• Glaucoma is one of the leading causes of irreversible blindness worldwide. • Diagnosing it in highly myopic patients is challenging due to anatomical alterations. • The domain-specific transfer learning method proposed improves model generalization. • Helps differentiate true glaucomatous patterns from high myopia structural changes. • First designed to detect glaucoma in highly myopic patients using only fundus images. Glaucoma is a leading cause of irreversible blindness worldwide, with early detection being essential to preventing vision loss. However, diagnosing glaucoma in highly myopic patients poses significant challenges due to anatomical alterations, such as optic nerve head deformation and retinal nerve fiber layer thinning, potentially obscuring key disease features and causing misdiagnosis. To address these limitations, we propose a novel deep learning-based framework for automated glaucoma screening in highly myopic eyes. Our approach leverages multi-transfer learning, integrating large-scale pretraining with domain-specific adaptation using ophthalmic disease datasets. This methodology enables the model to extract robust and highly discriminative features, improving sensitivity to glaucomatous changes in myopic eyes. Additionally, we incorporate domain transfer pipeline to address the distributional differences between standard datasets and myopia-specific cases, further enhancing the model’s generalization capabilities. To rigorously evaluate our approach, we conduct a comprehensive analysis of state-of-the-art architectures and transfer learning strategies, assessing their impact on classification performance. Experimental results demonstrate that the proposed model consistently outperforms baseline methods, achieving superior accuracy and robustness in glaucoma detection within highly myopic populations. These findings underscore the potential of AI-driven screening tools as reliable and accurate diagnostic aids, supporting clinicians in the early detection and effective management of glaucoma in complex cases.

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.

Essential tremor (ET) is a prevalent movement disorder characterized by motor and non-motor symptoms, often associated with neurodegeneration. Optical coherence tomography (OCT) has emerged as a valuable tool to identify retinal biomarkers in ET patients. This study presents a novel methodology for ET detection using 3D point clouds derived from retinal OCT layers. Leveraging advanced geometric deep learning (GDL) architectures, including PointTransformer, PointCNN, PointNet++ and SplineCNN, we evaluated the diagnostic potential of individual retinal layers, including the Retinal Nerve Fiber Layer (RNFL), Ganglion Cell Layer (GCL) and Bruch’s Membrane (BM), as well as their combined representation. Our approach achieved state-of-the-art results, with PointTransformer obtaining an F1-score of 0.85 using only BM retinal surface, while requiring just 2% of the original point cloud size. These findings underscore the diagnostic value of OCT-derived 3D data and demonstrate the potential of GDL for computational biomarker extraction in neurodegenerative disorders, offering a scalable and efficient framework for ET diagnosis.

The automatic identification and segmentation of the Foveal Avascular Zone (FAZ) is crucial for diagnosing and monitoring retinal diseases. However, the limited availability of Optical Coherence Tomography Angiography (OCTA) images with ground truth annotations poses a significant challenge for developing robust deep learning models. Traditional transfer learning techniques, such as ImageNet-based pre-training, require large datasets and struggle to adapt to domain-specific tasks in medical imaging. To address this issue, we propose a novel multi-deep transfer learning framework that leverages pretrained models on a generation task using both superficial (SCP) and deep capillary plexuses (DCP) of the retina. To the best of our knowledge, this is the first approach that integrates depth information from multiple vascular plexuses, allowing the model to capture cross-plexus structural relationships and enhance its ability to learn domain-specific vascular features while mitigating data scarcity constraints. We validated our framework through extensive experiments, demonstrating competitive or even superior segmentation performance compared to state-of-the-art pretraining models, despite using significantly smaller datasets. Our approach achieved a Dice coefficient score of 0.8238 ± 0.0263, a Jaccard Index of 0.7189 ± 0.0293, and a Correlation Index of 0.8328 ± 0.0229, highlighting the effectiveness of our multi-depth feature learning strategy in improving segmentation precision and generalization, even with limited data availability.

The registered version of this article, first published in “Pattern Recognition Letters 189, 31-37, 2025", is available online at the publisher's website: Elsevier, https://doi.org/10.1016/j.patrec.2025.01.003 La versión registrada de este artículo, publicado por primera vez en “Pattern Recognition Letters 189, 31-37, 2025.", está disponible en línea en el sitio web del editor: Elsevier, https://doi.org/10.1016/j.patrec.2025.01.003

The detection of vessel crossings and bifurcations in eye fundus images plays an important role in numerous applications, including the diagnosis of ophthalmic and systemic diseases, biometric authentication, and retinal image registration. Nowadays, deep neural networks are successfully used for the detection of these vascular landmarks. However, existing approaches could be limited by the lack of understanding of the retinal anatomy and the intricate retinal vasculature. In this context, we propose Recurrent Task Specialization, a novel approach that performs a recurrent forward process with two forward passes through the same network, each of them specialized in a different task. We apply the proposed approach to the detection of vessel crossings and bifurcations in the retina via heatmap regression, using the segmentation of the retinal vasculature as the auxiliary task. To validate our proposal, we perform comparative experiments on two public datasets, including common alternatives to leverage auxiliary tasks, such as standard multi-task learning and transfer learning. The proposed approach outperforms existing alternatives and achieves the best results in the state-of-the-art for the detection of vessel crossings and bifurcations in retinal images. In this regard, our experiments demonstrate the potential of the proposed approach to improve the performance of deep neural networks in applications where adequate auxiliary tasks can be constructed.

Multiple Sclerosis (MS), the leading cause of non-traumatic neurological impairment in young adults, manifests morphological changes in the retina observable in Optical Coherence Tomography (OCT) images. These changes in the Retinal Nerve Fibre Layer (RNFL) and the Ganglion Cell Layer - Bruch’s Membrane (GCL-BM) serve as potential computational biomarkers for MS. In this work, we propose a transformer-based graph learning approach for analyzing 3D point clouds generated from RNFL and GCL-BM contours, marking a first in the application of geometric deep learning (GDL) to MS diagnosis via OCT scans. Our proposal, tailored for efficiency, synergizes the global contextual strengths of transformers with the detailed, structure-aware capabilities of graph neural networks. Such integration allows for the nuanced analysis of complex retinal structures, significantly boosting the precision of MS detection by uncovering patterns not discernible to the human eye. Additionally, we conducted a comprehensive study on the optimal downsampling size of input 3D point clouds, ensuring efficient data processing without compromising diagnostic accuracy. Our optimal configuration achieved a test F1-Score of 0.88, using only 4.0% of total 3D points, showcasing the effectiveness of our method despite the higher computational demands compared to less complex, albeit less precise, configurations. These promising results are the first in the study of 3D analysis and transformer-based geometric deep learning for MS screening based on OCT images, which are revolutionizing neurophtalmological research.

Drusen, small lipid deposits located below the retina, are early biomarkers of age-related macular degeneration (AMD), a condition that leads to visual impairment worldwide, especially among the elderly. The presence of AMD is linked with Alzheimer’s Disease (AD) and dense deposit disease (DDD), emphasizing the critical need for early and accurate detection of drusen in retinal tissues. Optical Coherence Tomography (OCT), with its non-invasive and high-resolution imaging capabilities, stands as a pivotal tool for early AMD diagnosis through the identification of drusen. However, the reliance on manual segmentation of drusen in 3D OCT images introduces significant challenges: it is not only time-consuming but also subject to inter-observer variability, severely constraining its efficacy for widespread screening applications. These limitations underscore the critical need for an automated solution that can improve diagnostic workflows, ensure consistency across interpretations, and facilitate the processing of large datasets with improved accuracy and efficiency. In response, we propose a new deep learning-based, fully-automatic end-to-end approach for the segmentation of drusen in 3D OCT volumes. Complementary, a pivotal aspect of our research involves the first detailed comparative analysis between 2D and 3D end-to-end approaches. This comparison is crucial for understanding the impact of dimensional spatial information on the accuracy of drusen segmentation within OCT scans. The findings demonstrate that the 3D approach, by leveraging the depth and complexity of spatial data available in 3D OCT volumes, markedly surpasses the 2D approach. This superior performance underlines the importance of 3D spatial information in enhancing diagnostic precision. Through automating the segmentation process, our proposal not only makes AMD screening more efficient and precise but also significantly advances the diagnosis of retinal diseases, potentially enriching our understanding of systemic conditions connected through retinal biomarkers.

The prevalence of neurodegenerative diseases (NDDs) such as Alzheimer's (AD), Parkinson's (PD), Essential tremor (ET), and Multiple Sclerosis (MS) is increasing alongside the aging population. Recent studies suggest that these disorders can be identified through retinal imaging, allowing for early detection and monitoring via Optical Coherence Tomography (OCT) scans. This study is at the forefront of research, pioneering the application of multi-view OCT and 3D information to the neurological diseases domain. Our methodology consists of two main steps. In the first one, we focus on the segmentation of the retinal nerve fiber layer (RNFL) and a class layer grouping between the ganglion cell layer and Bruch's membrane (GCL-BM) in both macular and optic disc OCT scans. These are the areas where changes in thickness serve as a potential indicator of NDDs. The second phase is to select patients based on information about the retinal layers. We explore how the integration of both views (macula and optic disc) improves each screening scenario: Healthy Controls (HC) vs. NDD, AD vs. NDD, ET vs. NDD, MS vs. NDD, PD vs. NDD, and a final multi-class approach considering all four NDDs. For the segmentation task, we obtained satisfactory results for both 2D and 3D approaches in macular segmentation, in which 3D performed better due to the inclusion of depth and cross-sectional information. As for the optic disc view, transfer learning did not improve the metrics over training from scratch, but it did provide a faster training. As for screening, 3D computational biomarkers provided better results than 2D ones, and multi-view methods were usually better than the single-view ones. Regarding separability among diseases, MS and PD were the ones that provided better results in their screening approaches, being also the most represented classes. In conclusion, our methodology has been successfully validated with an extensive experimentation of configurations, techniques and OCT views, becoming the first multi-view analysis that merges data from both macula-centered and optic disc-centered perspectives. Besides, it is also the first effort to examine key retinal layers across four major NDDs within the framework of pathological screening.

Multi-task learning is a promising paradigm to leverage task interrelations during the training of deep neural networks. A key challenge in the training of multi-task networks is to adequately balance the complementary supervisory signals of multiple tasks. In that regard, although several task-balancing approaches have been proposed, they are usually limited by the use of per-task weighting schemes and do not completely address the uneven contribution of the different tasks to the network training. In contrast to classical approaches, we propose a novel Multi-Adaptive Optimization (MAO) strategy that dynamically adjusts the contribution of each task to the training of each individual parameter in the network. This automatically produces a balanced learning across tasks and across parameters, throughout the whole training and for any number of tasks. To validate our proposal, we perform comparative experiments on real-world datasets for computer vision, considering different experimental settings. These experiments allow us to analyze the performance obtained in several multi-task scenarios along with the learning balance across tasks, network layers and training steps. The results demonstrate that MAO outperforms previous task-balancing alternatives. Additionally, the performed analyses provide insights that allow us to comprehend the advantages of this novel approach for multi-task learning.

Optical Coherence Tomography (OCT) is the major diagnostic tool for the leading cause of blindness in developed countries: Diabetic Macular Edema (DME). Depending on the type of fluid accumulations, different treatments are needed. In particular, Cystoid Macular Edemas (CMEs) represent the most severe scenario, while Diffuse Retinal Thickening (DRT) is an early indicator of the disease but a challenging scenario to detect. While methodologies exist, their explanatory power is limited to the input sample itself. However, due to the complexity of these accumulations, this may not be enough for a clinician to assess the validity of the classification. Thus, in this work, we propose a novel approach based on multi-prototype networks with vision transformers to obtain an example-based explainable classification. Our proposal achieved robust results in two representative OCT devices, with a mean accuracy of 0.9099 ± 0.0083 and 0.8582 ± 0.0126 for CME and DRT-type fluid accumulations, respectively.

COVID-19 is a challenging disease that was declared as global pandemic in March 2020. As the main impact of this disease is located in the pulmonary regions, chest X-ray devices are very useful to understand the severity of the disease on each patient. In order to reduce the risk of cross-contamination, the radiologists are recommended to use portable devices instead of fixed machinery, as these devices are easier to decontaminate. Moreover, the development of reliable and robust methodologies of computer-aided diagnosis systems is very relevant to reduce the workload that expert clinicians are experiencing in the current moment. In this work, we propose a comprehensive analysis of the deep features extracted from portable chest X-ray captures to perform a COVID-19 screening. We also study the optimal characterization of the problem with a lower dimensionality, contrasting the results of the feature selection methods that were chosen. Results demonstrated that the proposed approach is robust and reliable, obtaining a 90.43% of accuracy for the test set, using only 46.85% of the deep features in the context of poor quality and low detail X-ray images obtained from portable devices.

Patients at high risk of contracting COVID-19 require specialized monitoring throughout their illness to ensure optimal treatment at each stage. To support this monitoring, Computer-Aided Diagnosis (CAD) methods analyze clinical data to estimate the most likely outcome for each patient, using various clinical variables such as symptoms, medical history, and laboratory results to predict outcomes. Despite the numerous proposals for COVID-19 diagnosis using CAD methods, the lack of explainability in many machine learning models poses a challenge in incorporating these methods into clinical practice. Additionally, other crucial tasks such as estimating the risk of death or severe forms of the disease must be considered to identify cases that require greater monitoring. To overcome these challenges, we propose an explainable methodology for estimating the risk of hospitalization and death in COVID-19 patients using clinical data. Our methodology employs four machine learning algorithms, three feature selection methods, and a decision tree to provide explainability. Our approach achieves an accuracy of 86.16% ± 0.74% for the estimation of hospitalization risk with 29 features, and an accuracy of 86.40% ± 1.80% for the estimation of the risk of death with 26 features. Moreover, our methodology provides valuable insights into the relationship between clinical variables and patient outcomes, which can inform more robust and informed clinical decision-making and improve our understanding of the disease. We demonstrate the potential of our transparent and effective CAD methods to support clinical decision-making in COVID-19 patient care and further research, offering a promising solution to overcome the challenges in incorporating CAD methods into clinical practice.

Abstract Optical Coherence Tomography Angiography or OCTA represents one of the main means of diagnosis of Age-related Macular Degeneration (AMD), the leading cause of blindness in developed countries. This eye disease is characterized by Macular Neovascularization (MNV), the formation of vessels that tear through the retinal tissues. Four types of MNV can be distinguished, each representing different levels of severity. Both the aggressiveness of the treatment and the recovery of the patient rely on an early detection and correct diagnosis of the stage of the disease. In this work, we propose the first fully-automatic grading methodology that considers all the four clinical types of MNV at the three most relevant OCTA scanning depths for the diagnosis of AMD. We perform both a comprehensive ablation study on the contribution of said depths and an analysis of the attention maps of the network in collaboration with experts of the domain. Our proposal aims to ease the diagnosis burden and decrease the influence of subjectivity on it, offering a explainable grading through the visualization of the attention of the expert models. Our grading proposal achieved satisfactory results with an AUC of 0.9224 ± 0.0381. Additionally, the qualitative analysis performed in collaboration with experts revealed the relevance of the avascular plexus in the grading of all three types of MNV (despite not being directly involved in some of them). Thus, our proposal is not only able to robustly detect MNV in complex scenarios, but also aided to discover previously unconsidered relationships between plexuses.

Abstract There are different retinal vascular imaging modalities widely used in clinical practice to diagnose different retinal pathologies. The joint analysis of these multimodal images is of increasing interest since each of them provides common and complementary visual information. However, if we want to facilitate the comparison of two images, obtained with different techniques and containing the same retinal region of interest, it will be necessary to make a previous registration of both images. Here, we present a weakly supervised deep learning methodology for robust deformable registration of multimodal retinal images, which is applied to implement a method for the registration of fluorescein angiography (FA) and optical coherence tomography angiography (OCTA) images. This methodology is strongly inspired by VoxelMorph, a general unsupervised deep learning framework of the state of the art for deformable registration of unimodal medical images. The method was evaluated in a public dataset with 172 pairs of FA and superficial plexus OCTA images. The degree of alignment of the common information (blood vessels) and preservation of the non-common information (image background) in the transformed image were measured using the Dice coefficient (DC) and zero-normalized cross-correlation (ZNCC), respectively. The average values of the mentioned metrics, including the standard deviations, were DC = 0.72 ± 0.10 and ZNCC = 0.82 ± 0.04. The time required to obtain each pair of registered images was 0.12 s. These results outperform rigid and deformable registration methods with which our method was compared.

Retinal Optical Coherence Tomography (OCT) allows the non-invasive direct observation of the central nervous system, enabling the measurement and extraction of biomarkers from neural tissue that can be helpful in the assessment of ocular, systemic and Neurological Disorders (ND). Deep learning models can be trained to segment the retinal layers for biomarker extraction. However, the onset of ND can have an impact on the neural tissue, which can lead to the degraded performance of models not exposed to images displaying signs of disease during training. We present a fully automatic approach for the retinal layer segmentation in multiple neurodegenerative disorder scenarios, using an annotated dataset of patients of the most prevalent NDs: Alzheimer's disease, Parkinson's disease, multiple sclerosis and essential tremor, along with healthy control patients. Furthermore, we present a two-part, comprehensive study on the effects of ND on the performance of these models. The results show that images of healthy patients may not be sufficient for the robust training of automated segmentation models intended for the analysis of ND patients, and that using images representative of different NDs can increase the model performance. These results indicate that the presence or absence of patients of ND in datasets should be taken into account when training deep learning models for retinal layer segmentation, and that the proposed approach can provide a valuable tool for the robust and reliable diagnosis in multiple scenarios of ND.

The study of the ciliary muscle represents a fundamental step in the diagnosis and treatment of many high-incidence diseases, such as glaucoma or myopia. Currently, Anterior Segment Optical Coherence Tomography (AS-OCT) is widely used by clinicians to analyse the morphological changes that affect this important ocular structure. AS-OCT is a non-invasive imaging technique that produces high-resolution cross-sectional images, allowing a precise visualization of the main ocular tissues of the anterior segment of the eye. In this work, we propose a novel methodology for the ciliary muscle segmentation using AS-OCT images, an emerging ophthalmic imaging technology with great potential to support early diagnosis of relevant ocular conditions. For this purpose, we have analysed the performance of the U-Net architecture with two different encoders (ResNet-18 and ResNet-34) combined with a transfer learning-based approach. The validation of the proposed system was performed through different and representative experiments, using an AS-OCT dataset that was specifically designed for this work. The results demonstrated that the proposed system is robust and reliable, achieving an average Precision of 0.8902 ± 0.0815, an average Recall of 0.8237 ± 0.1239, an average Accuracy of 0.9961 ± 0.0021, an average Jaccard of 0.7431 ± 0.1116 and an average Dice of 0.8445 ± 0.0870. These results demonstrate that the proposed method has a satisfactory performance that can help the clinicians to make a more accurate diagnosis and proceed with appropriate treatments of different diseases of interest.

Abstract Currently, optical coherence tomography is one of the most used medical imaging modalities, offering cross-sectional representations of the studied tissues. This image modality is specially relevant for the analysis of the retina, since it is the internal part of the human body that allows an almost direct examination without invasive techniques. One of the most representative cases of use of this medical imaging modality is for the identification and characterization of intraretinal fluid accumulations, critical for the diagnosis of one of the main causes of blindness in developed countries: the Diabetic Macular Edema. The study of these fluid accumulations is particularly interesting, both from the point of view of pattern recognition and from the different branches of health sciences. As these fluid accumulations are intermingled with retinal tissues, they present numerous variants according to their severity, and change their appearance depending on the configuration of the device; they are a perfect subject for an in-depth research, as they are considered to be a problem without a strict solution. In this work, we propose a comprehensive and detailed analysis of the patterns that characterize them. We employed a pool of 11 different texture and intensity feature families (giving a total of 510 markers) which we have analyzed using three different feature selection strategies and seven complementary classification algorithms. By doing so, we have been able to narrow down and explain the factors affecting this kind of accumulations and tissue lesions by means of machine learning techniques with a pipeline specially designed for this purpose.

Tuberculosis is an infectious disease that mainly affects the lung tissues. Therefore, chest X-ray imaging can be very useful to diagnose and to understand the evolution of the pathology. This image modality has a poorer quality in contrast with other techniques as the magnetic resonance or the computerized tomography, but chest X-ray is easier and cheaper to perform. Furthermore, data scarcity is challenging in the domain of biomedical imaging. In order to mitigate this problem, the use of Generative Adversarial Network models for image generation has proved to be a powerful approach to train the deep learning models with small datasets, representing an alternative to classic data augmentation strategies. In this work, we propose a fully automatic approach for the generation of novel synthetic chest X-ray images to mitigate the effect of data scarcity in order to improve the tuberculosis screening performance using 3 different publicly available representative datasets: Montgomery County, Shenzhen and TBX11K. Firstly, this approach trains image translation models with a large-sized dataset (TBX11K). Then, these models are used to generate the novel set of synthetic images using small-sized and medium-sized datasets (Montgomery County and Shenzhen, respectively). Finally, the novel set of generated images is added to the training set to improve the performance of an automatic tuberculosis screening. As a result, we obtained an 88.41% ± 5.27% of accuracy for the Montgomery County dataset and a 90.33% ± 1.41% for the Shenzhen dataset. These results demonstrate that the proposed method outperforms previous state-of-the-art approaches.

Optical Coherence Tomography Angiography (OCTA) images represent the newest ophthalmic image modality with a great potential. They are characterized by allowing the precise non-invasive visualization of the retinal vasculature at different depths and resolutions. Given the relevance of the vascular analysis in this novel image modality, we propose the automatic segmentation and measurement of the existing ischemic areas that represent damaged regions of the eye fundus. In particular, we designed a methodology based in three main steps: image preprocessing, initial segmentation of ischemic regions by using active contours and final segmentation refinement. We obtained satisfactory results in the validation process, with a high separability between defined ischemic degrees in retinal vein occlusion patients. This fully automatic tool, with objectivity and reproducibility, provides an accurate measurement of the ischemic regions in real time, considering an issue where the manual analysis of the specialist is unfeasible in clinical practice routine.

Central serous chorioretinopathy is one of the most frequent causes of vision impairment among middle-aged adults. Optical Coherence Tomography (OCT) is a non-invasive diagnostic technique that is commonly used for the monitoring of this relevant eye disease. In this context, this paper proposes a fully automatic system for the characterization of intraretinal pathological fluid regions associated with central serous chori-oretinopathy using OCT scans. To achieve this, we adapted an end-to-end fully convolutional architecture for semantic pixel-wise segmentation. The proposed methodology was tested using a heterogeneous set of 100 OCT scans of different patients. Satisfactory results were obtained, reaching values of 0.9954±0.0007, 0.8792 ±0.0079 and 0.9651 ±0.0041 for the mean Accuracy, mean Jaccard index and mean Dice coefficient, respectively. The proposed system also demonstrated its competitive performance with respect to other state-of-the-art approaches.

Nowadays, Diabetic Macular Edema (DME) is one of the leading causes of blindness in developed countries, and its characterized by the presence of pathological fluid accumulations inside the retinal layers. Currently, the main way to detect these fluid accumulations (as well as their severity) is through the use of Optical Coherence Tomography (OCT) imaging. In particular, this ophthalmological image modality allows a precise non-invasive analysis of the morphology of the retina and its structures. Due to the complexity of attempting to successfully segment these fluid accumulations, an alternative paradigm for their detection has been recently proposed. This paradigm, based on a diffuse representation of the pathological regions, creates an intuitive representation of the pathological regions based on a confidence map. Currently, there are only two approaches for this paradigm: one based on a predefined library of texture and intensity features with established machine learning algorithms and other based on deep learning methods. Both approaches have proven to offer satisfactory results, but each one of them performs better in different scenarios. In this work, we perform a complete analysis and comparison on the behaviour and performance of both strategies in a clinical screening scenario to evaluate the suitability of both approaches for the clinical practice as well as their performance as computer vision strategies.

Computer-aided diagnosis plays an important role in the COVID-19 pandemic. Currently, it is recommended to use X-ray imaging to diagnose and assess the evolution in patients. Particularly, radiologists are asked to use portable acquisition devices to minimize the risk of cross-infection, facilitating an effective separation of suspected patients with other low-risk cases. In this work, we present an automatic COVID-19 screening, considering 6 representative state-of-the-art deep network architectures on a portable chest X-ray dataset that was specifically designed for this proposal. Exhaustive experimentation demonstrates that the models can separate COVID-19 cases from NON-COVID-19 cases, achieving a 97.68% of global accuracy.

Coronavirus Disease 2019 (COVID-19), declared a global pandemic by the World Health Organization, mainly affects the pulmonary tissues, playing chest X-ray images an important role for its screening and early detection. In this context, portable X-ray devices are widely used, representing an alternative to fixed devices in order to reduce risks of cross-contamination. However, they provide lower quality and detailed images in terms of spatial resolution and contrast. In this work, given the low availability of images of this recent disease, we present new approaches to artificially increase the dimensionality of portable chest X-ray datasets for COVID-19 diagnosis. Hence, we combined 3 complementary CycleGAN architectures to perform a simultaneous oversampling using an unsupervised strategy and without the necessity of paired data. Despite the poor quality of the portable X-ray images, we provide an overall accuracy of 92.50% in a COVID-19 screening context, proving their suitability for COVID-19 diagnostic tasks.

The broad availability of medical images in current clinical practice provides a source of large image datasets. In order to use these datasets for training deep neural networks in detection and segmentation tools, it is necessary to provide pixel-wise annotations associated to each image. However, the image annotation is a tedious, time consuming and error prone process that requires the participation of experienced specialists. In this work, we propose different complementary context encoder self-supervised approaches to learn relevant characteristics for the restricted medical imaging domain of retinographies. In particular, we propose a patch-wise approach, inspired in the previous proposal of broad domain context encoders, and complementary fully convolutional approaches. These approaches take advantage of the restricted application domain to learn the relevant features of the eye fundus, situation that can be extrapolated to many medical imaging issues. Different representative experiments were conducted in order to evaluate the performance of the trained models, demonstrating the suitability of the proposed approaches in the understanding of the eye fundus characteristics. The proposed self-supervised models can serve as reference to support other domain-related issues through transfer or multi-task learning paradigms, like the detection and evaluation of the retinal structures or anomaly detections in the context of pathological analysis.

Computer-aided diagnosis using retinal fundus images is crucial for the early detection of many ocular and systemic diseases. Nowadays, deep learning-based approaches are commonly used for this purpose. However, training deep neural networks usually requires a large amount of annotated data, which is not always available. In practice, this issue is commonly mitigated with different techniques, such as data augmentation or transfer learning. Nevertheless, the latter is typically faced using networks that were pre-trained on additional annotated data. An emerging alternative to the traditional transfer learning source tasks is the use of self-supervised tasks that do not require manually annotated data for training. In that regard, we propose a novel self-supervised visual learning strategy for improving the retinal computer-aided diagnosis systems using unlabeled multimodal data. In particular, we explore the use of a multimodal reconstruction task between complementary retinal imaging modalities. This allows to take advantage of existent unlabeled multimodal data in the medical domain, improving the diagnosis of different ocular diseases with additional domain-specific knowledge that does not rely on manual annotation. To validate and analyze the proposed approach, we performed several experiments aiming at the diagnosis of different diseases, including two of the most prevalent impairing ocular disorders: glaucoma and age-related macular degeneration. Additionally, the advantages of the proposed approach are clearly demonstrated in the comparisons that we perform against both the common fully-supervised approaches in the literature as well as current self-supervised alternatives for retinal computer-aided diagnosis. In general, the results show a satisfactory performance of our proposal, which improves existing alternatives by leveraging the unlabeled multimodal visual data that is commonly available in the medical field.

The current COVID-19 pandemic, that has caused more than 100 million cases as well as more than two million deaths worldwide, demands the development of fast and accurate diagnostic methods despite the lack of available samples. This disease mainly affects the respiratory system of the patients and can lead to pneumonia and to severe cases of acute respiratory syndrome that result in the formation of several pathological structures in the lungs. These pathological structures can be explored taking advantage of chest X-ray imaging. As a recommendation for the health services, portable chest X-ray devices should be used instead of conventional fixed machinery, in order to prevent the spread of the pathogen. However, portable devices present several problems (specially those related with capture quality). Moreover, the subjectivity and the fatigue of the clinicians lead to a very difficult diagnostic process. To overcome that, computer-aided methodologies can be very useful even taking into account the lack of available samples that the COVID-19 affectation shows. In this work, we propose an improvement in the performance of COVID-19 screening, taking advantage of several cycle generative adversarial networks to generate useful and relevant synthetic images to solve the lack of COVID-19 samples, in the context of poor quality and low detail datasets obtained from portable devices. For validating this proposal for improved COVID-19 screening, several experiments were conducted. The results demonstrate that this data augmentation strategy improves the performance of a previous COVID-19 screening proposal, achieving an accuracy of 98.61% when distinguishing among NON-COVID-19 (i.e. normal control samples and samples with pathologies others than COVID-19) and genuine COVID-19 samples. It is remarkable that this methodology can be extrapolated to other pulmonary pathologies and even other medical imaging domains to overcome the data scarcity.

One of the main challenges in times of sanitary emergency is to quickly develop computer aided diagnosis systems with a limited number of available samples due to the novelty, complexity of the case and the urgency of its implementation. This is the case during the current pandemic of COVID-19. This pathogen primarily infects the respiratory system of the afflicted, resulting in pneumonia and in a severe case of acute respiratory distress syndrome. This results in the formation of different pathological structures in the lungs that can be detected by the use of chest X-rays. Due to the overload of the health services, portable X-ray devices are recommended during the pandemic, preventing the spread of the disease. However, these devices entail different complications (such as capture quality) that, together with the subjectivity of the clinician, make the diagnostic process more difficult and suggest the necessity for computer-aided diagnosis methodologies despite the scarcity of samples available to do so. To solve this problem, we propose a methodology that allows to adapt the knowledge from a well-known domain with a high number of samples to a new domain with a significantly reduced number and greater complexity. We took advantage of a pre-trained segmentation model from brain magnetic resonance imaging of a unrelated pathology and performed two stages of knowledge transfer to obtain a robust system able to segment lung regions from portable X-ray devices despite the scarcity of samples and lesser quality. This way, our methodology obtained a satisfactory accuracy of 0.9761±0.0100 for patients with COVID-19, 0.9801±0.0104 for normal patients and 0.9769±0.0111 for patients with pulmonary diseases with similar characteristics as COVID-19 (such as pneumonia) but not genuine COVID-19.

This paper presents a novel approach for the segmentation of the optic disc and cup in eye fundus images using deep learning. The accurate segmentation of these anatomical structures in the eye is important towards the early detection of glaucoma and, therefore, potentially avoiding severe vision loss. In order to improve the segmentation of the optic disc and cup, we propose a novel self-supervised pretraining consisting in the multi-modal reconstruction of eye fundus images. This novel approach aims at facilitating the segmentation task and avoiding the necessity of excessively large annotated datasets.To validate the proposal, we perform several experiments on different public datasets. The results show that the proposed multi-modal self-supervised pre-training leads to a significant improvement in the performance of the segmentation task. Consequently, the presented approach shows remarkable potential towards further improving the interpretable and early diagnosis of a relevant disease as is glaucoma.

Data scarcity represents an important constraint for the training of deep neural networks in medical imaging. Medical image labeling, especially if pixel-level annotations are required, is an expensive task that needs expert intervention and usually results in a reduced number of annotated samples. In contrast, extensive amounts of unlabeled data are produced in the daily clinical practice, including paired multimodal images from patients that were subjected to multiple imaging tests. This work proposes a novel self-supervised multimodal reconstruction task that takes advantage of this unlabeled multimodal data for learning about the domain without human supervision. Paired multimodal data is a rich source of clinical information that can be naturally exploited by trying to estimate one image modality from others. This multimodal reconstruction requires the recognition of domain-specific patterns that can be used to complement the training of image analysis tasks in the same domain for which annotated data is scarce. In this work, a set of experiments is performed using a multimodal setting of retinography and fluorescein angiography pairs that offer complementary information about the eye fundus. The evaluations performed on different public datasets, which include pathological and healthy data samples, demonstrate that a network trained for self-supervised multimodal reconstruction of angiography from retinography achieves unsupervised recognition of important retinal structures. These results indicate that the proposed self-supervised task provides relevant cues for image analysis tasks in the same domain.

This paper presents a novel approach that allows training convolutional neural networks for retinal vessel segmentation without manually annotated labels. In order to learn how to segment the retinal vessels, convolutional neural networks are typically trained with a set of pixel-level labels annotated by a clinical expert. This annotation is a tedious and error-prone task that limits the number of available training samples. To alleviate this problem, we propose the use of unlabeled multimodal data for learning about the retinal vasculature. Instead of using manually annotated labels, the networks learn to segment the retinal vessels from a complementary image modality where the vasculature is already highlighted. In this complementary modality, a vessel map can be easily constructed with simple image processing techniques. Then, a convolutional neural network is trained to learn the cross-modal mapping from the original modality to the automatically derived vessel maps. Using this strategy, the supervisory signal for training is automatically obtained from the unlabeled multimodal data. Thus, the number of training samples can be increased without any human annotation effort. Several experiments were conducted to evaluate the performance of the networks that were trained with the automatically derived labels, obtaining competitive results for retinal vessel segmentation in relevant public datasets. Furthermore, the results are promising towards including the presented approach in semi-supervised methods.

This paper explores the application of deep learning-based methods for the multimodal reconstruction of fluorescein angiography from retinography. The objective of this multimodal reconstruction is not only to estimate an invasive modality from a non-invasive one, but also to apply the learned models for transfer learning or domain adaption. Deep neural networks have demonstrated to be successful at learning the mapping between complementary image domains, using both paired or unpaired data. The paired data allows taking advantage of the rich information that is available from the pixelwise correspondence of paired images. However, this requires the pre-registration of the multimodal image pairs. In the case of the retinal images, the multimodal registration is a challenging task that may fail in complex scenarios, such as severe pathological cases or low quality samples. In contrast, the use of generative adversarial networks allows learning the mapping between image domains using unpaired data. This avoids the preregistration of the images and allows including all the available data for training. In this work, we analyze both paired and unpaired deep learning-based approaches for the multimodal reconstruction of retinal images. The objective is to understand the implications of each alternative and the considerations for their future usage. For that purpose, we perform several experiments with the focus on producing a fair comparison between paired and unpaired approaches.

Diabetic Macular Edema (DME) is one of the most common causes of vision impairment and blindness in individuals with diabetes. Among the different imaging modalities, Optical Coherence Tomography (OCT) is a non-invasive ophthalmological imaging technique that is commonly used for the diagnosis, monitoring and treatment of DME. In this context, this paper proposes a new methodology for the automatic classification of DME using OCT images. Firstly, the method extracts a set of deep features from the target OCT images using a transfer learning-based approach. Then, the most relevant subset of deep features is selected using different feature selection strategies. Finally, a machine learning approach is applied to test the potential of the implemented method. The proposed methodology was validated using an OCT image dataset retrieved from 400 different patients, being 200 with DME and 200 normal cases. The proposed system achieved satisfactory results, reaching a best accuracy of 97.50%, using only 14.65% of the deep features in the classification of this ocular pathology, demonstrating also its competitive performance with respect to others approaches of the state-of-the-art.

Optical Coherence Tomography (OCT) is a medical imaging modality that is currently the focus of many advancements in the field of ophthalmology. It is widely used to diagnose relevant diseases like Diabetic Macular Edema (DME) or Age-related Macular Degeneration (AMD), both among the principal causes of blindness. These diseases have in common the presence of pathological cystoid fluid accumulations inside the retinal layers that tear its tissues, hindering the correct vision of the patient. In the last years, several works proposed a variety of methodologies to obtain a precise segmentation of these fluid regions. However, many cystoid patterns present several difficulties that harden significantly the process. In particular, some of these cystoid bodies present diffuse limits, others are deformed by shadows, appear mixed with other tissues and other complex situations. To overcome these drawbacks, a regional analysis has been proven to be reliable in these problematic regions. In this work, we propose the use of the DenseNet architecture to perform this regional analysis instead of the classical machine learning approaches, and use it to represent the pathological identifications with an intuitive color map. We trained, validated and tested the DenseNet neural network with a dataset composed of 3247 samples labeled by an expert. They were extracted from 156 images taken with two of the principal OCT devices of the domain. Then, this network was used to generate the color map representations of the cystoid areas in the OCT images. Our proposal achieved robust results in these regions, with a satisfactory 97.48% ± 0.7611 mean test accuracy as well as a mean AUC of 0.9961 ± 0.0029.

Optical Coherence Tomography Angiography (OCTA) constitutes a novel ophthalmological image modality that is characterized for being a non-invasive capture technique that allows a profound analysis of the vascular characteristics of the eye fundus. Given the restricted field of view of the eye fundus that offers each scan, the specialists frequently capture several complementary images that may be simultaneously analyzed to offer a complete and accurate diagnosis of the patient. In this work, we propose a fully automatic method to register complementary OCTA images and provide compositions for the same patient, generating a wide field of representation that allows a simpler and more direct analysis than the traditional tedious manual procedures. To achieve this, we based our proposal in a robust combination of representative features that are filtered by an accurate identification of the main retinal vasculature. This way, given the characteristic high irregularity in the fundus of the OCTA images, we avoid many variable areas that may interfere in the registration process, restricting the analysis to the most representative and stable structure of this image modality, the main retinal vasculature. In particular, we use Speeded-Up Robust Features (SURF) algorithm to extract representative features in the main vascular region that is extracted using a method that combines the analysis of the Hessian matrix followed by an hysteresis threshold process. Then, using a K-NN model, we perform the registration of the resulting features from the different OCTA images to be analyzed. Finally, the Random sample consensus (RANSAC) method is exploited to produce the final target mosaic. The proposed method presented satisfactory results in the validation experiments, with accurate values for the MSE index of 1.2566 and 1.6725 pixels for the registration of paired images an mosaics, respectively.

Optical Coherence Tomography (OCT) is a well-established medical imaging technique that allows a complete analysis and evaluation of the main retinal structures and their histopathology properties. Diabetic Macular Edema (DME) implies the accumulation of intraretinal fluid within the macular region. Diffuse Retinal Thickening (DRT) edemas are considered a relevant case of DME disease, where the pathological regions are characterized by a “sponge-like” appearance and a reduced intraretinal reflectivity, being visible in OCT images. Additionally, the presence of other structures may alter the OCT image characteristics, confusing the pathological identification process. This is the case of the retinal vessels over all the eye fundus, whose presence produce shadow projections over the retinal layers that may hide the “sponge-like” appearance of the DRT edemas. Thus, in this paper, we present a proposal for the automatic extraction of DRT edemas, also using as reference the information provided by the automatic identifications of the retinal vessels in the OCT images. To do that, firstly, the system delimits three retinal regions of interest. These retinal regions facilitate the posterior identification of the vessel structures and the segmentation of the DRT regions. For the identification of the vessels structures, the method combined the localization of the upper bright vascular profiles with the presence of their corresponding lower dark vascular shadows. Finally, a learning strategy is implemented for the segmentation of the DRT edemas. Satisfactory results were obtained, reaching values of 0.8346 and 0.9051 of Jaccard index and Dice coefficient, respectively, for the extraction of the existing DRT edemas.

Optical Coherence Tomography Angiography (OCT-A) represents a new modality of ophthalmological imaging that stands out for being a non-invasive capture technique that facilitates the analysis of the vascular characteristics of the eye fundus. In this paper, we propose a complete automatic methodology that identifies the vascular and avascular zones in OCT-A images, quantifying each one of them for their posterior use in clinical analyses and diagnostic processes. To achieve this, we firstly intensify the vascular characteristics to facilitate the posterior extraction. Then, a set of image processing techniques are combined to differentiate both vascular and avascular regions and, finally, measure their representative parameters. The proposed methodology was tested on a set of images that were marked by an expert ophthalmologist, being used as reference in the validation of the method. The proposed approach presented satisfactory results in the validation experiments with the vascular and avascular measurements, demonstrating their utility for the diagnosis and monitoring of different vascular diseases that are frequently analysed through the retinal microcirculation.

The analysis of different image modalities is frequently performed in ophthalmology as it provides complementary information for the diagnosis and follow-up of relevant diseases, like hypertension or diabetes. This work presents a hybrid method for the multimodal registration of color fundus retinography and fluorescein angiography. The proposed method combines a feature-based approach, using domain-specific landmarks, with an intensity-based approach that employs a domain-adapted similarity metric. The methodology is tested on a dataset of 59 image pairs containing both healthy and pathological cases. The results show a satisfactory performance of the proposed combined approach in this multimodal scenario, improving the registration accuracy achieved by the feature-based and the intensity-based approaches.

22nd International Conference on Knowledge-Based and Intelligent Information and Engineering Systems, KES 2018, Belgrade, 3 September 2018 - 5 September 2018

The retinal vascular tortuosity is a commonly used parameter for the early diagnosis of several diseases that affects the circulatory system. The manual analysis of fundus images for the tortuosity characterization is a time-consuming and subjective task that presents a high inter-rater variability. Thus, automatic image processing methods allow the efficient computation of objective and stable parameters for the issue. The validation of these methods is crucial to ensure an objective and reliable environment for the retinal experts. This paper describes a multi-expert analysis that measures the clinical performance as well as a validation procedure of the computational tortuosity module of the Sirius framework, a computer-aided diagnosis platform for analyzing retinal images.

Diabetic retinopathy is one of the leading causes of vision impairment that is commonly associated to the Macular Edema (ME) disease. The Diffuse Retinal Thickening (DRT) is a ME type derived from the local intraretinal fluid accumulation in the lower retinal layers, producing significant morphological alterations in the eye fundus. The presence and properties of these intraretinal fluids are used by the ophthalmologists as significant indicators of the clinical stage of the ME disease. Given that, the precise identification and segmentation of the DRT edema type allow the early diagnosis of the ME disease which, therefore, permits a better adjustment of the treatments, reducing their costs as well as improving the life quality of the patients. This paper proposes a novel methodology for the automatic identification and segmentation of the DRT edemas using Optical Coherence Tomography (OCT) images as source of information. Firstly, the method identifies four of the principal retinal layers that are used as reference to delimit the outer retina, region where the DRT edemas are typically originated. Inside this region, a large and heterogeneous set of features was defined to recognize the characteristic “sponge-like” patterns of the DRT edema, using intensity, texture and clinically-defined features. For this analysis, four representative classifiers were employed with the best subsets of previously selected features. This methodology was tested using 70 OCT images from where 560 samples were extracted with the presence and absence of DRT edemas. The best results were achieved by the 7-kNN classifier, reaching in the detection stage an accuracy of 0.9366, whereas in the segmentation stage obtained values of 0.6625 and 0.7899 for the Jaccard and Dice coefficients, respectively.

Optical Coherence Tomography (OCT) is a well-established medical imaging technique that allows the analysis of the eye fundus characteristics in real time. These images enable the experts to make a clinical evaluation of the retinal vasculature, whose morphology provides relevant information for diseases like diabetes, hypertension or arteriosclerosis. In this paper, we present a novel proposal for the automatic vasculature identification in retinal OCT images. To achieve this, we analyse the intensity profiles between representative retinal layers, previously segmented. Then, two statistical models are generated using representative samples of vessel and non-vessel profiles. The analysis of both statistical models let us optimize the discrimination of both cathegories that is used, finally, to identify the vessel locations. The proposed method was adjusted and validated using 256 OCT images, including 1274 vascular structures that were labelled by an expert clinician. Satisfactory results were provided as a precision of 94.55% and a recall of 90.25% were obtained, respectively. The method facilitates the doctors’ work allowing better analysis and treatments of vascular diseases.

Optical Coherence Tomography (OCT) is, nowadays, a clinical standard imaging technique in opthalmology as it provides more information than other classical modalities as can be, for instance, retinographies. OCT scans show a 3D representation of the real layout of the eye fundus in a non-invasive way, letting clinicians inspect deeply the retinal layers in a cross-sectional visualization. For that reason, OCT scans are commonly used in the study of the retinal morphology and the identification of pathological structures. Among them, an appropriate identification and analysis of any present intraretinal cystoid region is crucial to perform an adequate diagnosis of the exudative macular disease, one of the main causes of blindness in developed countries. In this work, we analyzed and characterized the intraretinal cystoid regions in OCT images by the definition of a complete and heterogeneous set of 326 intensity and texture-based features. Relief-F and L0 feature selectors were used in order to identify the optimal feature subsets that provide the best discriminative power. Representative classifiers, as the Linear Bayes Normal Classifier (LDC), Quadratic Bayes Normal Classifier (QDC) and K-Nearest Neighbor Classifier (KNN) were finally used to evaluate the potential of identification of the feature subsets. The method was validated using 51 OCT images. From them, 363 and 360 samples of cystoid and non-cystoid regions were selected, respectively. The best results were offered by the LDC classifier that, using a feature subset identified by the L0 selector, provided an accuracy of 0.9060.

Audiology is the science of hearing and auditory processes study. The evaluation of hearing capacity is commonly performed by an audiologist using an audiometer, where the patient is asked to show some kind of sign when he or she recognizes the stimulus. This evaluation becomes much more complicated when the patient suffers some type of cognitive decline that hinders the emission of visible signs of recognition. With this group of patients, a typical question-answer interaction is not applicable, so the audiologist must focus his attention on the patient's spontaneous gestural reactions. This manual evaluation entails a number of problems: it is highly subjective, difficult to determine in real time (since the expert must pay attention simultaneously to the audiological process and the patient's reactions), etc. Considering this, in this paper, we present an automatic methodology for processing video sequences recorded during the performance of the hearing test in order to assist the audiologist in the detection of these spontaneous reactions. This screening method analyzes the movements that occur within the eye area, which has been pointed out by the audiologists as the most representative for these patients. By the analysis of these movements, the system helps the audiologist to determine when a positive gestural reaction has taken place increasing the objectivity and reproducibility.

This paper presents a new wavefront propagation method for dealing with the classic Eikonal equation. While classic Dijkstra-like graph-based techniques achieve the solution in O(M log M), they do not approximate the unique physically relevant solution very well. Fast Marching Methods (FMM) were created to efficiently solve the continuous problem. The proposed approximation tries to maintain the complexity, in order to make the algorithm useful in a wide range of contexts. The key idea behind our method is the creation of 'mini wave-fronts', which are combined to propagate the solution. Experimental results show the improvement in the accuracy with respect to the state of the art, while the average computational speed is maintained in O(M log M), similar to the FMM techniques.

This paper presents a novel methodology for modelling pedestrian trajectories over a scene, based in the hypothesis that, when people try to reach a destination, they use the path that takes less time, taking into account environmental information like the type of terrain or what other people did before. Thus, a minimal path approach can be used to model human trajectory behaviour. We develop a modified Fast Marching Method that allows us to include both velocity and orientation in the Front Propagation Approach, without increasing its computational complexity. Combining all the information, we create a time surface that shows the time a target need to reach any given position in the scene. We also create different metrics in order to compare the time surface against the real behaviour. Experimental results over a public dataset prove the initial hypothesis' correctness.

Hearing loss is a common disease which affects a large percentage of the population. Hearing loss may have a negative impact on health, social participation, and daily activities, so its diagnosis and monitoring is indeed important. The audiometric tests related to this diagnosis are constrained when the patient suffers from some form of cognitive impairment. In these cases, audilogist must try to detect particular facial reactions that may indicate auditory perception. With the aim of supporting the audiologist in this evaluation, a screening method that analyzes video sequences and seeks for facial reactions within the eye area was proposed. In this research, a comprehensive survey of one of the most relevent steps of this methodology is presented. This survey considers different alternatives for the detection of the interest points and the classsification techniques. The provided results allow to determine the most suitable configuration for this domain.

OCT technique for retinal imaging is establishing itself as a relevant modality among ophthalmologists due to its capacity to show more information than classical modalities. Nowadays, most image processing-based applications are emerging to extract that information automatically. As previous step of any automatic method to extract features from these images, the segmentation of the retinal layers has to be done. Graph-based methods provide good results for this problem, although their efficiency is an important limitation. In this work, a multiscale or pyramidal-based approach is studied in order to solve this problem. Different configurations are proposed to determine the optimal method. It is remarkable that this approach means an improvement not only in computation time, but also in segmentation results.

In this work Optical Coherence Tomography (OCT) retinal images are automatically processed to detect the presence of cysts. The methodology is composed by three phases: region of interest where cysts will be searched is delimited; a watershed algorithm is applied to find all the possible regions in the image which might conform cystic structures; finally, texture analysis is performed in each region from previous phase to final classification. Results show that accuracy achieved with this method is over 80%.

This paper presents an improved version of our specific methodology to detect the vessel tree in retinal angiographies. The automatic analysis of retinal vessel tree facilitates the computation of the arteriovenous index, which is essential for the diagnosis several eye diseases. The developed system is inspired in the classical snake but incorporating domain specific knowledge, such as blood vessels topological properties. It profits from the automatic localization of the optic disc, the vessel creases extraction and, as a recent innovation, the morphological vessel segmentation, all developed in our research group. After researching and testing our system, the parameter configuration has been enhanced. Significantly better results in the detection of arteriovenous structures are obtained, keeping a high efficiency, as shown by the systems performance evaluation on the publicly available DRIVE database.

Biometric features have been studied in order to be applied to authentication and identification systems due to its reliability. Among others, the retinal vessel tree have been proposed as a vessel pattern for personal authentication applications, since it is almost impossible to forge. In this kind of systems, the retinal vessel tree is computed from the retinal image, and then a registration process is made. Although reliable and remarkable results have been obtained in this vessel pattern-based system, the required computation effort is quite high, particularly to compute and extract the vessel tree. In this paper, a pixel parallel approach is proposed to tackle with the retinal vessel tree extraction to be used in a personal retinal authentication system, regarding the computation speed.