Focused Attention in Transformers for interpretable classification of retinal images

Ultrasound Image Analysis with Vision Transformers-Review

Ultrasound (US) has become a widely used imaging modality in clinical practice, characterized by its rapidly evolving technology, advantages, and unique challenges, such as a low imaging quality and high variability. There is a need to develop advanced automatic US image analysis methods to enhance its diagnostic accuracy and objectivity. Vision transformers, a recent innovation in machine learning, have demonstrated significant potential in various research fields, including general image analysis and computer vision, due to their capacity to process large datasets and learn complex patterns. Their suitability for automatic US image analysis tasks, such as classification, detection, and segmentation, has been recognized. This review provides an introduction to vision transformers and discusses their applications in specific US image analysis tasks, while also addressing the open challenges and potential future trends in their application in medical US image analysis. Vision transformers have shown promise in enhancing the accuracy and efficiency of ultrasound image analysis and are expected to play an increasingly important role in the diagnosis and treatment of medical conditions using ultrasound imaging as technology progresses.

Glaucoma detection model by exploiting multi-region and multi-scan-pattern OCT images with dynamical region score

Currently, deep learning-based methods have achieved success in glaucoma detection. However, most models focus on OCT images captured by a single scan pattern within a given region, holding the high risk of the omission of valuable features in the remaining regions or scan patterns. Therefore, we proposed a multi-region and multi-scan-pattern fusion model to address this issue. Our proposed model exploits comprehensive OCT images from three fundus anatomical regions (macular, middle, and optic nerve head regions) being captured by four scan patterns (radial, volume, single-line, and circular scan patterns). Moreover, to enhance the efficacy of integrating features across various scan patterns within a region and multiple regional features, we employed an attention multi-scan fusion module and an attention multi-region fusion module that auto-assign contribution to distinct scan-pattern features and region features adapting to characters of different samples, respectively. To alleviate the absence of available datasets, we have collected a specific dataset (MRMSG-OCT) comprising OCT images captured by four scan patterns from three regions. The experimental results and visualized feature maps both demonstrate that our proposed model achieves superior performance against the single scan-pattern models and single region-based models. Moreover, compared with the average fusion strategy, our proposed fusion modules yield superior performance, particularly reversing the performance degradation observed in some models relying on fixed weights, validating the efficacy of the proposed dynamic region scores adapted to different samples. Moreover, the derived region contribution scores enhance the interpretability of the model and offer an overview of the model's decision-making process, assisting ophthalmologists in prioritizing regions with heightened scores and increasing efficiency in clinical practice.

Automated detection of nine infantile fundus diseases and conditions in retinal images using a deep learning system

Purpose

We developed an Infant Retinal Intelligent Diagnosis System (IRIDS), an automated system to aid early diagnosis and monitoring of infantile fundus diseases and health conditions to satisfy urgent needs of ophthalmologists.

Methods

We developed IRIDS by combining convolutional neural networks and transformer structures, using a dataset of 7697 retinal images (1089 infants) from four hospitals. It identifies nine fundus diseases and conditions, namely, retinopathy of prematurity (ROP) (mild ROP, moderate ROP, and severe ROP), retinoblastoma (RB), retinitis pigmentosa (RP), Coats disease, coloboma of the choroid, congenital retinal fold (CRF), and normal. IRIDS also includes depth attention modules, ResNet-18 (Res-18), and Multi-Axis Vision Transformer (MaxViT). Performance was compared to that of ophthalmologists using 450 retinal images. The IRIDS employed a five-fold cross-validation approach to generate the classification results.

Results

Several baseline models achieved the following metrics: accuracy, precision, recall, F1-score (F1), kappa, and area under the receiver operating characteristic curve (AUC) with best values of 94.62% (95% CI, 94.34%-94.90%), 94.07% (95% CI, 93.32%-94.82%), 90.56% (95% CI, 88.64%-92.48%), 92.34% (95% CI, 91.87%-92.81%), 91.15% (95% CI, 90.37%-91.93%), and 99.08% (95% CI, 99.07%-99.09%), respectively. In comparison, IRIDS showed promising results compared to ophthalmologists, demonstrating an average accuracy, precision, recall, F1, kappa, and AUC of 96.45% (95% CI, 96.37%-96.53%), 95.86% (95% CI, 94.56%-97.16%), 94.37% (95% CI, 93.95%-94.79%), 95.03% (95% CI, 94.45%-95.61%), 94.43% (95% CI, 93.96%-94.90%), and 99.51% (95% CI, 99.51%-99.51%), respectively, in multi-label classification on the test dataset, utilizing the Res-18 and MaxViT models. These results suggest that, particularly in terms of AUC, IRIDS achieved performance that warrants further investigation for the detection of retinal abnormalities.

Conclusions

IRIDS identifies nine infantile fundus diseases and conditions accurately. It may aid non-ophthalmologist personnel in underserved areas in infantile fundus disease screening. Thus, preventing severe complications. The IRIDS serves as an example of artificial intelligence integration into ophthalmology to achieve better outcomes in predictive, preventive, and personalized medicine (PPPM / 3PM) in the treatment of infantile fundus diseases.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13167-024-00350-y.

WBC YOLO-ViT: 2 Way - 2 stage white blood cell detection and classification with a combination of YOLOv5 and vision transformer

Accurate detection and classification of white blood cells, otherwise known as leukocytes, play a critical role in diagnosing and monitoring various illnesses. However, conventional methods, such as manual classification by trained professionals, must be revised in terms of accuracy, efficiency, and potential bias. Moreover, applying deep learning techniques to detect and classify white blood cells using microscopic images is challenging owing to limited data, resolution noise, irregular shapes, and varying colors from different sources. This study presents a novel approach integrating object detection and classification for numerous type-white blood cell. We designed a 2-way approach to use two types of images: WBC and nucleus. YOLO (fast object detection) and ViT (powerful image representation capabilities) are effectively integrated into 16 classes. The proposed model demonstrates an exceptional 96.449% accuracy rate in classification.

Unsupervised Domain Adaptive Dose Prediction via Cross-Attention Transformer and Target-Specific Knowledge Preservation

Radiotherapy is one of the leading treatments for cancer. To accelerate the implementation of radiotherapy in clinic, various deep learning-based methods have been developed for automatic dose prediction. However, the effectiveness of these methods heavily relies on the availability of a substantial amount of data with labels, i.e. the dose distribution maps, which cost dosimetrists considerable time and effort to acquire. For cancers of low-incidence, such as cervical cancer, it is often a luxury to collect an adequate amount of labeled data to train a well-performing deep learning (DL) model. To mitigate this problem, in this paper, we resort to the unsupervised domain adaptation (UDA) strategy to achieve accurate dose prediction for cervical cancer (target domain) by leveraging the well-labeled high-incidence rectal cancer (source domain). Specifically, we introduce the cross-attention mechanism to learn the domain-invariant features and develop a cross-attention transformer-based encoder to align the two different cancer domains. Meanwhile, to preserve the target-specific knowledge, we employ multiple domain classifiers to enforce the network to extract more discriminative target features. In addition, we employ two independent convolutional neural network (CNN) decoders to compensate for the lack of spatial inductive bias in the pure transformer and generate accurate dose maps for both domains. Furthermore, to enhance the performance, two additional losses, i.e. a knowledge distillation loss (KDL) and a domain classification loss (DCL), are incorporated to transfer the domain-invariant features while preserving domain-specific information. Experimental results on a rectal cancer dataset and a cervical cancer dataset have demonstrated that our method achieves the best quantitative results with [Formula: see text], [Formula: see text], and HI of 1.446, 1.231, and 0.082, respectively, and outperforms other methods in terms of qualitative assessment.

Multi-Dataset Comparison of Vision Transformers and Convolutional Neural Networks for Detecting Glaucomatous Optic Neuropathy from Fundus Photographs

Glaucomatous optic neuropathy (GON) can be diagnosed and monitored using fundus photography, a widely available and low-cost approach already adopted for automated screening of ophthalmic diseases such as diabetic retinopathy. Despite this, the lack of validated early screening approaches remains a major obstacle in the prevention of glaucoma-related blindness. Deep learning models have gained significant interest as potential solutions, as these models offer objective and high-throughput methods for processing image-based medical data. While convolutional neural networks (CNN) have been widely utilized for these purposes, more recent advances in the application of Transformer architectures have led to new models, including Vision Transformer (ViT,) that have shown promise in many domains of image analysis. However, previous comparisons of these two architectures have not sufficiently compared models side-by-side with more than a single dataset, making it unclear which model is more generalizable or performs better in different clinical contexts. Our purpose is to investigate comparable ViT and CNN models tasked with GON detection from fundus photos and highlight their respective strengths and weaknesses. We train CNN and ViT models on six unrelated, publicly available databases and compare their performance using well-established statistics including AUC, sensitivity, and specificity. Our results indicate that ViT models often show superior performance when compared with a similarly trained CNN model, particularly when non-glaucomatous images are over-represented in a given dataset. We discuss the clinical implications of these findings and suggest that ViT can further the development of accurate and scalable GON detection for this leading cause of irreversible blindness worldwide.

Automatic prediction of acute coronary syndrome based on pericoronary adipose tissue and atherosclerotic plaques

Cardiovascular disease is the leading cause of human death worldwide, and acute coronary syndrome (ACS) is a common first manifestation of this. Studies have shown that pericoronary adipose tissue (PCAT) computed tomography (CT) attenuation and atherosclerotic plaque characteristics can be used to predict future adverse ACS events. However, radiomics-based methods have limitations in extracting features of PCAT and atherosclerotic plaques. Therefore, we propose a hybrid deep learning framework capable of extracting coronary CT angiography (CCTA) imaging features of both PCAT and atherosclerotic plaques for ACS prediction. The framework designs a two-stream CNN feature extraction (TSCFE) module to extract the features of PCAT and atherosclerotic plaques, respectively, and a channel feature fusion (CFF) to explore feature correlations between their features. Specifically, a trilinear-based fully-connected (FC) prediction module stepwise maps high-dimensional representations to low-dimensional label spaces. The framework was validated in retrospectively collected suspected coronary artery disease cases examined by CCTA. The prediction accuracy, sensitivity, specificity, and area under curve (AUC) are all higher than the classical image classification networks and state-of-the-art medical image classification methods. The experimental results show that the proposed method can effectively and accurately extract CCTA imaging features of PCAT and atherosclerotic plaques and explore the feature correlations to produce impressive performance. Thus, it has the potential value to be applied in clinical applications for accurate ACS prediction.

SF2Former: Amyotrophic lateral sclerosis identification from multi-center MRI data using spatial and frequency fusion transformer

Amyotrophic Lateral Sclerosis (ALS) is a complex neurodegenerative disorder characterized by motor neuron degeneration. Significant research has begun to establish brain magnetic resonance imaging (MRI) as a potential biomarker to diagnose and monitor the state of the disease. Deep learning has emerged as a prominent class of machine learning algorithms in computer vision and has shown successful applications in various medical image analysis tasks. However, deep learning methods applied to neuroimaging have not achieved superior performance in classifying ALS patients from healthy controls due to insignificant structural changes correlated with pathological features. Thus, a critical challenge in deep models is to identify discriminative features from limited training data. To address this challenge, this study introduces a framework called SF2Former, which leverages the power of the vision transformer architecture to distinguish ALS subjects from the control group by exploiting the long-range relationships among image features. Additionally, spatial and frequency domain information is combined to enhance the network's performance, as MRI scans are initially captured in the frequency domain and then converted to the spatial domain. The proposed framework is trained using a series of consecutive coronal slices and utilizes pre-trained weights from ImageNet through transfer learning. Finally, a majority voting scheme is employed on the coronal slices of each subject to generate the final classification decision. The proposed architecture is extensively evaluated with multi-modal neuroimaging data (i.e., T1-weighted, R2*, FLAIR) using two well-organized versions of the Canadian ALS Neuroimaging Consortium (CALSNIC) multi-center datasets. The experimental results demonstrate the superiority of the proposed strategy in terms of classification accuracy compared to several popular deep learning-based techniques.

Retinal Disease Detection Using Deep Learning Techniques: A Comprehensive Review

Millions of people are affected by retinal abnormalities worldwide. Early detection and treatment of these abnormalities could arrest further progression, saving multitudes from avoidable blindness. Manual disease detection is time-consuming, tedious and lacks repeatability. There have been efforts to automate ocular disease detection, riding on the successes of the application of Deep Convolutional Neural Networks (DCNNs) and vision transformers (ViTs) for Computer-Aided Diagnosis (CAD). These models have performed well, however, there remain challenges owing to the complex nature of retinal lesions. This work reviews the most common retinal pathologies, provides an overview of prevalent imaging modalities and presents a critical evaluation of current deep-learning research for the detection and grading of glaucoma, diabetic retinopathy, Age-Related Macular Degeneration and multiple retinal diseases. The work concluded that CAD, through deep learning, will increasingly be vital as an assistive technology. As future work, there is a need to explore the potential impact of using ensemble CNN architectures in multiclass, multilabel tasks. Efforts should also be expended on the improvement of model explainability to win the trust of clinicians and patients.

FSA-Net: Rethinking the attention mechanisms in medical image segmentation from releasing global suppressed information

Attention mechanism-based medical image segmentation methods have developed rapidly recently. For the attention mechanisms, it is crucial to accurately capture the distribution weights of the effective features contained in the data. To accomplish this task, most attention mechanisms prefer using the global squeezing approach. However, it will lead to a problem of over-focusing on the global most salient effective features of the region of interest, while suppressing the secondary salient ones. Making partial fine-grained features are abandoned directly. To address this issue, we propose to use a multiple-local perception method to aggregate global effective features, and design a fine-grained medical image segmentation network, named FSA-Net. This network consists of two key components: 1) the novel Separable Attention Mechanisms which replace global squeezing with local squeezing to release the suppressed secondary salient effective features. 2) a Multi-Attention Aggregator (MAA) which can fuse multi-level attention to efficiently aggregate task-relevant semantic information. We conduct extensive experimental evaluations on five publicly available medical image segmentation datasets: MoNuSeg, COVID-19-CT100, GlaS, CVC-ClinicDB, ISIC2018, and DRIVE datasets. Experimental results show that FSA-Net outperforms state-of-the-art methods in medical image segmentation.

A vision transformer architecture for the automated segmentation of retinal lesions in spectral domain optical coherence tomography images

Neovascular age-related macular degeneration (nAMD) is one of the major causes of irreversible blindness and is characterized by accumulations of different lesions inside the retina. AMD biomarkers enable experts to grade the AMD and could be used for therapy prognosis and individualized treatment decisions. In particular, intra-retinal fluid (IRF), sub-retinal fluid (SRF), and pigment epithelium detachment (PED) are prominent biomarkers for grading neovascular AMD. Spectral-domain optical coherence tomography (SD-OCT) revolutionized nAMD early diagnosis by providing cross-sectional images of the retina. Automatic segmentation and quantification of IRF, SRF, and PED in SD-OCT images can be extremely useful for clinical decision-making. Despite the excellent performance of convolutional neural network (CNN)-based methods, the task still presents some challenges due to relevant variations in the location, size, shape, and texture of the lesions. This work adopts a transformer-based method to automatically segment retinal lesion from SD-OCT images and qualitatively and quantitatively evaluate its performance against CNN-based methods. The method combines the efficient long-range feature extraction and aggregation capabilities of Vision Transformers with data-efficient training of CNNs. The proposed method was tested on a private dataset containing 3842 2-dimensional SD-OCT retina images, manually labeled by experts of the Franziskus Eye-Center, Muenster. While one of the competitors presents a better performance in terms of Dice score, the proposed method is significantly less computationally expensive. Thus, future research will focus on the proposed network's architecture to increase its segmentation performance while maintaining its computational efficiency.

Classification for thyroid nodule using ViT with contrastive learning in ultrasound images

The lack of representative features between benign nodules, especially level 3 of Thyroid Imaging Reporting and Data System (TI-RADS), and malignant nodules limits diagnostic accuracy, leading to inconsistent interpretation, overdiagnosis, and unnecessary biopsies. We propose a Vision-Transformer-based (ViT) thyroid nodule classification model using contrast learning, called TC-ViT, to improve accuracy of diagnosis and specificity of biopsy recommendations. ViT can explore the global features of thyroid nodules well. Nodule images are used as ROI to enhance the local features of the ViT. Contrast learning can minimize the representation distance between nodules of the same category, enhance the representation consistency of global and local features, and achieve accurate diagnosis of TI-RADS 3 or malignant nodules. The test results achieve an accuracy of 86.9%. The evaluation metrics show that the network outperforms other classical deep learning-based networks in terms of classification performance. TC-ViT can achieve automatic classification of TI-RADS 3 and malignant nodules on ultrasound images. It can also be used as a key step in computer-aided diagnosis for comprehensive analysis and accurate diagnosis. The code will be available at https://github.com/Jiawei217/TC-ViT.