Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

Digibug. Repositorio Institucional de la Universidad de Granada
Publicación Estudio, informe, memoria (Report)

Cancer Survivorship and AI for Well-being White Paper: A European research projects collaborative perspective

  • Callejas Carrión, Zoraida|||0000-0001-8891-5237
  • Flynn, Tom
The Cancer Survivorship - AI for Well-being cluster (#CS_AIW) has brought together 11 EU-funded projects working in artificial intelligence (AI) for healthcare and well-being with the aim of transcending the individual project experiences and stop working in silos. The White Paper presents the #CS_AIW initiative and our collective shared vision, providing an overview of the participating projects and it reports on the achievements and lessons learned from our collaboration, including both the benefits obtained and the difficulties addressed. The experience with #CS_AIW to date has proven that collaboration within the cluster has been extremely beneficial and useful to foster an ethos of cross-fertilization engagement, share best practices, boost cross-sector collaboration, increase the participation of end user communities, and reach a wider audience. The activities organized, which encompass meetings, roundtable discussions, podcasts, a better practices guide, a common data space and the white paper itself, have proven to be an inherited valued initiative to generate, activate and communicate knowledge and innovation in the areas of mental health, well-being, cancer recovery, patient support and participatory research., The projects in #CS_AIW have received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement no.: ASCAPE - Ref 875351 BD4QoL - Ref 875192 CAPABLE - Ref 875052 CLARIFY - Ref 860627 FAITH - Ref 875358 LIFECHAMPS - Ref 875329 MENHIR - Ref 823907 ONCORELIEF - Ref 875392 PERSIST - Ref 875406 QUALITOP - Ref 875171 REBECCA - Ref 65231
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Tesis doctoral (DoctoralThesis)

Leveraging routinely generated healthcare data through intelligent systems to enhance clinical decision-making for cancer survivors, Explotación de los datos sanitarios generados de manera rutinaria a través de sistemas inteligentes para mejorar la toma de decisiones clínicas para los sobrevivientes de cáncer, Explotación dos datos sanitarios xerados de forma rutinaria a través de sistemas intelixentes para mellorar a toma de decisións clínicas para os sobreviventes de cancro

  • González Castro, Lorena
Cancer survivorship care presents significant challenges due to the long-term effects and comorbidities associated with post-treatment care. These late effects, which may persist for years, can include physical, psychological, and cognitive sequelae, as well as an elevated risk of developing additional comorbid conditions. With advancements in early detection and treatment, cancer survival rates have improved considerably, leading to a growing population of survivors worldwide. This trend highlights an urgent need for personalized, continuous, and comprehensive follow-up care strategies. With the increasing adoption of electronic health records (EHRs), a wealth of clinical data is now available, offering unprecedented opportunities to enhance survivorship care through data-driven approaches. This thesis leverages routinely generated healthcare data and advanced machine learning (ML) techniques to support clinical decision-making and improve survivorship care. In the first place, we developed CASIDE (CAncer Survivorship Interoperable Data Elements), a data model based on the Fast Healthcare Interoperability Resources (FHIR) standard. CASIDE integrates diverse data sources, including electronic health records (EHRs), patient-reported outcomes, and wearable sensor data, into a unified framework for cancer survivorship information. This model ensures semantic interoperability and supports secondary use of data, enabling seamless information exchange across institutions. CASIDE was validated through a multicenter clinical study involving breast and colon cancer survivors, demonstrating its effectiveness in real-world settings, facilitating collaboration and enhancing the quality of survivorship data. Building upon this foundation, we explored the secondary use of structured and unstructured data from EHRs to predict 5-year breast cancer recurrence using ML algorithms. To this end, we constructed three datasets: one using structured data from EHRs, another based on features extracted from unstructured clinical notes, and a combined dataset integrating both sources. These datasets were used to evaluate whether combining structured and unstructured data could improve predictive accuracy over using either source alone. Our findings revealed that structured data alone provided the highest predictive performance, while unstructured could provide added value in cases where structured information was unavailable. This underscores the need for careful curation of EHR data to maximize the effectiveness of ML models in clinical applications. Subsequently, we examined the pivotal role of hyperparameter optimization in enhancing ML model performance. Five different models—Logistic Regression (LR), Decision Tree (DT), Gradient Boosting (GB), eXtreme Gradient Boosting (XGB), and Deep Neural Networks (DNN)—were evaluated before and after systematic hyperparameter tuning. Results showed significant performance improvements, especially in complex models like XGB and DNN. This highlights the importance of hyperparameter optimization as a standard practice for developing reliable and effective predictive tools in healthcare. By addressing key challenges in data interoperability and predictive analytics, this thesis contributes to advancing personalized and data-driven survivorship care. The findings highlight the transformative potential of integrating interoperable data models with AI techniques to translate healthcare data into actionable insights. These tools empower clinicians to make informed decisions, improve follow-up care, and ultimately enhance patient outcomes and quality of life for cancer survivors., La inteligencia artificial (IA) promete transformar los procesos de toma de decisiones clínicas ya que tiene el potencial de aprovechar la gran cantidad de datos clínicos que se están generando en todo el sistema de salud. Sin embargo, existen muchos desafíos que dificultan su adopción en la práctica clínica habitual, como la falta de estandarización, la naturaleza no estructurada de mucha información clínica o la falta de confianza en la efectividad de los modelos de IA. El objetivo general de la tesis será la aplicación de técnicas de Machine Learning (ML) para la extracción de información relevante y procesable de la Historia Clínica Electrónica (HCE) que permita apoyar la toma de decisiones clínicas, basada en la evidencia extraída de la práctica clínica, con el fin de mejorar la prestación de cuidados y aumentar la calidad de vida de los pacientes. Se aplicarán estándares de interoperabilidad médica, terminologías y ontologías del dominio biomédico para favorecer la interoperabilidad de los desarrollos, maximizando así su impacto. Se realizará un análisis de los datos de la HCE para identificar cohortes y trayectorias comunes, modelando los itinerarios seguidos por los pacientes. A partir de estas trayectorias, se desarrollarán modelos predictivos (p. ej. probabilidades de recaída, estratificación del riesgo) que puedan utilizarse para personalizar y optimizar el tratamiento, o para prevenir eventos adversos y reingresos hospitalarios. Se hará especial énfasis en la aplicación de las mejores prácticas para desarrollar modelos de ML robustos y confiables. Este estudio se centrará principalmente en mejorar la toma de decisiones clínicas en el ámbito del cáncer. El caso de estudio principal será la investigación y el desarrollo de herramientas y estrategias para mejorar la atención a la supervivencia del cáncer y la calidad de vida después del tratamiento. No obstante, las actividades de investigación podrían trasladarse a otros casos de uso, según los medios disponibles en ese momento. Este trabajo de investigación se lleva a cabo bajo la modalidad de “doctorado industrial”, en colaboración con GRADIANT, Centro Tecnológico de Telecomunicaciones de Galicia. GRADIANT participa y coordina el proyecto Horizonte 2020 PERSIST: “Patients-centered survival care plan after cancer treatments based on Big Data and Artificial Intelligence Technologies”. Este proyecto proporcionará el marco y el soporte necesario para el trabajo de investigación que se lleva a cabo en este doctorado., A intelixencia artificial (IA) promete transformar os procesos de toma de decisións clínicas xa que ten o potencial de aproveitar a gran cantidade de datos clínicos que se xeran no sistema sanitario. Non obstante, son moitos os retos que dificultan a súa adopción na práctica clínica rutinaria, como a falta de estandarización, a natureza non estruturada de moita información clínica ou a falta de confianza na eficacia dos modelos de IA. O obxectivo xeral da tese será a aplicación de técnicas de Machine Learning (ML) para a extracción de información relevante e accionable da Historia Clínica Electrónica (HCE) que permita apoiar a toma de decisións clínicas, a partir da evidencia extraída da práctica clínica, co fin de mellorar a prestación de coidados e aumentar a calidade de vida dos pacientes. Aplicaranse estándares de interoperabilidade médica, terminoloxías e ontoloxías do dominio biomédico para promover a interoperabilidade dos desenvolvementos, maximizando así o seu impacto. Realizarase unha análise dos datos de HCE para identificar cohortes e traxectorias comúns, modelando os itinerarios seguidos polos pacientes. A partir destas traxectorias, desenvolveranse modelos preditivos (por exemplo, probabilidades de recaída, estratificación do risco) que se poden utilizar para personalizar e optimizar o tratamento, ou para previr eventos adversos e reingresos hospitalarios. Farase especial fincapé na aplicación das mellores prácticas para desenvolver modelos de ML robustos e fiables. Este estudo centrarase principalmente en mellorar a toma de decisións clínicas no ámbito do cancro. O estudo de caso principal será a investigación e o desenvolvemento de ferramentas e estratexias para mellorar a atención de supervivencia do cancro e a calidade de vida despois do tratamento. Non obstante, as actividades de investigación poderían trasladarse a outros casos de uso, dependendo dos medios dispoñibles nese momento. Este traballo de investigación realízase na modalidade de “doutoramento industrial”, en colaboración con GRADIANT, Centro Tecnolóxico de Telecomunicacións de Galicia. GRADIANT participa e coordina o proxecto Horizon 2020 PERSIST: “Plan de coidados de supervivencia centrado no paciente despois de tratamentos contra o cancro baseados en tecnoloxías de Big Data e Intelixencia Artificial”. Este proxecto proporcionará o marco e o apoio necesarios para o traballo de investigación que se desenvolve neste doutoramento.
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Machine learning algorithms to predict breast cancer recurrence using structured and unstructured sources from electronic health records

  • González Castro, Lorena
  • Chávez, Marcela
  • Duflot, Patrick
  • Bleret, Valérie
  • Martin, Alistair G.
  • Zobel, Marc
  • Nateqi, Jama
  • Lin, Simon
  • Pazos Arias, José Juan
  • Del Fiol, Guilherme
  • López Nores, Martín
Recurrence is a critical aspect of breast cancer (BC) that is inexorably tied to mortality. Reuse of healthcare data through Machine Learning (ML) algorithms offers great opportunities to improve the stratification of patients at risk of cancer recurrence. We hypothesized that combining features from structured and unstructured sources would provide better prediction results for 5-year cancer recurrence than either source alone. We collected and preprocessed clinical data from a cohort of BC patients, resulting in 823 valid subjects for analysis. We derived three sets of features: structured information, features from free text, and a combination of both. We evaluated the performance of five ML algorithms to predict 5-year cancer recurrence and selected the best-performing to test our hypothesis. The XGB (eXtreme Gradient Boosting) model yielded the best performance among the five evaluated algorithms, with precision = 0.900, recall = 0.907, F1-score = 0.897, and area under the receiver operating characteristic AUROC = 0.807. The best prediction results were achieved with the structured dataset, followed by the unstructured dataset, while the combined dataset achieved the poorest performance. ML algorithms for BC recurrence prediction are valuable tools to improve patient risk stratification, help with post-cancer monitoring, and plan more effective follow-up. Structured data provides the best results when fed to ML algorithms. However, an approach based on natural language processing offers comparable results while potentially requiring less mapping effort., European Union | Ref. 875406, Fondo Europeo de Desarrollo Regional (FEDER), Xunta de Galicia