Skip to main content

INTELIGENCIA ARTIFICIAL INSESGADA Y EXPLICABLE PARA IMAGENES MEDICAS

PID2020-116907RB-I00

•
Funding agency name Agencia Estatal de Investigación
Funding agency acronym AEI
Program Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Call Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Call year 2020
State Plan 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Beneficiary centre UNIVERSITAT POLITECNICA DE CATALUNYA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 14
Encontrada(s) 1 página(s)

UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Structural networks for brain age prediction

  • Pina Benages, Òscar|||0009-0003-9755-6629
  • Cumplido Mayoral, Irene
  • Cacciaglia, Raffaele
  • González-de-Echávarri, Jose María
  • Gispert López, Juan Domingo
  • Vilaplana Besler, Verónica|||0000-0001-6924-9961
Biological networks have gained considerable attention within the Deep Learning community because of the promising framework of Graph Neural Networks (GNN), neural models that operate in complex networks. In the context of neuroimaging, GNNs have successfully been employed for functional MRI processing but their application to ROI-level structural MRI (sMRI) remains mostly unexplored. In this work we analyze the implementation of these geometric models with sMRI by building graphs of ROIs (ROI graphs) using tools from Graph Signal Processing literature and evaluate their performance in a downstream supervised task, age prediction. We first make a qualitative and quantitative comparison of the resulting networks obtained with common graph topology learning strategies. In a second stage, we train GNN-based models for brain age prediction. Since the order of every ROI graph is exactly the same and each vertex is an entity by itself (a ROI), we evaluate whether including ROI information during message-passing or global pooling operations is beneficial and compare the performance of GNNs against a Fully-Connected Neural Network baseline. The results show that ROI-level information is needed during the global pooling operation in order to achieve competitive results. However, no relevant improvement has been detected when it is incorporated during the message passing. These models achieve a MAE of 4.27 in hold-out test data, which is a performance very similar to the baseline, suggesting that the inductive bias included with the obtained graph connectivity is relevant and useful to reduce the dimensionality of the problem., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Bridging domains in melanoma diagnostics: predicting BRAF mutations and sentinel lymph node positivity with attention-based models in histological images

  • Hernández Pérez, Carlos|||0000-0002-0054-8573
  • Jiménez Martín, Lauren|||0009-0002-5824-7893
  • Vilaplana Besler, Verónica|||0000-0001-6924-9961
Whole Slide Images (WSIs) have significantly advanced the field of pathology by providing highly detailed views of tissue samples. Integrating Deep Learning (DL) into this area of research, particularly through transformer-based foundational models, has marked a new era in automated image analysis. These foundational models are adept at extracting features from WSIs, an essential step in their analysis process. The subsequent application of weakly supervised learning techniques combines these features to predict critical biomarkers, such as BRAF mutations and sentinel lymph node (SLN) biopsy positivity, which are vital in guiding patient treatment strategies. However, the limited availability of labelled datasets in pathology hinders the usefulness of DL models. Domain adaptation strategies adeptly overcome this hurdle, enabling model knowledge transfer between different tissue types, thus addressing data scarcity. Our study employs a form of domain adaptation by fine-tuning two DINOv2 models, one pre-trained on natural images and the other on WSI of colorectal cancer from the TCGA dataset, adapting them for melanoma analysis. We also incorporate a comparison with features extracted by a third DINOv1 model trained solely on WSIs of breast cancer. With this approach, we find some notable success in detecting BRAF mutations. Nonetheless, predicting SLN positivity presents a more intricate challenge, largely due to the indirect correlation between local histopathological features in WSIs of primary tumours and lymph node metastasis manifestation. This dual-faceted approach not only combats the issue of limited data but also showcases the potential for enhanced accuracy in the field of digital pathology., This research was supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the project 718/C/2019 with id 201923-30 and 201923-31 funded by Fundació la Marato de TV3. Also by the FI-AGAUR ( 2022 FI B 00634 ) grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya and the European Social Fund., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Self-supervised graph representations of WSIs

  • Pina Benages, Òscar|||0009-0003-9755-6629
  • Vilaplana Besler, Verónica|||0000-0001-6924-9961
In this manuscript we propose a framework for the analysis of whole slide images (WSI) on the cell entity space with self-supervised deep learning on graphs and explore its representation quality at different levels of application. It consists of a two step process in which the cell level analysis is performed locally, by clusters of nearby cells that can be seen as small regions of the image, in order to learn representations that capture the cell environment and distribution. In a second stage, a WSI graph is generated with these regions as nodes and the representations learned as initial node embeddings. The graph is leveraged for a downstream task, region of interest (ROI) detection addressed as a graph clustering. The representations outperform the evaluation baselines at both levels of application, which has been carried out predicting whether a cell, or region, is tumor or not based on its learned representations with a logistic regressor., This work has been supported by the Spanish Research Agency (AEI) under project PID2020- 116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Cell-DETR: Efficient cell detection and classification in WSIs with transformers

  • Pina Benages, Òscar|||0009-0003-9755-6629
  • Dorca, Eduard
  • Vilaplana Besler, Verónica|||0000-0001-6924-9961
Understanding cell interactions and subpopulation distribution is crucial for pathologists to support their diagnoses. This cell information is traditionally extracted from segmentation methods, which poses significant computational challenges on processing Whole Slide Images (WSIs) due to their giga-size nature. Nonetheless, the clinically relevant tasks are nuclei detection and classification rather than segmentation. In this manuscript, we undertake a comprehensive exploration of the applicability of detection transformers for cell detection and classification (Cell-DETR). Not only do we demonstrate the effectiveness of the method by achieving state-of-the-art performance on well-established benchmarks, but we also develop a pipeline to tackle these tasks on WSIs at scale to enable the development of downstream applications. We show its efficiency and feasibility by reporting a x3.4 faster inference time on a dataset featuring large WSIs. By addressing the challenges associated with large-scale cell detection, our work contributes valuable insights that paves the way for the development of scalable diagnosis pipelines based on cell-level information., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/AEI/10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Enhancing Ki-67 cell segmentation with Dual U-Net models: a step towards uncertainty-informed active learning

  • Anglada Rotger, David|||0009-0009-9792-5889
  • Sala Prat, Júlia
  • Marqués Acosta, Fernando|||0000-0001-8311-1168
  • Salembier Clairon, Philippe Jean|||0000-0001-8884-9604
  • Pardàs Feliu, Montse|||0000-0002-5861-6356
The diagnosis and prognosis of breast cancer relies on histopathology image analysis where markers such as Ki-67 are increasingly important. The diagnosis using this marker is based on quantification of proliferation, which implies counting of Ki-67 positive and negative tumoral cells, excluding stromal cells. A common problem for automatic quantification of these images derives from overlapping and clustering of cells. We propose in this paper an automatic segmentation and classification system that overcomes this problem using two Convolutional Neural Networks (Dual U-Net), whose results are combined with a watershed algorithm. Taking into account that a major issue for the development of reliable neural networks is the availability of labeled databases, we also introduce an approach for epistemic uncertainty estimation that can be used for active learning in instance segmentation applications. We use Monte Carlo Dropout within our networks to quantify the model's confidence across its predictions, offering insights into areas of high uncertainty. Our results show how the postprocessed uncertainty maps can be used to refine ground truth annotations and to generate new labeled data with reduced annotation effort. To initialize the labeling and further reduce this effort, we propose a tool for groundtruth generation which is based on candidate generation with maxtree. Candidates are filtered based on extracted features which can be adjusted for the specific image typology, thereby facilitating precise model training and evaluation., This reseach has been funded by the project PID2020-116907RB-100 [AIMING] and PID2020-117142GB-I00 [DeeLight] by MCIN/ AEI /10.13039/ 501100011033., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Color deconvolution applied to domain adaptation in HER2 histopathological images

  • Anglada Rotger, David|||0009-0009-9792-5889
  • Marqués Acosta, Fernando|||0000-0001-8311-1168
  • Pardàs Feliu, Montse|||0000-0002-5861-6356
Breast cancer early detection is crucial for improving patient outcomes. The Institut Catala de la Salut (ICS) has launched‘ the DigiPatICS project to develop and implement artificial intelligence algorithms to assist with the diagnosis of cancer. In this paper, we propose a new approach for facing the color normalization problem in HER2-stained histopathological images of breast cancer tissue, posed as an style transfer problem. We combine the Color Deconvolution technique with the Pix2Pix GAN network to present a novel approach to correct the color variations between different HER2 stain brands. Our approach focuses on maintaining the HER2 score of the cells in the transformed images, which is crucial for the HER2 analysis. Results demonstrate that our final model outperforms the state-of-the-art image style transfer methods in maintaining the cell classes in the transformed images and is as effective as them in generating realistic images., This reseach has been partially funded by the project DigiPatICS, from ICS, and the project PID2020-116907RB-100 AIMING, funded by MCIN/ AEI /10.13039/ 501100011033., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Unsupervised domain adaptation for multi-stain cell detection in breast cancer with transformers

  • Pina Benages, Òscar|||0009-0003-9755-6629
  • Vilaplana Besler, Verónica|||0000-0001-6924-9961
The complexity of digital pathology image analysis arises from histopathological slide variability, including tissue specimen differences and stain variations. While publicly available datasets primarily focus on hematoxylin and eosin (H&E) staining, pathologists often require analysis across multiple stains for comprehensive diagnosis. Deep learning pipelines' implementation in clinical settings is hindered by poor cross-stain generalization, necessitating exhaustive annotations for each stain, which are time-consuming to obtain. In this work, we address these challenges by focusing on breast cancer analysis across four crucial stains: ER, PR, HER2, and Ki-67. Given the necessity of cell-level information for diagnosis, we concentrate on cell detection tasks with detection transformers. Leveraging unsupervised domain adaptation techniques, we bridge the gap between publicly available, annotated H&E datasets and unlabeled data in other stains. We demonstrate the superiority of adversarial feature learning over source-only and image-level generative methods. Our work contributes to improving digital pathology image analysis by enabling robust and efficient computer-aided diagnosis pipelines across multiple stains, thereby improving diagnostic accuracy in practical settings. The code can be found at https://github.com/oscar97pina/stain-celldetr. © 2024 IEEE., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FIAGAUR grant funded by Direccio General de Recerca ´ (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Cell-DETR: Efficient cell detection and classification in WSIs with transformers

  • Pina Benages, Òscar
  • Dorca, Eduard
  • Vilaplana Besler, Verónica
Understanding cell interactions and subpopulation distribution is crucial for pathologists to support their diagnoses. This cell information is traditionally extracted from segmentation methods, which poses significant computational challenges on processing Whole Slide Images (WSIs) due to their giga-size nature. Nonetheless, the clinically relevant tasks are nuclei detection and classification rather than segmentation. In this manuscript, we undertake a comprehensive exploration of the applicability of detection transformers for cell detection and classification (Cell-DETR). Not only do we demonstrate the effectiveness of the method by achieving state-of-the-art performance on well-established benchmarks, but we also develop a pipeline to tackle these tasks on WSIs at scale to enable the development of downstream applications. We show its efficiency and feasibility by reporting a x3.4 faster inference time on a dataset featuring large WSIs. By addressing the challenges associated with large-scale cell detection, our work contributes valuable insights that paves the way for the development of scalable diagnosis pipelines based on cell-level information., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/AEI/10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Self-supervised graph representations of WSIs

  • Pina Benages, Òscar
  • Vilaplana Besler, Verónica
In this manuscript we propose a framework for the analysis of whole slide images (WSI) on the cell entity space with self-supervised deep learning on graphs and explore its representation quality at different levels of application. It consists of a two step process in which the cell level analysis is performed locally, by clusters of nearby cells that can be seen as small regions of the image, in order to learn representations that capture the cell environment and distribution. In a second stage, a WSI graph is generated with these regions as nodes and the representations learned as initial node embeddings. The graph is leveraged for a downstream task, region of interest (ROI) detection addressed as a graph clustering. The representations outperform the evaluation baselines at both levels of application, which has been carried out predicting whether a cell, or region, is tumor or not based on its learned representations with a logistic regressor., This work has been supported by the Spanish Research Agency (AEI) under project PID2020- 116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Unsupervised domain adaptation for multi-stain cell detection in breast cancer with transformers

  • Pina Benages, Òscar
  • Vilaplana Besler, Verónica
The complexity of digital pathology image analysis arises from histopathological slide variability, including tissue specimen differences and stain variations. While publicly available datasets primarily focus on hematoxylin and eosin (H&E) staining, pathologists often require analysis across multiple stains for comprehensive diagnosis. Deep learning pipelines' implementation in clinical settings is hindered by poor cross-stain generalization, necessitating exhaustive annotations for each stain, which are time-consuming to obtain. In this work, we address these challenges by focusing on breast cancer analysis across four crucial stains: ER, PR, HER2, and Ki-67. Given the necessity of cell-level information for diagnosis, we concentrate on cell detection tasks with detection transformers. Leveraging unsupervised domain adaptation techniques, we bridge the gap between publicly available, annotated H&E datasets and unlabeled data in other stains. We demonstrate the superiority of adversarial feature learning over source-only and image-level generative methods. Our work contributes to improving digital pathology image analysis by enabling robust and efficient computer-aided diagnosis pipelines across multiple stains, thereby improving diagnostic accuracy in practical settings. The code can be found at https://github.com/oscar97pina/stain-celldetr. © 2024 IEEE., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FIAGAUR grant funded by Direccio General de Recerca ´ (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Structural networks for brain age prediction

  • Pina Benages, Òscar
  • Cumplido Mayoral, Irene
  • Cacciaglia, Raffaele
  • González-de-Echávarri, Jose María
  • Gispert López, Juan Domingo
  • Vilaplana Besler, Verónica
Biological networks have gained considerable attention within the Deep Learning community because of the promising framework of Graph Neural Networks (GNN), neural models that operate in complex networks. In the context of neuroimaging, GNNs have successfully been employed for functional MRI processing but their application to ROI-level structural MRI (sMRI) remains mostly unexplored. In this work we analyze the implementation of these geometric models with sMRI by building graphs of ROIs (ROI graphs) using tools from Graph Signal Processing literature and evaluate their performance in a downstream supervised task, age prediction. We first make a qualitative and quantitative comparison of the resulting networks obtained with common graph topology learning strategies. In a second stage, we train GNN-based models for brain age prediction. Since the order of every ROI graph is exactly the same and each vertex is an entity by itself (a ROI), we evaluate whether including ROI information during message-passing or global pooling operations is beneficial and compare the performance of GNNs against a Fully-Connected Neural Network baseline. The results show that ROI-level information is needed during the global pooling operation in order to achieve competitive results. However, no relevant improvement has been detected when it is incorporated during the message passing. These models achieve a MAE of 4.27 in hold-out test data, which is a performance very similar to the baseline, suggesting that the inductive bias included with the obtained graph connectivity is relevant and useful to reduce the dimensionality of the problem., This work has been supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the FI-AGAUR grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Color deconvolution applied to domain adaptation in HER2 histopathological images

  • Anglada Rotger, David
  • Marqués Acosta, Fernando
  • Pardàs Feliu, Montse
Breast cancer early detection is crucial for improving patient outcomes. The Institut Catala de la Salut (ICS) has launched‘ the DigiPatICS project to develop and implement artificial intelligence algorithms to assist with the diagnosis of cancer. In this paper, we propose a new approach for facing the color normalization problem in HER2-stained histopathological images of breast cancer tissue, posed as an style transfer problem. We combine the Color Deconvolution technique with the Pix2Pix GAN network to present a novel approach to correct the color variations between different HER2 stain brands. Our approach focuses on maintaining the HER2 score of the cells in the transformed images, which is crucial for the HER2 analysis. Results demonstrate that our final model outperforms the state-of-the-art image style transfer methods in maintaining the cell classes in the transformed images and is as effective as them in generating realistic images., This reseach has been partially funded by the project DigiPatICS, from ICS, and the project PID2020-116907RB-100 AIMING, funded by MCIN/ AEI /10.13039/ 501100011033., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Enhancing Ki-67 cell segmentation with Dual U-Net models: a step towards uncertainty-informed active learning

  • Anglada Rotger, David
  • Sala Prat, Júlia
  • Marqués Acosta, Fernando
  • Salembier Clairon, Philippe Jean
  • Pardàs Feliu, Montse
The diagnosis and prognosis of breast cancer relies on histopathology image analysis where markers such as Ki-67 are increasingly important. The diagnosis using this marker is based on quantification of proliferation, which implies counting of Ki-67 positive and negative tumoral cells, excluding stromal cells. A common problem for automatic quantification of these images derives from overlapping and clustering of cells. We propose in this paper an automatic segmentation and classification system that overcomes this problem using two Convolutional Neural Networks (Dual U-Net), whose results are combined with a watershed algorithm. Taking into account that a major issue for the development of reliable neural networks is the availability of labeled databases, we also introduce an approach for epistemic uncertainty estimation that can be used for active learning in instance segmentation applications. We use Monte Carlo Dropout within our networks to quantify the model's confidence across its predictions, offering insights into areas of high uncertainty. Our results show how the postprocessed uncertainty maps can be used to refine ground truth annotations and to generate new labeled data with reduced annotation effort. To initialize the labeling and further reduce this effort, we propose a tool for groundtruth generation which is based on candidate generation with maxtree. Candidates are filtered based on extracted features which can be adjusted for the specific image typology, thereby facilitating precise model training and evaluation., This reseach has been funded by the project PID2020-116907RB-100 [AIMING] and PID2020-117142GB-I00 [DeeLight] by MCIN/ AEI /10.13039/ 501100011033., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Bridging domains in melanoma diagnostics: predicting BRAF mutations and sentinel lymph node positivity with attention-based models in histological images

  • Hernández Pérez, Carlos
  • Jiménez Martín, Lauren
  • Vilaplana Besler, Verónica
Whole Slide Images (WSIs) have significantly advanced the field of pathology by providing highly detailed views of tissue samples. Integrating Deep Learning (DL) into this area of research, particularly through transformer-based foundational models, has marked a new era in automated image analysis. These foundational models are adept at extracting features from WSIs, an essential step in their analysis process. The subsequent application of weakly supervised learning techniques combines these features to predict critical biomarkers, such as BRAF mutations and sentinel lymph node (SLN) biopsy positivity, which are vital in guiding patient treatment strategies. However, the limited availability of labelled datasets in pathology hinders the usefulness of DL models. Domain adaptation strategies adeptly overcome this hurdle, enabling model knowledge transfer between different tissue types, thus addressing data scarcity. Our study employs a form of domain adaptation by fine-tuning two DINOv2 models, one pre-trained on natural images and the other on WSI of colorectal cancer from the TCGA dataset, adapting them for melanoma analysis. We also incorporate a comparison with features extracted by a third DINOv1 model trained solely on WSIs of breast cancer. With this approach, we find some notable success in detecting BRAF mutations. Nonetheless, predicting SLN positivity presents a more intricate challenge, largely due to the indirect correlation between local histopathological features in WSIs of primary tumours and lymph node metastasis manifestation. This dual-faceted approach not only combats the issue of limited data but also showcases the potential for enhanced accuracy in the field of digital pathology., This research was supported by the Spanish Research Agency (AEI) under project PID2020-116907RB-I00 of the call MCIN/ AEI /10.13039/501100011033 and the project 718/C/2019 with id 201923-30 and 201923-31 funded by Fundació la Marato de TV3. Also by the FI-AGAUR ( 2022 FI B 00634 ) grant funded by Direcció General de Recerca (DGR) of Departament de Recerca i Universitats (REU) of the Generalitat de Catalunya and the European Social Fund., Peer Reviewed, Postprint (author's final draft)