Showing papers by "Miguel Ángel González Ballester published in 2021"

Balanced-MixUp for highly imbalanced medical image classification

Abstract: Highly imbalanced datasets are ubiquitous in medical image classification problems. In such problems, it is often the case that rare classes associated to less prevalent diseases are severely under-represented in labeled databases, typically resulting in poor performance of machine learning algorithms due to overfitting in the learning process. In this paper, we propose a novel mechanism for sampling training data based on the popular MixUp regularization technique, which we refer to as Balanced-MixUp. In short, Balanced-MixUp simultaneously performs regular (i.e., instance-based) and balanced (i.e., class-based) sampling of the training data. The resulting two sets of samples are then mixed-up to create a more balanced training distribution from which a neural network can effectively learn without incurring in heavily under-fitting the minority classes. We experiment with a highly imbalanced dataset of retinal images (55K samples, 5 classes) and a long-tail dataset of gastro-intestinal video frames (10K images, 23 classes), using two CNNs of varying representation capabilities. Experimental results demonstrate that applying Balanced-MixUp outperforms other conventional sampling schemes and loss functions specifically designed to deal with imbalanced data. Code is released at https://github.com/agaldran/balanced_mixup

Multiscale Regulation of the Intervertebral Disc: Achievements in Experimental, In Silico, and Regenerative Research.

Abstract: Intervertebral disc (IVD) degeneration is a major risk factor of low back pain. It is defined by a progressive loss of the IVD structure and functionality, leading to severe impairments with restricted treatment options due to the highly demanding mechanical exposure of the IVD. Degenerative changes in the IVD usually increase with age but at an accelerated rate in some individuals. To understand the initiation and progression of this disease, it is crucial to identify key top-down and bottom-up regulations’ processes, across the cell, tissue, and organ levels, in health and disease. Owing to unremitting investigation of experimental research, the comprehension of detailed cell signaling pathways and their effect on matrix turnover significantly rose. Likewise, in silico research substantially contributed to a holistic understanding of spatiotemporal effects and complex, multifactorial interactions within the IVD. Together with important achievements in the research of biomaterials, manifold promising approaches for regenerative treatment options were presented over the last years. This review provides an integrative analysis of the current knowledge about (1) the multiscale function and regulation of the IVD in health and disease, (2) the possible regenerative strategies, and (3) the in silico models that shall eventually support the development of advanced therapies.

Double Encoder-Decoder Networks for Gastrointestinal Polyp Segmentation.

Abstract: Polyps represent an early sign of the development of Colorectal Cancer. The standard procedure for their detection consists of colonoscopic examination of the gastrointestinal tract. However, the wide range of polyp shapes and visual appearances, as well as the reduced quality of this image modality, turn their automatic identification and segmentation with computational tools into a challenging computer vision task. In this work, we present a new strategy for the delineation of gastrointestinal polyps from endoscopic images based on a direct extension of common encoder-decoder networks for semantic segmentation. In our approach, two pretrained encoder-decoder networks are sequentially stacked: the second network takes as input the concatenation of the original frame and the initial prediction generated by the first network, which acts as an attention mechanism enabling the second network to focus on interesting areas within the image, thereby improving the quality of its predictions. Quantitative evaluation carried out on several polyp segmentation databases shows that double encoder-decoder networks clearly outperform their single encoder-decoder counterparts in all cases. In addition, our best double encoder-decoder combination attains excellent segmentation accuracy and reaches state-of-the-art performance results in all the considered datasets, with a remarkable boost of accuracy on images extracted from datasets not used for training.

Double Encoder-Decoder Networks for Gastrointestinal Polyp Segmentation

Abstract: Polyps represent an early sign of the development of Colorectal Cancer. The standard procedure for their detection consists of colonoscopic examination of the gastrointestinal tract. However, the wide range of polyp shapes and visual appearances, as well as the reduced quality of this image modality, turn their automatic identification and segmentation with computational tools into a challenging computer vision task. In this work, we present a new strategy for the delineation of gastrointestinal polyps from endoscopic images based on a direct extension of common encoder-decoder networks for semantic segmentation. In our approach, two pretrained encoder-decoder networks are sequentially stacked: the second network takes as input the concatenation of the original frame and the initial prediction generated by the first network, which acts as an attention mechanism enabling the second network to focus on interesting areas within the image, thereby improving the quality of its predictions. Quantitative evaluation carried out on several polyp segmentation databases shows that double encoder-decoder networks clearly outperform their single encoder-decoder counterparts in all cases. In addition, our best double encoder-decoder combination attains excellent segmentation accuracy and reaches state-of-the-art performance results in all the considered datasets, with a remarkable boost of accuracy on images extracted from datasets not used for training.

Exercise Capacity in Young Adults Born Small for Gestational Age.

Abstract: Importance Being born small for gestational age (SGA), approximately 10% of all births, is associated with increased risk of cardiovascular mortality in adulthood, but mechanistic pathways are unclear. Cardiac remodeling and dysfunction occur in fetuses SGA and children born SGA, but it is uncertain whether and how these changes persist into adulthood. Objective To evaluate baseline cardiac function and structure and exercise capacity in young adults born SGA. Design, setting, and participants This cohort study conducted from January 2015 to January 2018 assessed a perinatal cohort born at a tertiary university hospital in Spain between 1975 and 1995. Participants included 158 randomly selected young adults aged 20 to 40 years born SGA (birth weight below the 10th centile) or with intrauterine growth within standard reference ranges (controls). Participants provided their medical history, filled out questionnaires regarding smoking and physical activity habits, and underwent incremental cardiopulmonary exercise stress testing, cardiac magnetic resonance imaging, and a physical examination, with blood pressure, glucose level, and lipid profile data collected. Exposure Being born SGA. Main outcomes and measures Cardiac structure and function assessed by cardiac magnetic resonance imaging, including biventricular end-diastolic shape analysis. Exercise capacity assessed by incremental exercise stress testing. Results This cohort study included 81 adults born SGA (median age at study, 34.4 years [IQR, 30.8-36.7 years]; 43 women [53%]) and 77 control participants (median age at study, 33.7 years [interquartile range (IQR), 31.0-37.1 years]; 33 women [43%]). All participants were of White race/ethnicity and underwent imaging, whereas 127 participants (80% of the cohort; 66 control participants and 61 adults born SGA) completed the exercise test. Cardiac shape analysis showed minor changes at rest in right ventricular geometry (DeLong test z, 2.2098; P = .02) with preserved cardiac function in individuals born SGA. However, compared with controls, adults born SGA had lower exercise capacity, with decreased maximal workload (mean [SD], 180 [62] W vs 214 [60] W; P = .006) and oxygen consumption (median, 26.0 mL/min/kg [IQR, 21.5-33.5 mL/min/kg vs 29.5 mL/min/kg [IQR, 24.0-36.0 mL/min/kg]; P = .02). Exercise capacity was significantly correlated with left ventricular mass (ρ = 0.7934; P Conclusions and relevance This cohort of young adults born SGA had markedly reduced exercise capacity. These results support further research to clarify the causes of impaired exercise capacity and the potential association with increased cardiovascular mortality among adults born SGA.

Nonlinear interaction between APOE ε4 allele load and age in the hippocampal surface of cognitively intact individuals

Abstract: The e4 allele of the gene Apolipoprotein E is the major genetic risk factor for Alzheimer's Disease. APOE e4 has been associated with changes in brain structure in cognitively impaired and unimpaired subjects, including atrophy of the hippocampus, which is one of the brain structures that is early affected by AD. In this work we analyzed the impact of APOE e4 gene dose and its association with age, on hippocampal shape assessed with multivariate surface analysis, in a e4-enriched cohort of n = 479 cognitively healthy individuals. Furthermore, we sought to replicate our findings on an independent dataset of n = 969 individuals covering the entire AD spectrum. We segmented the hippocampus of the subjects with a multi-atlas-based approach, obtaining high-dimensional meshes that can be analyzed in a multivariate way. We analyzed the effects of different factors including APOE, sex, and age (in both cohorts) as well as clinical diagnosis on the local 3D hippocampal surface changes. We found specific regions on the hippocampal surface where the effect is modulated by significant APOE e4 linear and quadratic interactions with age. We compared between APOE and diagnosis effects from both cohorts, finding similarities between APOE e4 and AD effects on specific regions, and suggesting that age may modulate the effect of APOE e4 and AD in a similar way.

Identifying causative mechanisms linking early-life stress to psycho-cardio-metabolic multi-morbidity: The EarlyCause project

Abstract: Introduction Depression, cardiovascular diseases and diabetes are among the major non-communicable diseases, leading to significant disability and mortality worldwide. These diseases may share environmental and genetic determinants associated with multimorbid patterns. Stressful early-life events are among the primary factors associated with the development of mental and physical diseases. However, possible causative mechanisms linking early life stress (ELS) with psycho-cardio-metabolic (PCM) multi-morbidity are not well understood. This prevents a full understanding of causal pathways towards the shared risk of these diseases and the development of coordinated preventive and therapeutic interventions. Methods and analysis This paper describes the study protocol for EarlyCause, a large-scale and inter-disciplinary research project funded by the European Union’s Horizon 2020 research and innovation programme. The project takes advantage of human longitudinal birth cohort data, animal studies and cellular models to test the hypothesis of shared mechanisms and molecular pathways by which ELS shapes an individual’s physical and mental health in adulthood. The study will research in detail how ELS converts into biological signals embedded simultaneously or sequentially in the brain, the cardiovascular and metabolic systems. The research will mainly focus on four biological processes including possible alterations of the epigenome, neuroendocrine system, inflammatome, and the gut microbiome. Life-course models will integrate the role of modifying factors as sex, socioeconomics, and lifestyle with the goal to better identify groups at risk as well as inform promising strategies to reverse the possible mechanisms and/or reduce the impact of ELS on multi-morbidity development in high-risk individuals. These strategies will help better manage the impact of multi-morbidity on human health and the associated risk.

Memory-aware curriculum federated learning for breast cancer classification.

Abstract: For early breast cancer detection, regular screening with mammography imaging is recommended. Routinary examinations result in datasets with a predominant amount of negative samples. A potential solution to such class-imbalance is joining forces across multiple institutions. Developing a collaborative computer-aided diagnosis system is challenging in different ways. Patient privacy and regulations need to be carefully respected. Data across institutions may be acquired from different devices or imaging protocols, leading to heterogeneous non-IID data. Also, for learning-based methods, new optimization strategies working on distributed data are required. Recently, federated learning has emerged as an effective tool for collaborative learning. In this setting, local models perform computation on their private data to update the global model. The order and the frequency of local updates influence the final global model. Hence, the order in which samples are locally presented to the optimizers plays an important role. In this work, we define a memory-aware curriculum learning method for the federated setting. Our curriculum controls the order of the training samples paying special attention to those that are forgotten after the deployment of the global model. Our approach is combined with unsupervised domain adaptation to deal with domain shift while preserving data privacy. We evaluate our method with three clinical datasets from different vendors. Our results verify the effectiveness of federated adversarial learning for the multi-site breast cancer classification. Moreover, we show that our proposed memory-aware curriculum method is beneficial to further improve classification performance. Our code is publicly available at: this https URL.

A Hierarchical Multi-Task Approach to Gastrointestinal Image Analysis

Abstract: A large number of different lesions and pathologies can affect the human digestive system, resulting in life-threatening situations. Early detection plays a relevant role in the successful treatment and the increase of current survival rates to, e.g., colorectal cancer. The standard procedure enabling detection, endoscopic video analysis, generates large quantities of visual data that need to be carefully analyzed by an specialist. Due to the wide range of color, shape, and general visual appearance of pathologies, as well as highly varying image quality, such process is greatly dependent on the human operator experience and skill. In this work, we detail our solution to the task of multi-category classification of images from the gastrointestinal (GI) human tract within the 2020 Endotect Challenge. Our approach is based on a Convolutional Neural Network minimizing a hierarchical error function that takes into account not only the finding category, but also its location within the GI tract (lower/upper tract), and the type of finding (pathological finding/therapeutic intervention/anatomical landmark/mucosal views’ quality). We also describe in this paper our solution for the challenge task of polyp segmentation in colonoscopies, which was addressed with a pretrained double encoder-decoder network. Our internal cross-validation results show an average performance of 91.25 Mathews Correlation Coefficient (MCC) and 91.82 Micro-F1 score for the classification task, and a 92.30 F1 score for the polyp segmentation task. The organization provided feedback on the performance in a hidden test set for both tasks, which resulted in 85.61 MCC and 86.96 F1 score for classification, and 91.97 F1 score for polyp segmentation. At the time of writing no public ranking for this challenge had been released.

A Hierarchical Multi-Task Approach to Gastrointestinal Image Analysis

Abstract: A large number of different lesions and pathologies can affect the human digestive system, resulting in life-threatening situations. Early detection plays a relevant role in the successful treatment and the increase of current survival rates to, e.g., colorectal cancer. The standard procedure enabling detection, endoscopic video analysis, generates large quantities of visual data that need to be carefully analyzed by an specialist. Due to the wide range of color, shape, and general visual appearance of pathologies, as well as highly varying image quality, such process is greatly dependent on the human operator experience and skill. In this work, we detail our solution to the task of multi-category classification of images from the gastrointestinal (GI) human tract within the 2020 Endotect Challenge. Our approach is based on a Convolutional Neural Network minimizing a hierarchical error function that takes into account not only the finding category, but also its location within the GI tract (lower/upper tract), and the type of finding (pathological finding/therapeutic intervention/anatomical landmark/mucosal views' quality). We also describe in this paper our solution for the challenge task of polyp segmentation in colonoscopies, which was addressed with a pretrained double encoder-decoder network. Our internal cross-validation results show an average performance of 91.25 Mathews Correlation Coefficient (MCC) and 91.82 Micro-F1 score for the classification task, and a 92.30 F1 score for the polyp segmentation task. The organization provided feedback on the performance in a hidden test set for both tasks, which resulted in 85.61 MCC and 86.96 F1 score for classification, and 91.97 F1 score for polyp segmentation. At the time of writing no public ranking for this challenge had been released.

Right Ventricular Global and Regional Remodeling in American-Style Football Athletes: A Longitudinal 3D Echocardiographic Study

Abstract: Few data exist concerning the right ventricular (RV) physiological adaptation in American-style football (ASF) athletes. We aimed to analyze the RV global and regional responses among ASF-trained athletes. Fifty-nine (20 linemen and 39 non-linemen) ASF athletes were studied before and after inter-seasonal training. During this period, which lasted 7 months, all athletes were exposed to combined dynamic and static exercises. Cardiac longitudinal changes were examined using three-dimensional transthoracic echocardiography. A computational method based on geodesic distances was applied to volumetrically parcellate the RV into apical, outlet, and inlet regions. RV global and regional end-diastolic volumes increased significantly and similarly in linemen and non-linemen after training, with predominant changes in the apex and outlet regions. RV global and regional ejection fractions were preserved. Morphological changes were uniformly distributed among the four cardiac chambers, and it was independent of the field position. Assessment of RV end-diastolic global, inlet and apical volumes showed low intra-observer (3.3%, 4.1%, and 5.3%, respectively) and inter-observer (7%, 12.2%, and 9%, respectively) variability, whereas the outlet regional volumetric assessment was less reproducible. To conclude, ASF inter-seasonal training was associated with a proportionate biventricular enlargement, regardless of the field position. Regional RV analysis allowed us to quantify the amount of exercise-induced remodeling that was larger in the apical and outlet regions.

Convolutional Nets Versus Vision Transformers for Diabetic Foot Ulcer Classification.

Abstract: This paper compares well-established Convolutional Neural Networks (CNNs) to recently introduced Vision Transformers for the task of Diabetic Foot Ulcer Classification, in the context of the DFUC 2021 Grand-Challenge, in which this work attained the first position. Comprehensive experiments demonstrate that modern CNNs are still capable of outperforming Transformers in a low-data regime, likely owing to their ability for better exploiting spatial correlations. In addition, we empirically demonstrate that the recent Sharpness-Aware Minimization (SAM) optimization algorithm considerably improves the generalization capability of both kinds of models. Our results demonstrate that for this task, the combination of CNNs and the SAM optimization process results in superior performance than any other of the considered approaches.

Detection, growth quantification and malignancy prediction of pulmonary nodules using deep convolutional networks in follow-up CT scans.

Abstract: We address the problem of supporting radiologists in the longitudinal management of lung cancer. Therefore, we proposed a deep learning pipeline, composed of four stages that completely automatized from the detection of nodules to the classification of cancer, through the detection of growth in the nodules. In addition, the pipeline integrated a novel approach for nodule growth detection, which relied on a recent hierarchical probabilistic U-Net adapted to report uncertainty estimates. Also, a second novel method was introduced for lung cancer nodule classification, integrating into a two stream 3D-CNN network the estimated nodule malignancy probabilities derived from a pretrained nodule malignancy network. The pipeline was evaluated in a longitudinal cohort and reported comparable performances to the state of art.

Balanced-MixUp for Highly Imbalanced Medical Image Classification

Abstract: Highly imbalanced datasets are ubiquitous in medical image classification problems. In such problems, it is often the case that rare classes associated to less prevalent diseases are severely under-represented in labeled databases, typically resulting in poor performance of machine learning algorithms due to overfitting in the learning process. In this paper, we propose a novel mechanism for sampling training data based on the popular MixUp regularization technique, which we refer to as Balanced-MixUp. In short, Balanced-MixUp simultaneously performs regular (i.e., instance-based) and balanced (i.e., class-based) sampling of the training data. The resulting two sets of samples are then mixed-up to create a more balanced training distribution from which a neural network can effectively learn without incurring in heavily under-fitting the minority classes. We experiment with a highly imbalanced dataset of retinal images (55K samples, 5 classes) and a long-tail dataset of gastro-intestinal video frames (10K images, 23 classes), using two CNNs of varying representation capabilities. Experimental results demonstrate that applying Balanced-MixUp outperforms other conventional sampling schemes and loss functions specifically designed to deal with imbalanced data. Code is released at this https URL .

An Uncertainty-aware Hierarchical Probabilistic Network for Early Prediction, Quantification and Segmentation of Pulmonary Tumour Growth.

Abstract: Early detection and quantification of tumour growth would help clinicians to prescribe more accurate treatments and provide better surgical planning. However, the multifactorial and heterogeneous nature of lung tumour progression hampers identification of growth patterns. In this study, we present a novel method based on a deep hierarchical generative and probabilistic framework that, according to radiological guidelines, predicts tumour growth, quantifies its size and provides a semantic appearance of the future nodule. Unlike previous deterministic solutions, the generative characteristic of our approach also allows us to estimate the uncertainty in the predictions, especially important for complex and doubtful cases. Results of evaluating this method on an independent test set reported a tumour growth balanced accuracy of 74%, a tumour growth size MAE of 1.77 mm and a tumour segmentation Dice score of 78%. These surpassed the performances of equivalent deterministic and alternative generative solutions (i.e. probabilistic U-Net, Bayesian test dropout and Pix2Pix GAN) confirming the suitability of our approach.

Projection-Based Collision Detection Algorithm for Stereoelectroencephalography Electrode Risk Assessment and Re-Planning

Abstract: Surgical planning is crucial to Stereoelectroencephalography (SEEG), a minimally invasive procedure that requires clinicians to operate with no direct view of the brain. Decisions making involves different clinical specialties and requires analysis of multiple multimodal datasets. We present a DepthMap tool designed to localize, measure, and visualize surgical risk, and an AlternativeFinder tool, designed to search for alternative trajectories maintaining adherence to the initial trajectory with three different re-planning strategies: similar entry, similar target, or parallel trajectory. The two tools transform the 3D problem into the 2D domain using projective geometry and distance mapping. Both use the graphics processing unit (GPU) to create a 2D depth image used by DepthMap for measurement and visualization, and by AlternativeFinder to find alternative trajectories. Tools were tested with 12 SEEG cases using digital subtraction angiography. DepthMap was used to measure vessel distance. AlternativeFinder was then used to search for alternatives. Computation time and displacements of the entry and target points for each trajectory and adherence strategy were recorded. The DepthMap tool found vessels in 118 initial trajectories (out of 145). Ninety alternative trajectories were found to meet the required avascular constraints (average 820K alternatives evaluated per initial trajectory). The average computation time was 449 ms per initial trajectory (77 ms when alternatives were found). The tools presented helped clinicians examine and re-plan SEEG trajectories to avoid vascular risks using three adherence strategies. Quantitative measurement of the adherence shows the potential of this tool for clinical use.