Compositionally Equivariant Representation Learning.

Deep learning models often need sufficient supervision (i.e., labelled data) in order to be trained effectively. By contrast, humans can swiftly learn to identify important anatomy in medical images like MRI and CT scans, with minimal guidance. This recognition capability easily generalises to new images from different medical facilities and to new tasks in different settings. This rapid and generalisable learning ability is largely due to the compositional structure of image patterns in the human brain, which are not well represented in current medical models. In this paper, we study the utilisation of compositionality in learning more interpretable and generalisable representations for medical image segmentation. Overall, we propose that the underlying generative factors that are used to generate the medical images satisfy compositional equivariance property, where each factor is compositional (e.g., corresponds to human anatomy) and also equivariant to the task. Hence, a good representation that approximates well the ground truth factor has to be compositionally equivariant. By modelling the compositional representations with learnable von-Mises-Fisher (vMF) kernels, we explore how different design and learning biases can be used to enforce the representations to be more compositionally equivariant under un-, weakly-, and semi-supervised settings. Extensive results show that our methods achieve the best performance over several strong baselines on the task of semi-supervised domain-generalised medical image segmentation. Code will be made publicly available upon acceptance at https://github.com/vios-s.

Understanding the performance and reliability of NLP tools: a comparison of four NLP tools predicting stroke phenotypes in radiology reports.

Background: Natural language processing (NLP) has the potential to automate the reading of radiology reports, but there is a need to demonstrate that NLP methods are adaptable and reliable for use in real-world clinical applications. Methods: We tested the F1 score, precision, and recall to compare NLP tools on a cohort from a study on delirium using images and radiology reports from NHS Fife and a population-based cohort (Generation Scotland) that spans multiple National Health Service health boards. We compared four off-the-shelf rule-based and neural NLP tools (namely, EdIE-R, ALARM+, ESPRESSO, and Sem-EHR) and reported on their performance for three cerebrovascular phenotypes, namely, ischaemic stroke, small vessel disease (SVD), and atrophy. Clinical experts from the EdIE-R team defined phenotypes using labelling techniques developed in the development of EdIE-R, in conjunction with an expert researcher who read underlying images. Results: EdIE-R obtained the highest F1 score in both cohorts for ischaemic stroke, ≥93%, followed by ALARM+, ≥87%. The F1 score of ESPRESSO was ≥74%, whilst that of Sem-EHR is ≥66%, although ESPRESSO had the highest precision in both cohorts, 90% and 98%. For F1 scores for SVD, EdIE-R scored ≥98% and ALARM+ ≥90%. ESPRESSO scored lowest with ≥77% and Sem-EHR ≥81%. In NHS Fife, F1 scores for atrophy by EdIE-R and ALARM+ were 99%, dropping in Generation Scotland to 96% for EdIE-R and 91% for ALARM+. Sem-EHR performed lowest for atrophy at 89% in NHS Fife and 73% in Generation Scotland. When comparing NLP tool output with brain image reads using F1 scores, ALARM+ scored 80%, outperforming EdIE-R at 66% in ischaemic stroke. For SVD, EdIE-R performed best, scoring 84%, with Sem-EHR 82%. For atrophy, EdIE-R and both ALARM+ versions were comparable at 80%. Conclusions: The four NLP tools show varying F1 (and precision/recall) scores across all three phenotypes, although more apparent for ischaemic stroke. If NLP tools are to be used in clinical settings, this cannot be performed "out of the box." It is essential to understand the context of their development to assess whether they are suitable for the task at hand or whether further training, re-training, or modification is required to adapt tools to the target task.

The role of noise in denoising models for anomaly detection in medical images.

Pathological brain lesions exhibit diverse appearance in brain images, in terms of intensity, texture, shape, size, and location. Comprehensive sets of data and annotations are difficult to acquire. Therefore, unsupervised anomaly detection approaches have been proposed using only normal data for training, with the aim of detecting outlier anomalous voxels at test time. Denoising methods, for instance classical denoising autoencoders (DAEs) and more recently emerging diffusion models, are a promising approach, however naive application of pixelwise noise leads to poor anomaly detection performance. We show that optimization of the spatial resolution and magnitude of the noise improves the performance of different model training regimes, with similar noise parameter adjustments giving good performance for both DAEs and diffusion models. Visual inspection of the reconstructions suggests that the training noise influences the trade-off between the extent of the detail that is reconstructed and the extent of erasure of anomalies, both of which contribute to better anomaly detection performance. We validate our findings on two real-world datasets (tumor detection in brain MRI and hemorrhage/ischemia/tumor detection in brain CT), showing good detection on diverse anomaly appearances. Overall, we find that a DAE trained with coarse noise is a fast and simple method that gives state-of-the-art accuracy. Diffusion models applied to anomaly detection are as yet in their infancy and provide a promising avenue for further research. Code for our DAE model and coarse noise is provided at: https://github.com/AntanasKascenas/DenoisingAE.

Acute stroke CDS: automatic retrieval of thrombolysis contraindications from unstructured clinical letters.

Introduction: Thrombolysis treatment for acute ischaemic stroke can lead to better outcomes if administered early enough. However, contraindications exist which put the patient at greater risk of a bleed (e.g. recent major surgery, anticoagulant medication). Therefore, clinicians must check a patient's past medical history before proceeding with treatment. In this work we present a machine learning approach for accurate automatic detection of this information in unstructured text documents such as discharge letters or referral letters, to support the clinician in making a decision about whether to administer thrombolysis. Methods: We consulted local and national guidelines for thrombolysis eligibility, identifying 86 entities which are relevant to the thrombolysis decision. A total of 8,067 documents from 2,912 patients were manually annotated with these entities by medical students and clinicians. Using this data, we trained and validated several transformer-based named entity recognition (NER) models, focusing on transformer models which have been pre-trained on a biomedical corpus as these have shown most promise in the biomedical NER literature. Results: Our best model was a PubMedBERT-based approach, which obtained a lenient micro/macro F1 score of 0.829/0.723. Ensembling 5 variants of this model gave a significant boost to precision, obtaining micro/macro F1 of 0.846/0.734 which approaches the human annotator performance of 0.847/0.839. We further propose numeric definitions for the concepts of name regularity (similarity of all spans which refer to an entity) and context regularity (similarity of all context surrounding mentions of an entity), using these to analyse the types of errors made by the system and finding that the name regularity of an entity is a stronger predictor of model performance than raw training set frequency. Discussion: Overall, this work shows the potential of machine learning to provide clinical decision support (CDS) for the time-critical decision of thrombolysis administration in ischaemic stroke by quickly surfacing relevant information, leading to prompt treatment and hence to better patient outcomes.

Survey: Leakage and Privacy at Inference Time.

Leakage of data from publicly available Machine Learning (ML) models is an area of growing significance since commercial and government applications of ML can draw on multiple sources of data, potentially including users' and clients' sensitive data. We provide a comprehensive survey of contemporary advances on several fronts, covering involuntary data leakage which is natural to ML models, potential malicious leakage which is caused by privacy attacks, and currently available defence mechanisms. We focus on inference-time leakage, as the most likely scenario for publicly available models. We first discuss what leakage is in the context of different data, tasks, and model architectures. We then propose a taxonomy across involuntary and malicious leakage, followed by description of currently available defences, assessment metrics, and applications. We conclude with outstanding challenges and open questions, outlining some promising directions for future research.

Small-molecule screen reveals pathways that regulate C4 secretion in stem cell-derived astrocytes.

In the brain, the complement system plays a crucial role in the immune response and in synaptic elimination during normal development and disease. Here, we sought to identify pathways that modulate the production of complement component 4 (C4), recently associated with an increased risk of schizophrenia. To design a disease-relevant assay, we first developed a rapid and robust 3D protocol capable of producing large numbers of astrocytes from pluripotent cells. Transcriptional profiling of these astrocytes confirmed the homogeneity of this population of dorsal fetal-like astrocytes. Using a novel ELISA-based small-molecule screen, we identified epigenetic regulators, as well as inhibitors of intracellular signaling pathways, able to modulate C4 secretion from astrocytes. We then built a connectivity map to predict and validate additional key regulatory pathways, including one involving c-Jun-kinase. This work provides a foundation for developing therapies for CNS diseases involving the complement cascade.

Amyotrophic Lateral Sclerosis-Associated Persistent Organic Pollutant cis-Chlordane Causes GABAA-Independent Toxicity to Motor Neurons, Providing Evidence toward an Environmental Component of Sporadic Amyotrophic Lateral Sclerosis.

Amyotrophic lateral sclerosis (ALS) is an incurable neurodegenerative disease characterized by the death of upper and lower motor neurons. While causative genes have been identified, 90% of ALS cases are not inherited and are hypothesized to result from the accumulation of genetic and environmental risk factors. While no specific causative environmental toxin has been identified, previous work has indicated that the presence of the organochlorine pesticide cis-chlordane in the blood is highly correlated with ALS incidence. Never before tested on the motor system, here, we show that cis-chlordane is especially toxic to motor neurons in vitro- and in vivo-independent of its known antagonism of the GABAA receptor. We find that human stem-cell-derived motor neurons are more sensitive to cis-chlordane than other cell types and their action potential dynamics are altered. Utilizing zebrafish larvae, we show that cis-chlordane induces motor neuron and neuromuscular junction degeneration and subsequent motor deficits in a touch-evoked escape response. Together, our work points to cis-chlordane as a potential sporadic ALS exacerbating environmental pollutant.

Anti-SOD1 Nanobodies That Stabilize Misfolded SOD1 Proteins Also Promote Neurite Outgrowth in Mutant SOD1 Human Neurons.

ALS-linked mutations induce aberrant conformations within the SOD1 protein that are thought to underlie the pathogenic mechanism of SOD1-mediated ALS. Although clinical trials are underway for gene silencing of SOD1, these approaches reduce both wild-type and mutated forms of SOD1. Here, we sought to develop anti-SOD1 nanobodies with selectivity for mutant and misfolded forms of human SOD1 over wild-type SOD1. Characterization of two anti-SOD1 nanobodies revealed that these biologics stabilize mutant SOD1 in vitro. Further, SOD1 expression levels were enhanced and the physiological subcellular localization of mutant SOD1 was restored upon co-expression of anti-SOD1 nanobodies in immortalized cells. In human motor neurons harboring the SOD1 A4V mutation, anti-SOD1 nanobody expression promoted neurite outgrowth, demonstrating a protective effect of anti-SOD1 nanobodies in otherwise unhealthy cells. In vitro assays revealed that an anti-SOD1 nanobody exhibited selectivity for human mutant SOD1 over endogenous murine SOD1, thus supporting the preclinical utility of anti-SOD1 nanobodies for testing in animal models of ALS. In sum, the anti-SOD1 nanobodies developed and presented herein represent viable biologics for further preclinical testing in human and mouse models of ALS.

Evaluating the Impact on Clinical Task Efficiency of a Natural Language Processing Algorithm for Searching Medical Documents: Prospective Crossover Study.

BACKGROUND: Information retrieval (IR) from the free text within electronic health records (EHRs) is time consuming and complex. We hypothesize that natural language processing (NLP)-enhanced search functionality for EHRs can make clinical workflows more efficient and reduce cognitive load for clinicians. OBJECTIVE: This study aimed to evaluate the efficacy of 3 levels of search functionality (no search, string search, and NLP-enhanced search) in supporting IR for clinical users from the free text of EHR documents in a simulated clinical environment. METHODS: A clinical environment was simulated by uploading 3 sets of patient notes into an EHR research software application and presenting these alongside 3 corresponding IR tasks. Tasks contained a mixture of multiple-choice and free-text questions. A prospective crossover study design was used, for which 3 groups of evaluators were recruited, which comprised doctors (n=19) and medical students (n=16). Evaluators performed the 3 tasks using each of the search functionalities in an order in accordance with their randomly assigned group. The speed and accuracy of task completion were measured and analyzed, and user perceptions of NLP-enhanced search were reviewed in a feedback survey. RESULTS: NLP-enhanced search facilitated more accurate task completion than both string search (5.14%; P=.02) and no search (5.13%; P=.08). NLP-enhanced search and string search facilitated similar task speeds, both showing an increase in speed compared to the no search function, by 11.5% (P=.008) and 16.0% (P=.007) respectively. Overall, 93% of evaluators agreed that NLP-enhanced search would make clinical workflows more efficient than string search, with qualitative feedback reporting that NLP-enhanced search reduced cognitive load. CONCLUSIONS: To the best of our knowledge, this study is the largest evaluation to date of different search functionalities for supporting target clinical users in realistic clinical workflows, with a 3-way prospective crossover study design. NLP-enhanced search improved both accuracy and speed of clinical EHR IR tasks compared to browsing clinical notes without search. NLP-enhanced search improved accuracy and reduced the number of searches required for clinical EHR IR tasks compared to direct search term matching.

Causal machine learning for healthcare and precision medicine.

Causal machine learning (CML) has experienced increasing popularity in healthcare. Beyond the inherent capabilities of adding domain knowledge into learning systems, CML provides a complete toolset for investigating how a system would react to an intervention (e.g. outcome given a treatment). Quantifying effects of interventions allows actionable decisions to be made while maintaining robustness in the presence of confounders. Here, we explore how causal inference can be incorporated into different aspects of clinical decision support systems by using recent advances in machine learning. Throughout this paper, we use Alzheimer's disease to create examples for illustrating how CML can be advantageous in clinical scenarios. Furthermore, we discuss important challenges present in healthcare applications such as processing high-dimensional and unstructured data, generalization to out-of-distribution samples and temporal relationships, that despite the great effort from the research community remain to be solved. Finally, we review lines of research within causal representation learning, causal discovery and causal reasoning which offer the potential towards addressing the aforementioned challenges.

Learning disentangled representations in the imaging domain.

Disentangled representation learning has been proposed as an approach to learning general representations even in the absence of, or with limited, supervision. A good general representation can be fine-tuned for new target tasks using modest amounts of data, or used directly in unseen domains achieving remarkable performance in the corresponding task. This alleviation of the data and annotation requirements offers tantalising prospects for applications in computer vision and healthcare. In this tutorial paper, we motivate the need for disentangled representations, revisit key concepts, and describe practical building blocks and criteria for learning such representations. We survey applications in medical imaging emphasising choices made in exemplar key works, and then discuss links to computer vision applications. We conclude by presenting limitations, challenges, and opportunities.

MOOD 2020: A Public Benchmark for Out-of-Distribution Detection and Localization on Medical Images.

Detecting Out-of-Distribution (OoD) data is one of the greatest challenges in safe and robust deployment of machine learning algorithms in medicine. When the algorithms encounter cases that deviate from the distribution of the training data, they often produce incorrect and over-confident predictions. OoD detection algorithms aim to catch erroneous predictions in advance by analysing the data distribution and detecting potential instances of failure. Moreover, flagging OoD cases may support human readers in identifying incidental findings. Due to the increased interest in OoD algorithms, benchmarks for different domains have recently been established. In the medical imaging domain, for which reliable predictions are often essential, an open benchmark has been missing. We introduce the Medical-Out-Of-Distribution-Analysis-Challenge (MOOD) as an open, fair, and unbiased benchmark for OoD methods in the medical imaging domain. The analysis of the submitted algorithms shows that performance has a strong positive correlation with the perceived difficulty, and that all algorithms show a high variance for different anomalies, making it yet hard to recommend them for clinical practice. We also see a strong correlation between challenge ranking and performance on a simple toy test set, indicating that this might be a valuable addition as a proxy dataset during anomaly detection algorithm development.

Resolving cell state in iPSC-derived human neural samples with multiplexed fluorescence imaging.

Human induced pluripotent stem cell-derived (iPSC) neural cultures offer clinically relevant models of human diseases, including Amyotrophic Lateral Sclerosis, Alzheimer's, and Autism Spectrum Disorder. In situ characterization of the spatial-temporal evolution of cell state in 3D culture and subsequent 2D dissociated culture models based on protein expression levels and localizations is essential to understanding neural cell differentiation, disease state phenotypes, and sample-to-sample variability. Here, we apply PRobe-based Imaging for Sequential Multiplexing (PRISM) to facilitate multiplexed imaging with facile, rapid exchange of imaging probes to analyze iPSC-derived cortical and motor neuron cultures that are relevant to psychiatric and neurodegenerative disease models, using over ten protein targets. Our approach permits analysis of cell differentiation, cell composition, and functional marker expression in complex stem-cell derived neural cultures. Furthermore, our approach is amenable to automation, offering in principle the ability to scale-up to dozens of protein targets and samples.

Free energy calculations of ALS-causing SOD1 mutants reveal common perturbations to stability and dynamics along the maturation pathway.

With over 150 heritable mutations identified as disease-causative, superoxide dismutase 1 (SOD1) has been a main target of amyotrophic lateral sclerosis (ALS) research and therapeutic efforts. However, recent evidence has suggested that neither loss of function nor protein aggregation is responsible for promoting neurotoxicity. Furthermore, there is no clear pattern to the nature or the location of these mutations that could suggest a molecular mechanism behind SOD1-linked ALS. Here, we utilize reliable and accurate computational techniques to predict the perturbations of 10 such mutations to the free energy changes of SOD1 as it matures from apo monomer to metallated dimer. We find that the free energy perturbations caused by these mutations strongly depend on maturational progress, indicating the need for state-specific therapeutic targeting. We also find that many mutations exhibit similar patterns of perturbation to native and non-native maturation, indicating strong thermodynamic coupling between the dynamics at various sites of maturation within SOD1. These results suggest the presence of an allosteric network in SOD1 which is vulnerable to disruption by these mutations. Analysis of these perturbations may contribute to uncovering a unifying molecular mechanism which explains SOD1-linked ALS and help to guide future therapeutic efforts.

A Multi-Organ Nucleus Segmentation Challenge.

Generalized nucleus segmentation techniques can contribute greatly to reducing the time to develop and validate visual biomarkers for new digital pathology datasets. We summarize the results of MoNuSeg 2018 Challenge whose objective was to develop generalizable nuclei segmentation techniques in digital pathology. The challenge was an official satellite event of the MICCAI 2018 conference in which 32 teams with more than 80 participants from geographically diverse institutes participated. Contestants were given a training set with 30 images from seven organs with annotations of 21,623 individual nuclei. A test dataset with 14 images taken from seven organs, including two organs that did not appear in the training set was released without annotations. Entries were evaluated based on average aggregated Jaccard index (AJI) on the test set to prioritize accurate instance segmentation as opposed to mere semantic segmentation. More than half the teams that completed the challenge outperformed a previous baseline. Among the trends observed that contributed to increased accuracy were the use of color normalization as well as heavy data augmentation. Additionally, fully convolutional networks inspired by variants of U-Net, FCN, and Mask-RCNN were popularly used, typically based on ResNet or VGG base architectures. Watershed segmentation on predicted semantic segmentation maps was a popular post-processing strategy. Several of the top techniques compared favorably to an individual human annotator and can be used with confidence for nuclear morphometrics.

Aggregated SOD1 causes selective death of cultured human motor neurons.

Most human neurodegenerative diseases share a phenotype of neuronal protein aggregation. In Amyotrophic Lateral Sclerosis (ALS), the abundant protein superoxide dismutase (SOD1) or the TAR-DNA binding protein TDP-43 can aggregate in motor neurons. Recently, numerous studies have highlighted the ability of aggregates to spread from neuron to neuron in a prion-like fashion. These studies have typically focused on the use of neuron-like cell lines or neurons that are not normally affected by the specific aggregated protein being studied. Here, we have investigated the uptake of pre-formed SOD1 aggregates by cultures containing pluripotent stem cell-derived human motor neurons. We found that all cells take up aggregates by a process resembling fluid-phase endocytosis, just as found in earlier studies. However, motor neurons, despite taking up smaller amounts of SOD1, were much more vulnerable to the accumulating aggregates. Thus, the propagation of disease pathology depends less on selective uptake than on selective response to intracellular aggregates. We further demonstrate that anti-SOD1 antibodies, being considered as ALS therapeutics, can act by blocking the uptake of SOD1, but also by blocking the toxic effects of intracellular SOD1. This work demonstrates the importance of using disease relevant cells even in studying phenomena such as aggregate propagation.

Material microenvironmental properties couple to induce distinct transcriptional programs in mammalian stem cells.

Variations in a multitude of material microenvironmental properties have been observed across tissues in vivo, and these have profound effects on cell phenotype. Phenomenological experiments have suggested that certain of these features of the physical microenvironment, such as stiffness, could sensitize cells to other features; meanwhile, mechanistic studies have detailed a number of biophysical mechanisms for this sensing. However, the broad molecular consequences of these potentially complex and nonlinear interactions bridging from biophysical sensing to phenotype have not been systematically characterized, limiting the overall understanding and rational deployment of these biophysical cues. Here, we explore these interactions by employing a 3D cell culture system that allows for the independent control of culture substrate stiffness, stress relaxation, and adhesion ligand density to systematically explore the transcriptional programs affected by distinct combinations of biophysical parameters using RNA-seq. In mouse mesenchymal stem cells and human cortical neuron progenitors, we find dramatic coupling among these substrate properties, and that the relative contribution of each property to changes in gene expression varies with cell type. Motivated by the bioinformatic analysis, the stiffness of hydrogels encapsulating mouse mesenchymal stem cells was found to regulate the secretion of a wide range of cytokines, and to accordingly influence hematopoietic stem cell differentiation in a Transwell coculture model. These results give insights into how biophysical features are integrated by cells across distinct tissues and offer strategies to synthetic biologists and bioengineers for designing responses to a cell's biophysical environment.

In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images.

Microscopy is a central method in life sciences. Many popular methods, such as antibody labeling, are used to add physical fluorescent labels to specific cellular constituents. However, these approaches have significant drawbacks, including inconsistency; limitations in the number of simultaneous labels because of spectral overlap; and necessary perturbations of the experiment, such as fixing the cells, to generate the measurement. Here, we show that a computational machine-learning approach, which we call "in silico labeling" (ISL), reliably predicts some fluorescent labels from transmitted-light images of unlabeled fixed or live biological samples. ISL predicts a range of labels, such as those for nuclei, cell type (e.g., neural), and cell state (e.g., cell death). Because prediction happens in silico, the method is consistent, is not limited by spectral overlap, and does not disturb the experiment. ISL generates biological measurements that would otherwise be problematic or impossible to acquire.

Using intracellular markers to identify a novel set of surface markers for live cell purification from a heterogeneous hIPSC culture.

Human embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) can provide sources for midbrain dopaminergic (mDA) neural progenitors (NPCs) for cell therapy to treat Parkinson's disease (PD) patients. However, the well-known line-to-cell line variability in the differentiation capacity of individual cell lines needs to be improved for the success of this therapy. To address this issue, we sought to identify mDA NPC specific cell surface markers for fluorescence activated cell sorting (FACS). Through RNA isolation after sorting for NPCs based on staining for cell-specific transcription factors followed by microarray, we identified two positive cell surface markers (CORIN and CD166) and one negative cell surface marker (CXCR4) for mDA NPC sorting. These three markers can enrich floor plate NPCs to 90% purity, and the sorted NPCs more efficiently differentiate to mature dopaminergic neurons compared to unsorted or CORIN+ alone mDA NPCs. This surface marker identification strategy can be used broadly to facilitate isolation of cell subtypes of interest from heterogeneous cultures.