing Readability and Detection of Age-Related Macular Degeneration Using Optical Coherence Tomography Imaging: An AI Approach

“…They also compared their model with other methods such as MCME Studies in the classification paradigm explore the integration of attention mechanisms to enhance AMD classification accuracy. Xu et al proposed a hybrid attention mechanism for retinal disease classification using OCT images [16]. Their method, called MHANet, combines parallel spatial and channel attention mechanisms to extract key features of the lesion areas and reduce the influence of background information.…”

“…They reported that their method achieved 96.5% and 99.76% classification accuracy on Dataset1 and Dataset2, respectively, outperforming the other recent employed models such as VGG, ResNet, and SENet. They also visualized the attention maps of their method and showed that it can more accurately locate the lesion regions in the images [16]. Li et al have proposed two convolutional neural network (CNN) models to classify four types of age-related macular degeneration (AMD) from retinal images [17].…”

“…Santos et al [20] introduced a methodology using geostatistical functions for diagnosing AMD from OCT images, showing high accuracy and AUROC. Additionally, Wang and colleagues [16,18] also developed a novel method for DME detection using labeled and unlabeled data, which demonstrated superior performance according to various metrics. Lastly, the paper introduced a VGG-19 based AI model for classifying OCT retinal images, trained on a large and diverse dataset.…”