A major contributor to permanent vision loss is glaucoma. Early diagnosis is crucial for preventing vision loss due to glaucoma, making glaucoma screening essential. A more affordable method of glaucoma screening can be achieved by applying artificial intelligence to evaluate color fundus photographs (CFPs). We present the Justified Referral in AI Glaucoma Screening (JustRAIGS) challenge to further develop these AI algorithms for glaucoma screening and to assess their efficacy. To support this challenge, we have generated a distinctive big dataset containing more than 110,000 meticulously labeled CFPs obtained from approximately 60,000 patients and 500 distinct screening centers in the USA. Our objective is to assess the practicality of creating advanced and dependable AI systems that can take a CFP as input and produce the probability of referable glaucoma, as well as outputs for glaucoma justification by integrating both binary and multi-label classification tasks. This paper presents the evaluation of solutions provided by nine teams, recognizing the team with the highest level of performance. The highest achieved score of sensitivity at a specificity level of 95% was 85%, and the highest achieved score of Hamming losses average was 0.13. Additionally, we test the top three participants' algorithms on an external dataset to validate the performance and generalization of these models. The outcomes of this research can offer valuable insights into the development of intelligent systems for detecting glaucoma. Ultimately, findings can aid in the early detection and treatment of glaucoma patients, hence decreasing preventable vision impairment and blindness caused by glaucoma.

Diabetic Retinopathy (DR), an ocular complication of diabetes, is a leading cause of blindness worldwide. Traditionally, DR is monitored using Color Fundus Photography (CFP), a widespread 2-D imaging modality. However, DR classifications based on CFP have poor predictive power, resulting in suboptimal DR management. Optical Coherence Tomography Angiography (OCTA) is a recent 3-D imaging modality offering enhanced structural and functional information (blood flow) with a wider field of view. This paper investigates automatic DR severity assessment using 3-D OCTA. A straightforward solution to this task is a 3-D neural network classifier. However, 3-D architectures have numerous parameters and typically require many training samples. A lighter solution consists in using 2-D neural network classifiers processing 2-D en-face (or frontal) projections and/or 2-D cross-sectional slices. Such an approach mimics the way ophthalmologists analyze OCTA acquisitions: (1) en-face flow maps are often used to detect avascular zones and neovascularization, and (2) cross-sectional slices are commonly analyzed to detect macular edemas, for instance. However, arbitrary data reduction or selection might result in information loss. Two complementary strategies are thus proposed to optimally summarize OCTA volumes with 2-D images: (1) a parametric en-face projection optimized through deep learning and (2) a cross-sectional slice selection process controlled through gradient-based attribution. The full summarization and DR classification pipeline is trained from end to end. The automatic 2-D summary can be displayed in a viewer or printed in a report to support the decision. We show that the proposed 2-D summarization and classification pipeline outperforms direct 3-D classification with the advantage of improved interpretability.

Vision Transformers (ViT)-based models are witnessing an exponential growth in the medical imaging community. Among desirable properties, ViTs provide a powerful modeling of long-range pixel relationships, contrary to inherently local convolutional neural networks (CNN). These emerging models can be categorized either as hybrid-based when used in conjunction with CNN layers (CNN-ViT) or purely Transformers-based. In this work, we conduct a comparative quantitative analysis to study the differences between a range of available Transformers-based models using controlled brain tumor segmentation experiments. We also investigate to what extent such models could benefit from modality interaction schemes in a multi-modal setting. Results on the publicly-available BraTS2021 dataset show that hybrid-based pipelines generally tend to outperform simple Transformers-based models. In these experiments, no particular improvement using multi-modal interaction schemes was observed.

The extraction of blood vessels has recently experienced a widespread interest in medical image analysis. Automatic vessel segmentation is highly desirable to guide clinicians in computer-assisted diagnosis, therapy or surgical planning. Despite a good ability to extract large anatomical structures, the capacity of U-Net inspired architectures to automatically delineate vascular systems remains a major issue, especially given the scarcity of existing datasets. In this paper, we present a novel approach that integrates into deep segmentation shape priors from a Semi-Overcomplete Convolutional Auto-Encoder (S-OCAE) embedding. Compared to standard Convolutional Auto-Encoders (CAE), it exploits an over-complete branch that projects data onto higher dimensions to better characterize tiny structures. Experiments on retinal and liver vessel extraction, respectively performed on publicly-available DRIVE and 3D-IRCADb datasets, highlight the effectiveness of our method compared to U-Net trained without and with shape priors from a traditional CAE.

Automatic segmentation of abdominal organs in CT scans plays an important role in clinical practice. However, most existing benchmarks and datasets only focus on segmentation accuracy, while the model efficiency and its accuracy on the testing cases from different medical centers have not been evaluated. To comprehensively benchmark abdominal organ segmentation methods, we organized the first Fast and Low GPU memory Abdominal oRgan sEgmentation (FLARE) challenge, where the segmentation methods were encouraged to achieve high accuracy on the testing cases from different medical centers, fast inference speed, and low GPU memory consumption, simultaneously. The winning method surpassed the existing state-of-the-art method, achieving a 19× faster inference speed and reducing the GPU memory consumption by 60% with comparable accuracy. We provide a summary of the top methods, make their code and Docker containers publicly available, and give practical suggestions on building accurate and efficient abdominal organ segmentation models. The FLARE challenge remains open for future submissions through a live platform for benchmarking further methodology developments at https://flare.grand-challenge.org/.

Automatic segmentation of magnetic resonance (MR) images is crucial for morphological evaluation of the pediatric musculoskeletal system in clinical practice. However, the accuracy and generalization performance of individual segmentation models are limited due to the restricted amount of annotated pediatric data. Hence, we propose to train a segmentation model on multiple datasets, arising from different parts of the anatomy, in a multi-task and multi-domain learning framework. This approach allows to overcome the inherent scarcity of pediatric data while benefiting from a more robust shared representation. The proposed segmentation network comprises shared convolutional filters, domain-specific batch normalization parameters that compute the respective dataset statistics and a domain-specific segmentation layer. Furthermore, a supervised contrastive regularization is integrated to further improve generalization capabilities, by promoting intra-domain similarity and impose inter-domain margins in embedded space. We evaluate our contributions on two pediatric imaging datasets of the ankle and shoulder joints for bone segmentation. Results demonstrate that the proposed model outperforms state-of-the-art approaches.

In this paper, we propose a new Conditional Region Growing (CRG) approach with the ability of finding the accurate MC boundaries starting from selected seed points. The starting seed points are determined based on regional maxima detection and superpixel analysis. The region growing step is controlled by a set of criteria derived from prior knowledge to characterize MCs. The key feature is to highlight below each MC to estimate the appropriate criteria and not to use the same parameters for all of them. Defined criteria can be divided into two categories. The first one concerns the neighbourhood searching size. The second one deals with the gradient information and shape evolution within the growing process. Experimental results show the benefits of used criteria in terms of improving the MC delineation qualities.

We analyze the problem of how to correctly construct dense point trajectories from optical flow fields. First, we show that simple Euler integration is unavoidably inaccurate, no matter how good is the optical flow estimator. Then, an inverse integration scheme is analyzed which is more robust to bias and input noise and shows better stability properties. Our contribution is threefold: 1) a theoretical analysis that demonstrates why and in what sense inverse integration is more accurate; 2) a rich experimental validation both on synthetic and real (image) data; and 3) an algorithm for approximate online inverse integration. This new technique is precious whether one is trying to propagate information densely available on a reference frame to the other frames in the sequence or, conversely, to assign information densely over each frame by pulling it from the reference.

Accurate estimation of dense point correspondences between two distant frames of a video sequence is a challenging task. To address this problem, we present a combinatorial multistep integration procedure which allows one to obtain a large set of candidate motion fields between the two distant frames by considering multiple motion paths across the video sequence. Given this large candidate set, we propose to perform the optimal motion vector selection by combining a global optimization stage with a new statistical processing. Instead of considering a selection only based on intrinsic motion field quality and spatial regularization, the statistical processing exploits the spatial distribution of candidates and introduces an intra-candidate quality based on forward-backward consistency. Experiments evaluate the effectiveness of our method for distant motion estimation in the context of video editing.

International audience

Dense point matching and tracking in image sequences is an open issue with implications in several domains, from content analysis to video editing. We observe that for long term dense point matching, some regions of the image are better matched by concatenation of consecutive motion vectors, while for others a direct long term matching is preferred. We propose a method to optimally estimate the correspondence of a point w.r.t. a reference image from a set of input motion estimations over different temporal intervals. Results on texture insertion by point tracking in the context of video editing are presented and compared with a state-of-the-art approach.