Does imbalance in chest X-ray datasets produce biased deep learning approaches for COVID-19 screening?

BACKGROUND: The health crisis resulting from the global COVID-19 pandemic highlighted more than ever the need for rapid, reliable and safe methods of diagnosis and monitoring of respiratory diseases. To study pulmonary involvement in detail, one of the most common resources is the use of different lung imaging modalities (like chest radiography) to explore the possible affected areas. METHODS: The study of patient characteristics like sex and age in pathologies of this type is crucial for gaining knowledge of the disease and for avoiding biases due to the clear scarcity of data when developing representative systems. In this work, we performed an analysis of these factors in chest X-ray images to identify biases. Specifically, 11 imbalance scenarios were defined with female and male COVID-19 patients present in different proportions for the sex analysis, and 6 scenarios where only one specific age range was used for training for the age factor. In each study, 3 different approaches for automatic COVID-19 screening were used: Normal vs COVID-19, Pneumonia vs COVID-19 and Non-COVID-19 vs COVID-19. The study was validated using two public chest X-ray datasets, allowing a reliable analysis to support the clinical decision-making process. RESULTS: The results for the sex-related analysis indicate this factor slightly affects the system in the Normal VS COVID-19 and Pneumonia VS COVID-19 approaches, although the identified differences are not relevant enough to worsen considerably the system. Regarding the age-related analysis, this factor was observed to be influencing the system in a more consistent way than the sex factor, as it was present in all considered scenarios. However, this worsening does not represent a major factor, as it is not of great magnitude. CONCLUSIONS: Multiple studies have been conducted in other fields in order to determine if certain patient characteristics such as sex or age influenced these deep learning systems. However, to the best of our knowledge, this study has not been done for COVID-19 despite the urgency and lack of COVID-19 chest x-ray images. The presented results evidenced that the proposed methodology and tested approaches allow a robust and reliable analysis to support the clinical decision-making process in this pandemic scenario.

Fully-Automatic 3D Intuitive Visualization of Age-Related Macular Degeneration Fluid Accumulations in OCT Cubes.

Age-related macular degeneration is the leading cause of vision loss in developed countries, and wet-type AMD requires urgent treatment and rapid diagnosis because it causes rapid irreversible vision loss. Currently, AMD diagnosis is mainly carried out using images obtained by optical coherence tomography. This diagnostic process is performed by human clinicians, so human error may occur in some cases. Therefore, fully automatic methodologies are highly desirable adding a layer of robustness to the diagnosis. In this work, a novel computer-aided diagnosis and visualization methodology is proposed for the rapid identification and visualization of wet AMD. We adapted a convolutional neural network for segmentation of a similar domain of medical images to the problem of wet AMD segmentation, taking advantage of transfer learning, which allows us to work with and exploit a reduced number of samples. We generate a 3D intuitive visualization where the existence, position and severity of the fluid were represented in a clear and intuitive way to facilitate the analysis of the clinicians. The 3D visualization is robust and accurate, obtaining satisfactory 0.949 and 0.960 Dice coefficients in the different evaluated OCT cube configurations, allowing to quickly assess the presence and extension of the fluid associated to wet AMD.

High prevalence of SARS-CoV-2 detection and prolonged viral shedding in stools: A systematic review and cohort study.

OBJECTIVES: To: 1. Describe the frequency of viral RNA detection in stools in a cohort of patients infected with SARS-CoV-2, and 2. Perform a systematic review to assess the clearance time in stools of SARS-CoV-2. METHODS: We conducted a prospective cohort study in two centers between March and May 2020. We included SARS-CoV-2 infected patients of any age and severity. We collected seriated nasopharyngeal swabs and stool samples to detect SARS-CoV-2. After, we performed a systematic review of the prevalence and clearance of SARS-CoV-2 in stools (PROSPERO-ID: CRD42020192490). We estimated prevalence using a random-effects model. We assessed clearance time by using Kaplan-Meier curves. RESULTS: We included 32 patients; mean age was 43.7±17.7 years, 43.8% were female, and 40.6% reported gastrointestinal symptoms. Twenty-five percent (8/32) of patients had detectable viral RNA in stools. The median clearance time in stools of the cohort was 11[10-15] days. Systematic review included 30 studies (1392 patients) with stool samples. Six studies were performed in children and 55% were male. The pooled prevalence of viral detection in stools was 34.6% (twenty-four studies, 1393 patients; 95%CI:25.4-45.1); heterogeneity was high (I2:91.2%, Q:208.6; p≤0.001). A meta-regression demonstrates an association between female-gender and lower presence in stools (p=0.004). The median clearance time in stools was 22 days (nineteen studies, 140 patients; 95%CI:19-25). After 34 days, 19.9% (95%CI:11.3-29.7) of patients have a persistent detection in stools. CONCLUSIONS: Detection of SARS-CoV-2 in stools is a frequent finding. The clearance of SARS-CoV-2 in stools is prolonged and it takes longer than nasopharyngeal secretions.

Fully automatic deep convolutional approaches for the analysis of COVID-19 using chest X-ray images.

Covid-19 is a new infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Given the seriousness of the situation, the World Health Organization declared a global pandemic as the Covid-19 rapidly around the world. Among its applications, chest X-ray images are frequently used for an early diagnostic/screening of Covid-19 disease, given the frequent pulmonary impact in the patients, critical issue to prevent further complications caused by this highly infectious disease. In this work, we propose 4 fully automatic approaches for the classification of chest X-ray images under the analysis of 3 different categories: Covid-19, pneumonia and healthy cases. Given the similarity between the pathological impact in the lungs between Covid-19 and pneumonia, mainly during the initial stages of both lung diseases, we performed an exhaustive study of differentiation considering different pathological scenarios. To address these classification tasks, we evaluated 6 representative state-of-the-art deep network architectures on 3 different public datasets: (I) Chest X-ray dataset of the Radiological Society of North America (RSNA); (II) Covid-19 Image Data Collection; (III) SIRM dataset of the Italian Society of Medical Radiology. To validate the designed approaches, several representative experiments were performed using 6,070 chest X-ray radiographs. In general, satisfactory results were obtained from the designed approaches, reaching a global accuracy values of 0.9706 ± 0.0044, 0.9839 ± 0.0102, 0.9744 ± 0.0104 and 0.9744 ± 0.0104, respectively, thus helping the work of clinicians in the diagnosis and consequently in the early treatment of this relevant pandemic pathology.

Multi-stage transfer learning for lung segmentation using portable X-ray devices for patients with COVID-19.

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.

The Role of Anxiety in Patients With Persistent Atrial Fibrillation Undergoing Elective Cardioversion: An Observational Study.

OBJECTIVE: Anxiety is often present among patients with atrial fibrillation (AF). This condition has been associated with greater symptom severity and worse quality of life in these patients. However, the influence of anxiety on the risk of AF recurrence is not well known. We aimed to define the level of anxiety in patients with persistent AF undergoing elective cardioversion (EC) and determine whether there is an association between anxiety and the risk of early AF recurrence after EC. METHODS: Anxiety was measured before EC using the State-Trait Anxiety Inventory. Early AF recurrence was assessed with a control electrocardiogram at 30-day follow-up. RESULTS: We included 107 patients undergoing effective EC. Early AF recurrence was diagnosed in 40 patients (37.4%). Compared with those who remained in sinus rhythm, individuals with early AF recurrence had significantly higher levels of trait anxiety (23.1 [10.4] versus 17.9 [9.5]; p = .013) and larger left atrial volume index (45.8 [12.3] versus 37.9 [13.3] ml/m; p = .004). Both variables remained independently associated with early AF recurrence after multivariate analysis. A predictive model including trait anxiety score >20 and left atrial volume index >41 ml/m showed acceptable accuracy for the diagnosis of early AF recurrence (area under the curve = 0.733; 95% confidence interval = 0.634-0.832; p < .001). CONCLUSIONS: Our study shows that trait anxiety is an independent risk factor for early AF recurrence after EC. Further studies are warranted to assess the beneficial role of anxiety-reducing strategies on the outcomes of patients with AF.

Ocular surface analysis and automatic non-invasive assessment of tear film breakup location, extension and progression in patients with glaucoma.

BACKGROUND: Tear film stability is the key event in ocular surface diseases. The purpose of this study is to evaluate spatial and temporal progression of the tear film breakup using an automatic non-invasive device. METHODS: Non-invasive tear breakup time (NITBUT) parameters, such as First NITBUT (F-NITBUT) and Average NITBUT (A-NITBUT), were evaluated in 132 glaucoma and 87 control eyes with the Keratograph 5 M device. Further analysis of this data was used to determine size, location and progression of tear film breakup with automatically identified breakup areas (BUA). The progression from First BUA (F-BUA) to total BUA (T-BUA) was expressed as Dry Area Growth Rate (DAGR). Differences between both groups were analysed using Student t-test for parametric data and Mann-Whitney U test for non-parametric data. Pearson's correlation coefficient was used to assess the relationship between parametric variables and Spearman in the case of non-parametric variables. RESULTS: F-NITBUT was 11.43 ± 7.83 s in the control group and 8.17 ± 5.73 in the glaucoma group (P = 0.010). A-NITBUT was 14.04 ± 7.21 and 11.82 ± 6.09 s in control and glaucoma groups, respectively (P = 0.028). F-BUA was higher in the glaucoma group than in the control group (2.73 and 2.28; P = 0.022) and was more frequently located at the centre of the cornea in the glaucoma group (P = 0.039). T-BUA was also higher in the glaucoma group than in the control group (13.24 and 9.76%; P = 0.012) and the DAGR was steeper in the glaucoma group than in the control group (34.38° and 27.15°; P = 0.009). CONCLUSIONS: Shorter NITBUT values and bigger, more central tear film breakup locations were observed in the glaucoma group than in the control group. The DAGR indicates that tear film rupture is bigger and increases faster in glaucomatous eyes than in normal eyes.

Intraretinal Fluid Pattern Characterization in Optical Coherence Tomography Images.

Optical Coherence Tomography (OCT) has become a relevant image modality in the ophthalmological clinical practice, as it offers a detailed representation of the eye fundus. This medical imaging modality is currently one of the main means of identification and characterization of intraretinal cystoid regions, a crucial task in the diagnosis of exudative macular disease or macular edema, among the main causes of blindness in developed countries. This work presents an exhaustive analysis of intensity and texture-based descriptors for its identification and classification, using a complete set of 510 texture features, three state-of-the-art feature selection strategies, and seven representative classifier strategies. The methodology validation and the analysis were performed using an image dataset of 83 OCT scans. From these images, 1609 samples were extracted from both cystoid and non-cystoid regions. The different tested configurations provided satisfactory results, reaching a mean cross-validation test accuracy of 92.69%. The most promising feature categories identified for the issue were the Gabor filters, the Histogram of Oriented Gradients (HOG), the Gray-Level Run-Length matrix (GLRL), and the Laws' texture filters (LAWS), being consistently and considerably selected along all feature selector algorithms in the top positions of different relevance rankings.

Joint Diabetic Macular Edema Segmentation and Characterization in OCT Images.

The automatic identification and segmentation of edemas associated with diabetic macular edema (DME) constitutes a crucial ophthalmological issue as they provide useful information for the evaluation of the disease severity. According to clinical knowledge, the DME disorder can be categorized into three main pathological types: serous retinal detachment (SRD), cystoid macular edema (CME), and diffuse retinal thickening (DRT). The implementation of computational systems for their automatic extraction and characterization may help the clinicians in their daily clinical practice, adjusting the diagnosis and therapies and consequently the life quality of the patients. In this context, this paper proposes a fully automatic system for the identification, segmentation and characterization of the three ME types using optical coherence tomography (OCT) images. In the case of SRD and CME edemas, different approaches were implemented adapting graph cuts and active contours for their identification and precise delimitation. In the case of the DRT edemas, given their fuzzy regional appearance that requires a complex extraction process, an exhaustive analysis using a learning strategy was designed, exploiting intensity, texture, and clinical-based information. The different steps of this methodology were validated with a heterogeneous set of 262 OCT images, using the manual labeling provided by an expert clinician. In general terms, the system provided satisfactory results, reaching Dice coefficient scores of 0.8768, 0.7475, and 0.8913 for the segmentation of SRD, CME, and DRT edemas, respectively.

Engineering of industrially important microorganisms for assimilation of cellulosic biomass: towards consolidated bioprocessing.

Conversion of cellulosic biomass (non-edible plant material) to products such as chemical feedstocks and liquid fuels is a major goal of industrial biotechnology and an essential component of plans to move from an economy based on fossil carbon to one based on renewable materials. Many microorganisms can effectively degrade cellulosic biomass, but attempts to engineer this ability into industrially useful strains have met with limited success, suggesting an incomplete understanding of the process. The recent discovery and continuing study of enzymes involved in oxidative depolymerisation, as well as more detailed study of natural cellulose degradation processes, may offer a way forward.

Automatic Identification and Representation of the Cornea-Contact Lens Relationship Using AS-OCT Images.

The clinical study of the cornea-contact lens relationship is widely used in the process of adaptation of the scleral contact lens (SCL) to the ocular morphology of patients. In that sense, the measurement of the adjustment between the SCL and the cornea can be used to study the comfort or potential damage that the lens may produce in the eye. The current analysis procedure implies the manual inspection of optical coherence tomography of the anterior segment images (AS-OCT) by the clinical experts. This process presents several limitations such as the inability to obtain complex metrics, the inaccuracies of the manual measurements or the requirement of a time-consuming process by the expert in a tedious process, among others. This work proposes a fully-automatic methodology for the extraction of the areas of interest in the study of the cornea-contact lens relationship and the measurement of representative metrics that allow the clinicians to measure quantitatively the adjustment between the lens and the eye. In particular, three distance metrics are herein proposed: Vertical, normal to the tangent of the region of interest and by the nearest point. Moreover, the images are classified to characterize the analysis as belonging to the central cornea, peripheral cornea, limbus or sclera (regions where the inner layer of the lens has already joined the cornea). Finally, the methodology graphically presents the results of the identified segmentations using an intuitive visualization that facilitates the analysis and diagnosis of the patients by the clinical experts.

Automatic Identification and Intuitive Map Representation of the Epiretinal Membrane Presence in 3D OCT Volumes.

Optical Coherence Tomography (OCT) is a medical image modality providing high-resolution cross-sectional visualizations of the retinal tissues without any invasive procedure, commonly used in the analysis of retinal diseases such as diabetic retinopathy or retinal detachment. Early identification of the epiretinal membrane (ERM) facilitates ERM surgical removal operations. Moreover, presence of the ERM is linked to other retinal pathologies, such as macular edemas, being among the main causes of vision loss. In this work, we propose an automatic method for the characterization and visualization of the ERM's presence using 3D OCT volumes. A set of 452 features is refined using the Spatial Uniform ReliefF (SURF) selection strategy to identify the most relevant ones. Afterwards, a set of representative classifiers is trained, selecting the most proficient model, generating a 2D reconstruction of the ERM's presence. Finally, a post-processing stage using a set of morphological operators is performed to improve the quality of the generated maps. To verify the proposed methodology, we used 20 3D OCT volumes, both with and without the ERM's presence, totalling 2428 OCT images manually labeled by a specialist. The most optimal classifier in the training stage achieved a mean accuracy of 91 . 9 % . Regarding the post-processing stage, mean specificity values of 91 . 9 % and 99 . 0 % were obtained from volumes with and without the ERM's presence, respectively.

Automatic Visual Acuity Estimation by Means of Computational Vascularity Biomarkers Using Oct Angiographies.

Optical Coherence Tomography Angiography (OCTA) constitutes a new non-invasive ophthalmic image modality that allows the precise visualization of the micro-retinal vascularity that is commonly used to analyze the foveal region. Given that there are many systemic and eye diseases that affect the eye fundus and its vascularity, the analysis of that region is crucial to diagnose and estimate the vision loss. The Visual Acuity (VA) is typically measured manually, implying an exhaustive and time-consuming procedure. In this work, we propose a method that exploits the information of the OCTA images to automatically estimate the VA with an accurate error of 0.1713.

Artery/Vein Vessel Tree Identification in Near-Infrared Reflectance Retinographies.

An accurate identification of the retinal arteries and veins is a relevant issue in the development of automatic computer-aided diagnosis systems that facilitate the analysis of different relevant diseases that affect the vascular system as diabetes or hypertension, among others. The proposed method offers a complete analysis of the retinal vascular tree structure by its identification and posterior classification into arteries and veins using optical coherence tomography (OCT) scans. These scans include the near-infrared reflectance retinography images, the ones we used in this work, in combination with the corresponding histological sections. The method, firstly, segments the vessel tree and identifies its characteristic points. Then, Global Intensity-Based Features (GIBS) are used to measure the differences in the intensity profiles between arteries and veins. A k-means clustering classifier employs these features to evaluate the potential of artery/vein identification of the proposed method. Finally, a post-processing stage is applied to correct misclassifications using context information and maximize the performance of the classification process. The methodology was validated using an OCT image dataset retrieved from 46 different patients, where 2,392 vessel segments and 97,294 vessel points were manually labeled by an expert clinician. The method achieved satisfactory results, reaching a best accuracy of 93.35% in the identification of arteries and veins, being the first proposal that faces this issue in this image modality.

Retinal vascular tortuosity assessment: inter-intra expert analysis and correlation with computational measurements.

BACKGROUND: The retinal vascular tortuosity can be a potential indicator of relevant vascular and non-vascular diseases. However, the lack of a precise and standard guide for the tortuosity evaluation hinders its use for diagnostic and treatment purposes. This work aims to advance in the standardization of the retinal vascular tortuosity as a clinical biomarker with diagnostic potential, allowing, thereby, the validation of objective computational measurements on the basis of the entire spectrum of the expert knowledge. METHODS: This paper describes a multi-expert validation process of the computational vascular tortuosity measurements of reference. A group of five experts, covering the different clinical profiles of an ophthalmological service, and a four-grade scale from non-tortuous to severe tortuosity as well as non-tortuous / tortuous and asymptomatic / symptomatic binary classifications are considered for the analysis of the the multi-expert validation procedure. The specialists rating process comprises two rounds involving all the experts and a joint round to establish consensual rates. The expert agreement is analyzed throughout the rating procedure and, then, the consensual rates are set as the reference to validate the prognostic performance of four computational tortuosity metrics of reference. RESULTS: The Kappa indexes for the intra-rater agreement analysis were obtained between 0.35 and 0.83 whereas for the inter-rater agreement in the asymptomatic / symptomatic classification were between 0.22 and 0.76. The Area Under the Curve (AUC) for each expert against the consensual rates were placed between 0.61 and 0.83 whereas the prognostic performance of the best objective tortuosity metric was 0.80. CONCLUSIONS: There is a high inter and intra-rater variability, especially for the case of the four grade scale. The prognostic performance of the tortuosity measurements is close to the experts' performance, especially for Grisan measurement. However, there is a gap between the automatic effectiveness and the expert perception given the lack of clinical criteria in the computational measurements.

Molecular Dissection of the Forces Responsible for Viral Capsid Assembly and Stabilization by Decoration Proteins.

Complex double-stranded DNA viruses utilize a terminase enzyme to package their genomes into a preassembled procapsid shell. DNA packaging triggers a major conformational change in the proteins assembled into the shell and most often subsequent addition of a decoration protein that is required to stabilize the structure. In bacteriophage λ, DNA packaging drives a procapsid expansion transition to afford a larger but fragile shell. The gpD decoration protein adds to the expanded shell as trimeric spikes at each of the 140 three-fold axes. The spikes provide mechanical strength to the shell such that it can withstand the tremendous internal forces generated by the packaged DNA in addition to environmental insults. Hydrophobic, electrostatic, and aromatic-proline noncovalent interactions have been proposed to mediate gpD trimer spike assembly at the expanded shell surface. Here, we directly examine each of these interactions and demonstrate that hydrophobic interactions play the dominant role. In the course of this study, we unexpectedly found that Trp308 in the λ major capsid protein (gpE) plays a critical role in shell assembly. The gpE-W308A mutation affords a soluble, natively folded protein that does not further assemble into a procapsid shell, despite the fact that it retains binding interactions with the scaffolding protein, the shell assembly chaparone protein. The data support a model in which the λ procapsid shell assembles via cooperative interaction of monomeric capsid proteins, as observed in the herpesviruses and phages such as P22. The significance of the results with respect to capsid assembly, maturation, and stability is discussed.

Multi-Adaptive Optimization for multi-task learning with deep neural networks.

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.

Explainable artificial intelligence for the automated assessment of the retinal vascular tortuosity.

Retinal vascular tortuosity is an excessive bending and twisting of the blood vessels in the retina that is associated with numerous health conditions. We propose a novel methodology for the automated assessment of the retinal vascular tortuosity from color fundus images. Our methodology takes into consideration several anatomical factors to weigh the importance of each individual blood vessel. First, we use deep neural networks to produce a robust extraction of the different anatomical structures. Then, the weighting coefficients that are required for the integration of the different anatomical factors are adjusted using evolutionary computation. Finally, the proposed methodology also provides visual representations that explain the contribution of each individual blood vessel to the predicted tortuosity, hence allowing us to understand the decisions of the model. We validate our proposal in a dataset of color fundus images providing a consensus ground truth as well as the annotations of five clinical experts. Our proposal outperforms previous automated methods and offers a performance that is comparable to that of the clinical experts. Therefore, our methodology demonstrates to be a viable alternative for the assessment of the retinal vascular tortuosity. This could facilitate the use of this biomarker in clinical practice and medical research.

Adapted generative latent diffusion models for accurate pathological analysis in chest X-ray images.

Respiratory diseases have a significant global impact, and assessing these conditions is crucial for improving patient outcomes. Chest X-ray is widely used for diagnosis, but expert evaluation can be challenging. Automatic computer-aided diagnosis methods can provide support for clinicians in these tasks. Deep learning has emerged as a set of algorithms with exceptional potential in such tasks. However, these algorithms require a vast amount of data, often scarce in medical imaging domains. In this work, a new data augmentation methodology based on adapted generative latent diffusion models is proposed to improve the performance of an automatic pathological screening in two high-impact scenarios: tuberculosis and lung nodules. The methodology is evaluated using three publicly available datasets, representative of real-world settings. An ablation study obtained the highest-performing image generation model configuration regarding the number of training steps. The results demonstrate that the novel set of generated images can improve the performance of the screening of these two highly relevant pathologies, obtaining an accuracy of 97.09%, 92.14% in each dataset of tuberculosis screening, respectively, and 82.19% in lung nodules. The proposal notably improves on previous image generation methods for data augmentation, highlighting the importance of the contribution in these critical public health challenges.

Intra- and Inter-expert Validation of an Automatic Segmentation Method for Fluid Regions Associated with Central Serous Chorioretinopathy in OCT Images.

Central Serous Chorioretinopathy (CSC) is a retinal disorder caused by the accumulation of fluid, resulting in vision distortion. The diagnosis of this disease is typically performed through Optical Coherence Tomography (OCT) imaging, which displays any fluid buildup between the retinal layers. Currently, these fluid regions are manually detected by visual inspection a time-consuming and subjective process that can be prone to errors. A series of six deep learning-based automatic segmentation architectural configurations of different levels of complexity were trained and compared in order to determine the best model intended for the automatic segmentation of CSC-related lesions in OCT images. The best performing models were then evaluated in an external validation study. Furthermore, an intra- and inter-expert analysis was conducted in order to compare the manual segmentation performed by expert ophthalmologists with the automatic segmentation provided by the models. Test results of the best performing configuration achieved a mean Dice of 0.868 ± 0.056 in the internal dataset. In the external validation set, these models achieved a level of agreement with human experts of up to 0.960 in terms of Kappa coefficient, contrasting with a value of 0.951 for agreement between human experts. Overall, the models reached a better agreement with either of the human experts than these experts with each other, suggesting that automatic segmentation models for the detection of CSC-related lesions in OCT imaging can be useful tools for assessing this disease, reducing the workload of manual inspection and leading to a more robust and objective diagnosis method.