Focused active learning for histopathological image classification.

Active Learning (AL) has the potential to solve a major problem of digital pathology: the efficient acquisition of labeled data for machine learning algorithms. However, existing AL methods often struggle in realistic settings with artifacts, ambiguities, and class imbalances, as commonly seen in the medical field. The lack of precise uncertainty estimations leads to the acquisition of images with a low informative value. To address these challenges, we propose Focused Active Learning (FocAL), which combines a Bayesian Neural Network with Out-of-Distribution detection to estimate different uncertainties for the acquisition function. Specifically, the weighted epistemic uncertainty accounts for the class imbalance, aleatoric uncertainty for ambiguous images, and an OoD score for artifacts. We perform extensive experiments to validate our method on MNIST and the real-world Panda dataset for the classification of prostate cancer. The results confirm that other AL methods are 'distracted' by ambiguities and artifacts which harm the performance. FocAL effectively focuses on the most informative images, avoiding ambiguities and artifacts during acquisition. For both experiments, FocAL outperforms existing AL approaches, reaching a Cohen's kappa of 0.764 with only 0.69% of the labeled Panda data.

Learning from crowds for automated histopathological image segmentation.

Automated semantic segmentation of histopathological images is an essential task in Computational Pathology (CPATH). The main limitation of Deep Learning (DL) to address this task is the scarcity of expert annotations. Crowdsourcing (CR) has emerged as a promising solution to reduce the individual (expert) annotation cost by distributing the labeling effort among a group of (non-expert) annotators. Extracting knowledge in this scenario is challenging, as it involves noisy annotations. Jointly learning the underlying (expert) segmentation and the annotators' expertise is currently a commonly used approach. Unfortunately, this approach is frequently carried out by learning a different neural network for each annotator, which scales poorly when the number of annotators grows. For this reason, this strategy cannot be easily applied to real-world CPATH segmentation. This paper proposes a new family of methods for CR segmentation of histopathological images. Our approach consists of two coupled networks: a segmentation network (for learning the expert segmentation) and an annotator network (for learning the annotators' expertise). We propose to estimate the annotators' behavior with only one network that receives the annotator ID as input, achieving scalability on the number of annotators. Our family is composed of three different models for the annotator network. Within this family, we propose a novel modeling of the annotator network in the CR segmentation literature, which considers the global features of the image. We validate our methods on a real-world dataset of Triple Negative Breast Cancer images labeled by several medical students. Our new CR modeling achieves a Dice coefficient of 0.7827, outperforming the well-known STAPLE (0.7039) and being competitive with the supervised method with expert labels (0.7723). The code is available at https://github.com/wizmik12/CRowd_Seg.

Bibliometric analysis of the global scientific production on machine learning applied to different cancer types.

Cancer disease is one of the main causes of death in the world, with million annual cases in the last decades. The need to find a cure has stimulated the search for efficient treatments and diagnostic procedures. One of the most promising tools that has emerged against cancer in recent years is machine learning (ML), which has raised a huge number of scientific papers published in a relatively short period of time. The present study analyzes global scientific production on ML applied to the most relevant cancer types through various bibliometric indicators. We find that over 30,000 studies have been published so far and observe that cancers with the highest number of published studies using ML (breast, lung, and colon cancer) are those with the highest incidence, being the USA and China the main scientific producers on the subject. Interestingly, the role of China and Japan in stomach cancer is correlated with the number of cases of this cancer type in Asia (78% of the worldwide cases). Knowing the countries and institutions that most study each area can be of great help for improving international collaborations between research groups and countries. Our analysis shows that medical and computer science journals lead the number of publications on the subject and could be useful for researchers in the field. Finally, keyword co-occurrence analysis suggests that ML-cancer research trends are focused not only on the use of ML as an effective diagnostic method, but also for the improvement of radiotherapy- and chemotherapy-based treatments.

Probabilistic Attention Based on Gaussian Processes for Deep Multiple Instance Learning.

Multiple instance learning (MIL) is a weakly supervised learning paradigm that is becoming increasingly popular because it requires less labeling effort than fully supervised methods. This is especially interesting for areas where the creation of large annotated datasets remains challenging, as in medicine. Although recent deep learning MIL approaches have obtained state-of-the-art results, they are fully deterministic and do not provide uncertainty estimations for the predictions. In this work, we introduce the attention Gaussian process (AGP) model, a novel probabilistic attention mechanism based on Gaussian processes (GPs) for deep MIL. AGP provides accurate bag-level predictions as well as instance-level explainability and can be trained end-to-end. Moreover, its probabilistic nature guarantees robustness to overfit on small datasets and uncertainty estimations for the predictions. The latter is especially important in medical applications, where decisions have a direct impact on the patient's health. The proposed model is validated experimentally as follows. First, its behavior is illustrated in two synthetic MIL experiments based on the well-known MNIST and CIFAR-10 datasets, respectively. Then, it is evaluated in three different real-world cancer detection experiments. AGP outperforms state-of-the-art MIL approaches, including deterministic deep learning ones. It shows a strong performance even on a small dataset with less than 100 labels and generalizes better than competing methods on an external test set. Moreover, we experimentally show that predictive uncertainty correlates with the risk of wrong predictions, and therefore it is a good indicator of reliability in practice. Our code is publicly available.

Scalable Variational Gaussian Processes for Crowdsourcing: Glitch Detection in LIGO.

In the last years, crowdsourcing is transforming the way classification training sets are obtained. Instead of relying on a single expert annotator, crowdsourcing shares the labelling effort among a large number of collaborators. For instance, this is being applied in the laureate laser interferometer gravitational waves observatory (LIGO), in order to detect glitches which might hinder the identification of true gravitational-waves. The crowdsourcing scenario poses new challenging difficulties, as it has to deal with different opinions from a heterogeneous group of annotators with unknown degrees of expertise. Probabilistic methods, such as Gaussian processes (GP), have proven successful in modeling this setting. However, GPs do not scale up well to large data sets, which hampers their broad adoption in real-world problems (in particular LIGO). This has led to the very recent introduction of deep learning based crowdsourcing methods, which have become the state-of-the-art for this type of problems. However, the accurate uncertainty quantification provided by GPs has been partially sacrificed. This is an important aspect for astrophysicists in LIGO, since a glitch detection system should provide very accurate probability distributions of its predictions. In this work, we first leverage a standard sparse GP approximation (SVGP) to develop a GP-based crowdsourcing method that factorizes into mini-batches. This makes it able to cope with previously-prohibitive data sets. This first approach, which we refer to as scalable variational Gaussian processes for crowdsourcing (SVGPCR), brings back GP-based methods to a state-of-the-art level, and excels at uncertainty quantification. SVGPCR is shown to outperform deep learning based methods and previous probabilistic ones when applied to the LIGO data. Its behavior and main properties are carefully analyzed in a controlled experiment based on the MNIST data set. Moreover, recent GP inference techniques are also adapted to crowdsourcing and evaluated experimentally.

Learning from crowds in digital pathology using scalable variational Gaussian processes.

The volume of labeled data is often the primary determinant of success in developing machine learning algorithms. This has increased interest in methods for leveraging crowds to scale data labeling efforts, and methods to learn from noisy crowd-sourced labels. The need to scale labeling is acute but particularly challenging in medical applications like pathology, due to the expertise required to generate quality labels and the limited availability of qualified experts. In this paper we investigate the application of Scalable Variational Gaussian Processes for Crowdsourcing (SVGPCR) in digital pathology. We compare SVGPCR with other crowdsourcing methods using a large multi-rater dataset where pathologists, pathology residents, and medical students annotated tissue regions breast cancer. Our study shows that SVGPCR is competitive with equivalent methods trained using gold-standard pathologist generated labels, and that SVGPCR meets or exceeds the performance of other crowdsourcing methods based on deep learning. We also show how SVGPCR can effectively learn the class-conditional reliabilities of individual annotators and demonstrate that Gaussian-process classifiers have comparable performance to similar deep learning methods. These results suggest that SVGPCR can meaningfully engage non-experts in pathology labeling tasks, and that the class-conditional reliabilities estimated by SVGPCR may assist in matching annotators to tasks where they perform well.

Deep Gaussian processes for biogeophysical parameter retrieval and model inversion.

Parameter retrieval and model inversion are key problems in remote sensing and Earth observation. Currently, different approximations exist: a direct, yet costly, inversion of radiative transfer models (RTMs); the statistical inversion with in situ data that often results in problems with extrapolation outside the study area; and the most widely adopted hybrid modeling by which statistical models, mostly nonlinear and non-parametric machine learning algorithms, are applied to invert RTM simulations. We will focus on the latter. Among the different existing algorithms, in the last decade kernel based methods, and Gaussian Processes (GPs) in particular, have provided useful and informative solutions to such RTM inversion problems. This is in large part due to the confidence intervals they provide, and their predictive accuracy. However, RTMs are very complex, highly nonlinear, and typically hierarchical models, so that very often a single (shallow) GP model cannot capture complex feature relations for inversion. This motivates the use of deeper hierarchical architectures, while still preserving the desirable properties of GPs. This paper introduces the use of deep Gaussian Processes (DGPs) for bio-geo-physical model inversion. Unlike shallow GP models, DGPs account for complicated (modular, hierarchical) processes, provide an efficient solution that scales well to big datasets, and improve prediction accuracy over their single layer counterpart. In the experimental section, we provide empirical evidence of performance for the estimation of surface temperature and dew point temperature from infrared sounding data, as well as for the prediction of chlorophyll content, inorganic suspended matter, and coloured dissolved matter from multispectral data acquired by the Sentinel-3 OLCI sensor. The presented methodology allows for more expressive forms of GPs in big remote sensing model inversion problems.