Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 12
Encontrada(s) 1 página(s)

RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Retina-Focused Preprocessing for Enhanced Deep Learning Classification of Multiple Sclerosis from OCT Scans

  • Álvarez-Rodríguez, Lorena
  • Cordón Ciordia, Beatriz
  • García-Martín, Elena
  • Novo Buján, Jorge
  • Moura, Joaquim de
  • Ortega Hortas, Marcos
Presentado en: EUROCAST 2026, 20th International Conference on Computer Aided Systems Theory Las Palmas de Gran Canaria, Spain, February 23- 27, 2026, [Abstract]: This work researches the impact of retina-focused preprocessing on OCT-based deep learning classification of Multiple Sclerosis (MS). Raw OCT B-scans were compared with anatomically preprocessed images, where the retina was segmented, flattened, and isolated from background structures. Experiments with several convolutional architectures showed that focusing on the retinal region enhances classification accuracy and consistency across repetitions, supporting the relevance of anatomically guided preprocessing in OCT-based neurological screening., This work was supported by the Instituto de Salud Carlos III (ISCIII), Government of Spain [grant numbers PI17/01726, PI20/00437, PI23/00935, RD21/0007/0022 (Inflammatory Disease Network- RICORS), FORT23/00010 (Programa FORTALECE)]; MICIU/AEI/10.13039/501100011033 and by “ERDF A way of making Europe” [grant numbers PID2023-148913OB-I00, PID2024-161024OB-I00]; the Consellería de Educación, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva [grant number ED431C 2024/33]; and by the Government of Aragon [group B23_23R and PROY_B50_24].
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Detección de glistenings en lentes intraoculares con OCT-SA mediante deep learning multi-tarea

  • Álvarez-Rodríguez, Lorena
  • Moura, Joaquim de
  • Fernández-Vigo, José Ignacio
  • Macarro-Merino, Ana
  • Fernández-Vigo, José A.
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Poster presentado en BioIntegraSaúde - BIS 2025, Santiago de Compostela, 15 de maio de 2025., [Resumen]: Las lentes intraoculares (LIO) son dispositivos ópticos implantados quirúrgicamente en la cápsula posterior del cristalino, habitualmente tras cirugía de cataratas u otros trastornos del cristalino. Estas lentes pueden desarrollar glistenings, microvacuolas llenas de agua que aparecen en su interior y pueden afectar a la transparencia y calidad óptica. Tradicionalmente, su evaluación se ha realizado mediante fotografía con lámpara de hendidura, una técnica sensible a factores externos como la iluminación o la dilatación pupilar. Recientemente, se ha propuesto la tomografía de coherencia óptica del segmento anterior (OCT-SA) como alternativa más rápida y objetiva, con buenos resultados preliminares. Con el objetivo de facilitar la evaluación de las LIO y validar el uso de la OCT-SA, presentamos un enfoque automático basado en deep learning para detectar y caracterizar glistenings en imágenes OCT-SA. Desarrollamos un modelo multi-tarea que identifica simultáneamente las zonas con glistenings y el área ocupada por la LIO (LIOa). Este enfoque mejora la precisión, ya que la localización de la lente aporta un contexto anatómico clave para interpretar correctamente la presencia y distribución de los glistenings. A su vez, la detección precisa de la LIOa es esencial para su evaluación estructural y seguimiento clínico. Nuestro estudio es el primero en aplicar esta metodología combinada y validarla sobre un conjunto de datos clínico altamente heterogéneo, con imágenes de 19 modelos distintos de LIO y diversos grados de opacificación. El modelo alcanzó un alto grado de acuerdo con las anotaciones de expertos (ICC > 0.93), a la par de estudios previos con enfoques más simples y menos variedad de datos. Estos resultados confirman el potencial de los métodos automáticos basados en deep learning para facilitar una evaluación más rápida, objetiva y reproducible de los glistenings en entornos clínicos, con especial utilidad en el seguimiento de pacientes con LIO., Este trabajo es parte del proyecto PID2023-148913OB-I00 financiado por MICIU/AEI/10.13039/501100011033 y por ”FEDER Una manera de hacer Europa”. Ha sido también financiado por el Instituto de Salud Carlos III (Programa FORTALECE, FORT23/00010) y la Xunta de Galicia mediante la Consellería de Educación, Universidade e Formación Profesional (Grupos de Referencia Competitiva, ED431C 2024/33). Las organizaciones financiadoras no participaron en el diseño o ejecución de esta investigación., Xunta de Galicia; ED431C 2024/33
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Fully-automatic end-to-end approaches for 3D drusen segmentation in Optical Coherence Tomography images

  • Goyanes, Elena
  • Leyva Santarén, Saúl
  • Herrero, Paula
  • Moura, Joaquim de
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Presented at: 28th International Conference on Knowledge Based and Intelligent information and Engineering Systems (KES 2024), [Abstract]: 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., This work was supported by Ministerio de Ciencia e Innovacion, Government of Spain through the research project with [grant numbers PID2019-108435RB-I00, TED2021-131201B-I00, and PDC2022-133132-I00]; Consellería de Educacion, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [grant number ED431C 2020/24] and predoctoral grant [grant number ED481A-2023-152]. Furthermore, this work was supported by the Instituto de Salud Carlos III (ISCIII) under the grant [FORT23/00010] as part of the Programa FORTALECE del Ministerio de Ciencia e Innovacion., Xunta de Galicia; ED431C 2020/24, Xunta de Galicia; ED481A-2023-152
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Multi-depth transfer learning-based approaches via generative models for foveal avascular zone segmentation in OCTA images

  • Regueiro Feal, Ángel
  • Goyanes, Elena
  • Moura, Joaquim de
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
This version of the article has been accepted for publication, after peer review. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The Version of Record is available online at: https://doi.org/10.1109/IJCNN64981.2025.11227613 Traballo presentado en:2025 International Joint Conference on Neural Networks (IJCNN), Roma, Italia, 30 June-05 July 2025, [Abstract]: 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., This work was supported by Ministerio de Ciencia e Innovación, Government of Spain through the research project with [grant numbers PID2023- 148913OB-I00, TED2021-131201B-I00, and PDC2022-133132-I00]; Consellería de Educación, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [grant number ED431C 2024/33], predoctoral grant [grant number ED481A-2023-152]. Also supported by the ISCIII under the grant [FORT23/00010] as part of the Programa FORTALECE of Ministerio de Ciencia e Innovación., Xunta de Galicia; ED431C 2024/33, Xunta de Galicia; ED481A-2023-152
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Análisis multicentro del cribado automático de glaucoma en pacientes con alta miopía a través de retinografías

  • Millán Guerrero, Ulises
  • Goyanes, Elena
  • Moura, Joaquim de
  • Magadán Salazar, Andrea
  • Moctezuma Ochoa, Daniela
  • Martínez-de-la-Casa, José María
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Poster presentado a BioIntegraSaúde 2025, co-organizado por los tres institutos: IDIS, IISGS e INIBIC, en Santiago de Compostela, 15 de mayo de 2025., [Resumen]: El glaucoma es una causa principal de ceguera irreversible en todo el mundo, siendo la detección temprana esencial para prevenir la pérdida de visión. Sin embargo, diagnosticar glaucoma en pacientes altamente miopes presenta desafíos significativos debido a alteraciones anatómicas, como la deformación de la cabeza del nervio óptico y el adelgazamiento de la capa de fibras nerviosas de la retina, que pueden oscurecer los cambios relacionados con la enfermedad y llevar a un diagnóstico erróneo. Para abordar estas limitaciones, proponemos un sistema basado en técnicas avanzadas de inteligencia artificial para la detección automatizada de glaucoma en ojos altamente miopes. Nuestro enfoque aprovecha el aprendizaje multi-transferencia, integrando pre-entrenamiento a gran escala con adaptación específica del dominio utilizando un conjunto formado por 10,000 imágenes de 5,000 pacientes. Además, se realiza un análisis entre múltiples centros clínicos, cada uno empleando un retinógrafo distinto, lo que permite evaluar la robustez y aplicabilidad del modelo en escenarios reales y diversos. Los centros participantes son el Hospital Clínico San Carlos (HCSC), la Clínica Universidad de Navarra (CUN), el Hospital Ramón y Cajal (RYC) y el Instituto Universitario Fernández-Vega de Oviedo (IRU). Esta colaboración multicéntrica asegura una mayor representatividad de los datos, y además refuerza la capacidad del sistema para adaptarse a las variaciones en los equipos de adquisición, un desafío clave en la práctica clínica. Los resultados experimentales demuestran que el modelo propuesto supera consistentemente los métodos de referencia, logrando una precisión y robustez superiores en la detección de glaucoma dentro de poblaciones altamente miopes. Estos hallazgos subrayan el potencial de las herramientas de detección impulsadas por IA como ayudas diagnósticas confiables y precisas, apoyando a los clínicos en la detección temprana y el manejo efectivo del glaucoma en casos complejos., Este trabajo es parte del proyecto PID2023-148913OB-I00 financiado por MICIU/AEI/10.13039/501100011033 y por ”FEDER Una manera de hacer Europa”. Ha sido también financiado por la Consellería de Educación, Universidade e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [código ED431C 2024/33], ayuda predoctoral [ED481A-2023-152]. Además, está apoyado por el ISCIII bajo la ayuda [FORT23/00010] como parte del Programa FORTALECE del Ministerio de Ciencia e Innovación; y el proyecto de investigación [PI23/00828]: Desarrollo evaluación de un algoritmo de detección de glaucoma a partir de un abordaje multimodal en pacientes con miopía magna., Xunta de Galicia; ED431C 2024/33, Xunta de Galicia; ED481A-2023-152
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Deep Learning-Based Models for Sickle Cell Anemia Characterization in Retinal Fundus Images

  • Carrasco, L.
  • Ramos, Lucía
  • Barreira, Noelia
  • S. Hervella, Álvaro
  • Raffa, Lina H.
This version of the paper has been accepted for publication. The final published paper is available online at: https://doi.org/10.1109/CBMS65348.2025.00165. Presented at: 2025 IEEE 38th International Symposium on Computer-Based Medical Systems (CBMS), 18-20 June 2025, Madrid, Spain., [Abstract]: Sickle cell disease (SCD) is a severe hereditary disorder that affects multiple systems and compromises blood flow, potentially leading to serious complications such as tissue damage and vision loss. Despite treatment developments, early detection is essential, even more so in regions with limited resources. In this context, clinical manifestations observed in ocular fundus images, such as vascular tortuosity, provide the opportunity to detect SCD non-invasively by applying artificial intelligence algorithms and techniques. This study analyzes the potential of deep learning methods to detect SCD using ocular fundus images and to identify relevant retinal patterns in this disorder, which leads to improved comprehension and clinical management of the disease. To this end, we used ocular fundus images from SCD patients and healthy controls to train and evaluate several models of neural networks models, including CNN and a hybrid CNN-Transformer Vision model. In addition, activation maps were built to identify the most relevant retinal characteristics for the classification problem. ResNet-50 and EfficientNet-b0 models showed better performance in the F1 score metric, getting 88 % and 83 % values, respectively. The activation maps analyze highlighted vascular tortuosity as an important feature of disorder detection. Notwithstanding certain limitations, such as the size of the data set or the variability between the images, the results obtained are promising. Solving these problems could improve the effectiveness of the models for the detection and characterization of this disorder., This work was supported by the R&D&I project under grant ref. PID2023-148913OB-I00 funded by MICIU/AEI/10.13039/501100011033 and by “ERDF/EU”, the Consellería de Educación, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, under grant ref ED431C 2024/33 and postdoctoral fellowship ref. ED481B-2022-025. Additional support was provided by the Instituto de Salud Carlos III (ISCIII) under grant ref. FORT23/00010, part of the Programa FORTALECE from the Ministerio de Ciencia e Innovación., Xunta de Galicia; ED431C 2024/33, Xunta de Galicia; ED481B-2022-025
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Recurrent Task Specialization Network for Segmentation-aided Vascular Landmarks Detection in Retinal Images

  • Hervella, Álvaro S.
  • Rouco, José
  • Novo Buján, Jorge
  • Sánchez, Clara I.
  • Ortega Hortas, Marcos
Presented at the 27th European Conference on Artificial Intelligence, ECAI 2024, Santiago de Compostela 19-24 October 2024., [Abstract]: 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., This work is supported by Ministerio de Ciencia e Innovación, Government of Spain, through the PID2019-108435RB-I00, TED2021-131201B-I00, and PDC2022-133132-I00 research projects; Consellería de Cultura, Educación e Universidade, Xunta de Galicia, through Grupos de Referencia Competitiva ref. ED431C 2020/24 and the postdoctoral fellowship ref. ED481B-2022-025; and Instituto de Salud Carlos III under the grant [FORT23/00010] as part of Programa FORTALECE from Ministerio de Ciencia e Innovación., Xunta de Galicia; ED431C 2020/24, Xunta de Galicia; ED481B-2022-025
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Sistema automático y explicable para el cribado simultáneo de múltiples patologías pulmonares en radiografías de tórax mediante aprendizaje profundo

  • Goyanes, Elena
  • Moura, Joaquim de
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Poster presentado a BioIntegraSaúde 2025, co-organizado por los tres institutos: IDIS, IISGS e INIBIC, en Santiago de Compostela, 15 de mayo de 2025., [Resumen]: En un contexto clínico donde la rapidez y precisión diagnóstica son fundamentales, este estudio presenta un enfoque basado en inteligencia artificial para el cribado automatizado y simultáneo de múltiples enfermedades pulmonares a partir de radiografías de tórax. A través de un modelo de aprendizaje profundo de clasificación multi- etiqueta, se ha desarrollado una herramienta capaz de identificar de manera eficiente 14 patologías pulmonares distintas (atelectasia, cardiomegalia, derrame pleural, infiltración, masa, nódulos, neumonía, neumotórax, consolidación, edema, enfisema, fibrosis, engrosamiento pleural y hernia), abordando así uno de los principales desafíos del diagnóstico por imagen: la coocurrencia de enfermedades y la complejidad inherente a su diferenciación visual. El sistema propuesto fue entrenado y validado con un conjunto masivo proveniente del National Institutes of Health (NIH) Clinical Center de Estados Unidos, compuesto por 112120 radiografías de 30805 pacientes, lo que garantiza una base sólida y representativa para evaluar su rendimiento en escenarios clínicos reales. Los resultados obtenidos muestran un desempeño robusto y generalizable, posicionando a este sistema como una solución prometedora para aliviar la carga diagnóstica en servicios radiológicos de alta demanda. Además, el modelo integra una herramienta de explicabilidad a través de mapas de activación Grad-CAM, lo que permite visualizar las regiones clave de las radiografías utilizadas para cada predicción. Esta capacidad de interpretación no solo incrementa la transparencia del sistema, sino que también refuerza la confianza de los profesionales de la salud, facilitando una toma de decisiones más precisa., Este trabajo es parte del proyecto PID2023-148913OB-I00 financiado por MICIU/AEI/10.13039/501100011033 y por ”FEDER Una manera de hacer Europa”. Ha sido también financiado por la Consellería de Educación, Universidade e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [código ED431C 2024/33], ayuda predoctoral [ED481A-2023-152]. Además, está apoyado por el ISCIII bajo la ayuda [FORT23/00010] como parte del Programa FORTALECE del Ministerio de Ciencia e Innovación., Xunta de Galicia; ED431C 2024/33, Xunta de Galicia; ED481A-2023-152
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Geometric Deep Learning for Essential Tremor Screening Using

  • Álvarez-Rodríguez, Lorena
  • Moura, Joaquim de
  • Viladés, Elisa
  • García-Martín, Elena
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
This version of the article has been accepted for publication, after peer review. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The Version of Record is available online at: https://doi.org/10.1109/IJCNN64981.2025.11229335 Traballo presentado en: 2025 International Joint Conference on Neural Networks (IJCNN), Rome, Italy, 30 June - 05 July 2025, [Abstract]: 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., This work was supported by the Instituto de Salud Carlos III (ISCIII), Government of Spain [grant numbers PI17/01726, PI20/00437, PI23/00935, RD21/0002/0050 (Inflammatory Disease Network - RICORS), FORT23/00010 (Programa FORTALECE)], the Ministerio de Ciencia e Innovacion, Government of Spain [grant numbers PID2023-148913OB-I00, TED2021-131201B-I00, and PDC2022-133132-I00], the Consellería de Educacion, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva [grant number ED431C 2024/33] and by the Government of Aragon [group B2323R]. This work was also supported by the Horizon Europe Programme through the ACHILLES/101189689 project (HORIZON-CL4-2024-DATA-01-01)., Xunta de Galicia; ED431C 2024/33
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Multi-Disease Explainable Deep Learning System for Automated Chest X-ray Screening

  • Goyanes, Elena
  • Fernández Moreira, Breogán
  • Moura, Joaquim de
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Presentado en: EUROCAST 2026, 20th International Conference on Computer Aided Systems Theory Las Palmas de Gran Canaria, Spain, February 23- 27, 2026, [Abstract]: This paper presents an explainable deep learning system for the simultaneous screening of multiple pulmonary diseases in chest Xrays. Ten state-of-the-art convolutional and transformer-based architectures were evaluated to identify the most effective model for multi-label classification, with the best-performing model achieving an average AUC of 0.841. The interpretability of the system was enhanced using Grad-CAM, providing visual explanations consistent with clinical findings., This work was supported by Government of Spain through the research projects with [grant number PID2023-148913OB-I00] and [grant number PID2024-161024OB-I00], funded by MICIU/AEI/10.13039/501100011033 and by ERDF, EU; Consellería de Educación, Universidade e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [grant number ED431C 2024/33], predoctoral grant [grant number ED481A-2023-152]. Also supported by the ISCIII under the grant [FORT23/00010] as part of the Programa FORTALECE of MICIU., Xunta de Galicia; ED431C 2024/33, Xunta de Galicia; ED481A-2023-152
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

3D Point Cloud Analysis via Transformer-Based Graph Learning for Multiple Sclerosis Screening in OCT Images

  • Álvarez-Rodríguez, Lorena
  • García Prego, Iván
  • Moura, Joaquim de
  • Pueyo-Bestué, Ana
  • Viladés, Elisa
  • García-Martín, Elena
  • Sánchez, Clara I.
  • Novo Buján, Jorge
  • Ortega Hortas, Marcos
Presented at: 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES 2024), [Abstract]: 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., Funded by Instituto de Salud Carlos III (ISCIII), Government of Spain, [projects PI17/01726 and PI20/00437]; Inflammatory Disease Network (RICORS) [project RD21/0002/0050]; Ministerio de Ciencia e Innovacion, Government of Spain [ projects PID2019-108435RB-I00, TED2021-131201B-I00, and PDC2022-133132-I00]; Consellería de Educacion, Universidade, e Formación Profesional, Xunta de Galicia, Grupos de Referencia Competitiva, [grant number ED431C 2020/24]. Also supported by the ISCIII under the grant [FORT23/00010] as part of the Programa FORTALECE of Ministerio de Ciencia e Innovacion., Xunta de Galicia; ED431C 2020/24
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Contribución de congreso (ConferenceOutput)

Geometric Deep Learning for Essential Tremor Screening Using OCT-Derived 3D Point Clouds

  • Álvarez-Rodríguez, Lorena
  • de Moura, Joaquim
  • Vilades, Elisa
  • Garcia-Martin, Elena
  • Novo, Jorge
  • Ortega, Marcos
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.