Pasar al contenido principal

ASPECTOS MATEMATICOS DEL PROCESAMIENTO DE IMAGENES BIOMEDICAS: METODOS TOPOLOGICOS Y DE CIENCIA DE DATOS

PID2020-115225RB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2020
Plan Estatal 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDAD DE LA RIOJA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

La perplejidad como herramienta para estimar la asignación de nivel de competencia en escritos de una lengua extranjera, Perplexity as a tool for the allocation of proficiency levels to utterances written by foreign language learners

  • Agustín Llach, María del Pilar
  • Heras, Jonathan
  • Mata, Gadea
  • Rubio, Julio
La asignación de niveles de competencia a escritos producidos por aprendices de una lengua es una tarea altamente subjetiva. Es por esto que el desarrollo de métodos que evalúen escritos de manera automática puede ayudar tanto al profesorado como al alumnado. En este trabajo, hemos explorado dos vías mediante el uso del corpus CAES. Dicho corpus está formado por escritos de aprendices de español y etiquetado con niveles CEFR (hasta el C1). La primera aproximación es un modelo de aprendizaje profundo llamado Deep-ELE que asigna niveles de competencia a las frases. La segunda aproximación llevada a cabo ha consistido en estudiar la perplejidad de las frases de los estudiantes de distintos niveles, para luego clasificarlos en niveles. Ambas aproximaciones han sido evaluadas, y se ha comprobado que pueden usarse de manera exitosa para clasificar frases por niveles. En concreto, el modelo Deep-ELE obtiene una accuracy de 81,3% y un QWK de 0,83. Como conclusión, este trabajo es un paso para entender cómo las herramientas del procesado de lenguaje natural pueden ayudar a las personas que aprenden un segundo idioma.
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Deep Learning Models for Justified Referral in AI Glaucoma Screening

  • Casado-García, Ángela
  • Heras, Jónathan
  • Ortega Hortas, Marcos
  • Ramos, Lucía
Presented at: 2024 IEEE International Symposium on Biomedical Imaging (ISBI), 27-30 May 2024, Athens, Greece 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/ISBI56570.2024.10635680, [Abstract]: Glaucoma is an optic disease that leads to blindness, but this might be avoided with an early diagnosis thanks to a screening test. The JustRAIGS Challenge was organised to develop solutions for glaucoma screening from retinal fundus images that not only classify fundus images as "referrable"or "no referable"but also identify specific characteristics or abnormalities that may be present in the fundus images of glaucoma patients. In this work, we present our solution to this challenge based on the study of several combinations of fundus images and their associated optic disc and cup. For the binary task, our solution consists of the ensemble of a ConvNext model and a Swin model that achieves a sensitivity at 95% specificity of 0.8570; whereas, for the multi-label task, our ConvNext model achieved a Hamming Loss of 0.1930., This work was partially supported by Grant PID2020-115225RBI00 funded by MCIN/AEI/10.13039/501100011033, and by Ministerio de Ciencia e Innovación, Government of Spain through the research project with grant number PID2019-108435RB-I00. Ángela Casado-García has a FPI grant from Community of La Rioja 2020, at the moment, she is in the Postdoctoral Orientation Period (POP).
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Application of multi-regression machine learning algorithms to solve ocean water mass mixing in the Atlantic Ocean

  • Romera-Castillo, Cristina
  • Heras, Jónathan
  • Álvarez-Rodríguez, Marta
  • Álvarez-Salgado, Xosé Antón
  • Mata, Gadea
  • Sánez-de-Cabezón, Eduardo
14 pages, 10 figures, 6 tables.-- This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY), The distribution of any non-conservative variable in the deep open ocean results from the circulation and mixing of water masses (WMs) of contrasting origin and from the initial preformed composition, modified during ongoing simultaneous biological and/or geochemical processes. Estimating the contribution of the WMs composing a sample is useful to trace the distribution of each water mass and to quantitatively separate the physical (mixing) and biogeochemical components of the variability of any, non- conservative variable (e.g., dissolved organic carbon, prokaryote biomass) in the ocean. Other than potential temperature and salinity, additional semi-conservative and non-conservative variables have been used to solve the mixing of more than three water masses using Optimum Multi-Parameter (OMP) approaches. Successful application of an OMP analysis requires knowledge of the characteristics of the water masses in their source regions as well as their circulation and mixing patterns. Here, we propose the application of multi-regression machine learning models to solve ocean water mass mixing. The models tested were trained using the solutions from OMP analyses previously applied to samples from cruises in the Atlantic Ocean. Extremely Randomized Trees algorithm yielded the highest score (R2 = 0.9931; mse = 0.000227). Our model allows solving the mixing of water masses in the Atlantic Ocean using potential temperature, salinity, latitude, longitude and depth. Therefore, basic hydrographic data collected during typical research cruises or autonomous systems can be used as input variables and provide results in real time. The model can be fed with new solutions from compatible OMP analyses as well as with new water masses not previously considered in it. Our tool will provide knowledge on water mass composition and distribution to a broader community of marine scientists not specialized in OMP analysis and/or in the oceanography of the studied area. This will allow a quantitative analysis of the effect of water mass mixing on the variables or processes under study, CR-C was funded by grant PID2019-109889RJ-I00 / AEI / 10.13039/501100011033 (Ministerio de Ciencia e Innovación and Agencia Estatal de Investigación, Spain). GM and JH were partially supported by Ministerio de Ciencia e Innovación [PID2020-115225RB-I00 / AEI / 10.13039/501100011033]. ES-d-C was partially supported by Ministerio de Ciencia e Innovación [PID2020-116641GB-100 / AEI / 10.13039/501100011033]. XA–S was partially funded by grant number PID2019-109084RB-C22 (Ministerio de Ciencia e Innovación, Spain). MA was funded by RADPROF and RADIALES IEO-CSIC projects, With the institutional support of the ‘Severo Ochoa Centre of Excellence’ accreditation (CEX2019-000928-S), Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Otros (Other)

Replace, Paraphrase or Fine-tune? Evaluating Automatic Simplification for Medical Texts in Spanish

  • Campillos-Llanos, Leonardo
  • Terroba Reinares, Ana Rosa
  • Bartolomé, Rocío
  • Valverde, Ana
  • González, Cristina
  • Capllonch Carrión, Adrián
  • Heras, Jónathan
Este capítulo está sujeto a una licencia CC BY-NC 4.0, Patients can not always completely understand medical documents given the myriad of technical terms they contain. Automatic text simplification techniques can help, but they must guarantee that the content is transmitted rigorously and not creating wrong information. In this work, we tested: 1) lexicon-based simplification approaches, using a Spanish lexicon of technical and laymen terms collected for this task (SimpMedLexSp); 2) deep-learning (DL) based methods, with BART-based and prompt-learning-based models; and 3) a combination of both techniques. As a test set, we used 5000 parallel (technical and laymen) sentence pairs: 3800 manually aligned sentences from the CLARA-MeD corpus; and 1200 sentences from clinical trials simplified by linguists. We conducted a quantitative evaluation with standard measures (BLEU, ROUGE and SARI) and a human evaluation, in which eleven subjects scored the simplification output of several methods. In our experiments, the lexicon improved the quantitative results when combined with the DL models. The simplified sentences using only the lexicon were assessed with the highest scores regarding semantic adequacy; however, their fluency needs to be improved. The prompt-method had similar ratings in this aspect and in simplification. We make available the models and the data to reproduce our results., This work was done in project CLARAMED (PID2020-116001RA-C33) funded by MCIN/AEI/10.13039/501100011033/, in call: "Proyectos I+D+i Retos Investigación"; and also partially supported by Grant PID2020-115225RBI00 funded by MCIN/AEI/10.13039/501100011033., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Transformer-based relation extraction and concept normalization using an annotated clinical trials corpus

  • Campillos-Llanos, Leonardo
  • Valverde Mateos, Ana
  • Capllonch Carrión, Adrián
  • González-Quevedo, David
  • López-Urbán, María Rosa
  • Hernando-Tundidor, Soledad
  • Zakhir Puig, Sofía
  • Heras, Jónathan
Healthcare professionals manually review electronic health records to select patients who meet eligibility criteria of clinical trials. Natural language processing offers a complement for this task, although few initiatives exist in Spanish. We present version 3 of the CT-EBM-SP corpus of 1200 clinical trials (292173 tokens), annotated with 23 entity types and 18 relation types, covering Unified Medical Language System (UMLS) semantic groups, drug-related information, temporal data, and negation/speculation. We encoded 11 attributes (e.g., event temporality and experiencer status) and normalized entities to UMLS Concept Unique Identifiers. The corpus contains 87037 entities, including nested and discontinuous entities, 16597 attributes and 68206 relationships. Inter-annotator agreement (IAA) achieved average F1 values of 0.861 (entities), 0.810 (attributes), and 0.791 (relations). 81.75% of entities were normalized (IAA: F1 = 0.966). We benchmarked this dataset by fine-tuning Transformer models for relation extraction (RE) and medical concept normalization (MCN). In RE, the average F1 ranged from 0.858 to 0.879, and for MCN, the accuracy at rank 1 was 0.896. The corpus and models are publicly available., This work was conducted in the CLARA-MED project (PID2020-116001RA-C33), the ExPlain4Health project (PID2024-158912NB-I00), and partly (JH) by project PID2020-115225RB-I00, all funded by MCIN/AEI/10.13039/501100011033, in project call: Proyectos I+D+i Generación de conocimiento (Retos Investigación). Sofía Zakhir was funded by a JAE Intro 2021 scholarship at CSIC., Peer reviewed