Federated learning for predicting histological response to neoadjuvant chemotherapy in triple-negative breast cancer.

Triple-negative breast cancer (TNBC) is a rare cancer, characterized by high metastatic potential and poor prognosis, and has limited treatment options. The current standard of care in nonmetastatic settings is neoadjuvant chemotherapy (NACT), but treatment efficacy varies substantially across patients. This heterogeneity is still poorly understood, partly due to the paucity of curated TNBC data. Here we investigate the use of machine learning (ML) leveraging whole-slide images and clinical information to predict, at diagnosis, the histological response to NACT for early TNBC women patients. To overcome the biases of small-scale studies while respecting data privacy, we conducted a multicentric TNBC study using federated learning, in which patient data remain secured behind hospitals' firewalls. We show that local ML models relying on whole-slide images can predict response to NACT but that collaborative training of ML models further improves performance, on par with the best current approaches in which ML models are trained using time-consuming expert annotations. Our ML model is interpretable and is sensitive to specific histological patterns. This proof of concept study, in which federated learning is applied to real-world datasets, paves the way for future biomarker discovery using unprecedentedly large datasets.

External control arm analysis: an evaluation of propensity score approaches, G-computation, and doubly debiased machine learning.

BACKGROUND: An external control arm is a cohort of control patients that are collected from data external to a single-arm trial. To provide an unbiased estimation of efficacy, the clinical profiles of patients from single and external arms should be aligned, typically using propensity score approaches. There are alternative approaches to infer efficacy based on comparisons between outcomes of single-arm patients and machine-learning predictions of control patient outcomes. These methods include G-computation and Doubly Debiased Machine Learning (DDML) and their evaluation for External Control Arms (ECA) analysis is insufficient. METHODS: We consider both numerical simulations and a trial replication procedure to evaluate the different statistical approaches: propensity score matching, Inverse Probability of Treatment Weighting (IPTW), G-computation, and DDML. The replication study relies on five type 2 diabetes randomized clinical trials granted by the Yale University Open Data Access (YODA) project. From the pool of five trials, observational experiments are artificially built by replacing a control arm from one trial by an arm originating from another trial and containing similarly-treated patients. RESULTS: Among the different statistical approaches, numerical simulations show that DDML has the smallest bias followed by G-computation. In terms of mean squared error, G-computation usually minimizes mean squared error. Compared to other methods, DDML has varying Mean Squared Error performances that improves with increasing sample sizes. For hypothesis testing, all methods control type I error and DDML is the most conservative. G-computation is the best method in terms of statistical power, and DDML has comparable power at [Formula: see text] but inferior ones for smaller sample sizes. The replication procedure also indicates that G-computation minimizes mean squared error whereas DDML has intermediate performances in between G-computation and propensity score approaches. The confidence intervals of G-computation are the narrowest whereas confidence intervals obtained with DDML are the widest for small sample sizes, which confirms its conservative nature. CONCLUSIONS: For external control arm analyses, methods based on outcome prediction models can reduce estimation error and increase statistical power compared to propensity score approaches.

An artificial intelligence model predicts the survival of solid tumour patients from imaging and clinical data.

BACKGROUND: The need for developing new biomarkers is increasing with the emergence of many targeted therapies. Artificial Intelligence (AI) algorithms have shown great promise in the medical imaging field to build predictive models. We developed a prognostic model for solid tumour patients using AI on multimodal data. PATIENTS AND METHODS: Our retrospective study included examinations of patients with seven different cancer types performed between 2003 and 2017 in 17 different hospitals. Radiologists annotated all metastases on baseline computed tomography (CT) and ultrasound (US) images. Imaging features were extracted using AI models and used along with the patients' and treatments' metadata. A Cox regression was fitted to predict prognosis. Performance was assessed on a left-out test set with 1000 bootstraps. RESULTS: The model was built on 436 patients and tested on 196 patients (mean age 59, IQR: 51-6, 411 men out of 616 patients). On the whole, 1147 US images were annotated with lesions delineation, and 632 thorax-abdomen-pelvis CTs (total of 301,975 slices) were fully annotated with a total of 9516 lesions. The developed model reaches an average concordance index of 0.71 (0.67-0.76, 95% CI). Using the median predicted risk as a threshold value, the model is able to significantly (log-rank test P value < 0.001) isolate high-risk patients from low-risk patients (respective median OS of 11 and 31 months) with a hazard ratio of 3.5 (2.4-5.2, 95% CI). CONCLUSION: AI was able to extract prognostic features from imaging data, and along with clinical data, allows an accurate stratification of patients' prognoses.

A deep learning method for predicting knee osteoarthritis radiographic progression from MRI.

BACKGROUND: The identification of patients with knee osteoarthritis (OA) likely to progress rapidly in terms of structure is critical to facilitate the development of disease-modifying drugs. METHODS: Using 9280 knee magnetic resonance (MR) images (3268 patients) from the Osteoarthritis Initiative (OAI) database , we implemented a deep learning method to predict, from MR images and clinical variables including body mass index (BMI), further cartilage degradation measured by joint space narrowing at 12 months. RESULTS: Using COR IW TSE images, our classification model achieved a ROC AUC score of 65%. On a similar task, trained radiologists obtained a ROC AUC score of 58.7% highlighting the difficulty of the classification task. Additional analyses conducted in parallel to predict pain grade evaluated by the WOMAC pain index achieved a ROC AUC score of 72%. Attention maps provided evidence for distinct specific areas as being relevant in those two predictive models, including the medial joint space for JSN progression and the intra-articular space for pain prediction. CONCLUSIONS: This feasibility study demonstrates the interest of deep learning applied to OA, with a potential to support even trained radiologists in the challenging task of identifying patients with a high-risk of disease progression.

Integrating deep learning CT-scan model, biological and clinical variables to predict severity of COVID-19 patients.

The SARS-COV-2 pandemic has put pressure on intensive care units, so that identifying predictors of disease severity is a priority. We collect 58 clinical and biological variables, and chest CT scan data, from 1003 coronavirus-infected patients from two French hospitals. We train a deep learning model based on CT scans to predict severity. We then construct the multimodal AI-severity score that includes 5 clinical and biological variables (age, sex, oxygenation, urea, platelet) in addition to the deep learning model. We show that neural network analysis of CT-scans brings unique prognosis information, although it is correlated with other markers of severity (oxygenation, LDH, and CRP) explaining the measurable but limited 0.03 increase of AUC obtained when adding CT-scan information to clinical variables. Here, we show that when comparing AI-severity with 11 existing severity scores, we find significantly improved prognosis performance; AI-severity can therefore rapidly become a reference scoring approach.

A deep learning model to predict RNA-Seq expression of tumours from whole slide images.

Deep learning methods for digital pathology analysis are an effective way to address multiple clinical questions, from diagnosis to prediction of treatment outcomes. These methods have also been used to predict gene mutations from pathology images, but no comprehensive evaluation of their potential for extracting molecular features from histology slides has yet been performed. We show that HE2RNA, a model based on the integration of multiple data modes, can be trained to systematically predict RNA-Seq profiles from whole-slide images alone, without expert annotation. Through its interpretable design, HE2RNA provides virtual spatialization of gene expression, as validated by CD3- and CD20-staining on an independent dataset. The transcriptomic representation learned by HE2RNA can also be transferred on other datasets, even of small size, to increase prediction performance for specific molecular phenotypes. We illustrate the use of this approach in clinical diagnosis purposes such as the identification of tumors with microsatellite instability.

Comprehensive Molecular and Pathologic Evaluation of Transitional Mesothelioma Assisted by Deep Learning Approach: A Multi-Institutional Study of the International Mesothelioma Panel from the MESOPATH Reference Center.

INTRODUCTION: Histologic subtypes of malignant pleural mesothelioma are a major prognostic indicator and decision denominator for all therapeutic strategies. In an ambiguous case, a rare transitional mesothelioma (TM) pattern may be diagnosed by pathologists either as epithelioid mesothelioma (EM), biphasic mesothelioma (BM), or sarcomatoid mesothelioma (SM). This study aimed to better characterize the TM subtype from a histological, immunohistochemical, and molecular standpoint. Deep learning of pathologic slides was applied to this cohort. METHODS: A random selection of 49 representative digitalized sections from surgical biopsies of TM was reviewed by 16 panelists. We evaluated BAP1 expression and CDKN2A (p16) homozygous deletion. We conducted a comprehensive, integrated, transcriptomic analysis. An unsupervised deep learning algorithm was trained to classify tumors. RESULTS: The 16 panelists recorded 784 diagnoses on the 49 cases. Even though a Kappa value of 0.42 is moderate, the presence of a TM component was diagnosed in 51%. In 49% of the histological evaluation, the reviewers classified the lesion as EM in 53%, SM in 33%, or BM in 14%. Median survival was 6.7 months. Loss of BAP1 observed in 44% was less frequent in TM than in EM and BM. p16 homozygous deletion was higher in TM (73%), followed by BM (63%) and SM (46%). RNA sequencing unsupervised clustering analysis revealed that TM grouped together and were closer to SM than to EM. Deep learning analysis achieved 94% accuracy for TM identification. CONCLUSION: These results revealed that the TM pattern should be classified as non-EM or at minimum as a subgroup of the SM type.

Predicting Survival After Hepatocellular Carcinoma Resection Using Deep Learning on Histological Slides.

BACKGROUND AND AIMS: Standardized and robust risk-stratification systems for patients with hepatocellular carcinoma (HCC) are required to improve therapeutic strategies and investigate the benefits of adjuvant systemic therapies after curative resection/ablation. APPROACH AND RESULTS: In this study, we used two deep-learning algorithms based on whole-slide digitized histological slides (whole-slide imaging; WSI) to build models for predicting survival of patients with HCC treated by surgical resection. Two independent series were investigated: a discovery set (Henri Mondor Hospital, n = 194) used to develop our algorithms and an independent validation set (The Cancer Genome Atlas [TCGA], n = 328). WSIs were first divided into small squares ("tiles"), and features were extracted with a pretrained convolutional neural network (preprocessing step). The first deep-learning-based algorithm ("SCHMOWDER") uses an attention mechanism on tumoral areas annotated by a pathologist whereas the second ("CHOWDER") does not require human expertise. In the discovery set, c-indices for survival prediction of SCHMOWDER and CHOWDER reached 0.78 and 0.75, respectively. Both models outperformed a composite score incorporating all baseline variables associated with survival. Prognostic value of the models was further validated in the TCGA data set, and, as observed in the discovery series, both models had a higher discriminatory power than a score combining all baseline variables associated with survival. Pathological review showed that the tumoral areas most predictive of poor survival were characterized by vascular spaces, the macrotrabecular architectural pattern, and a lack of immune infiltration. CONCLUSIONS: This study shows that artificial intelligence can help refine the prediction of HCC prognosis. It highlights the importance of pathologist/machine interactions for the construction of deep-learning algorithms that benefit from expert knowledge and allow a biological understanding of their output.

Deep learning-based classification of mesothelioma improves prediction of patient outcome.

Malignant mesothelioma (MM) is an aggressive cancer primarily diagnosed on the basis of histological criteria1. The 2015 World Health Organization classification subdivides mesothelioma tumors into three histological types: epithelioid, biphasic and sarcomatoid MM. MM is a highly complex and heterogeneous disease, rendering its diagnosis and histological typing difficult and leading to suboptimal patient care and decisions regarding treatment modalities2. Here we have developed a new approach-based on deep convolutional neural networks-called MesoNet to accurately predict the overall survival of mesothelioma patients from whole-slide digitized images, without any pathologist-provided locally annotated regions. We validated MesoNet on both an internal validation cohort from the French MESOBANK and an independent cohort from The Cancer Genome Atlas (TCGA). We also demonstrated that the model was more accurate in predicting patient survival than using current pathology practices. Furthermore, unlike classical black-box deep learning methods, MesoNet identified regions contributing to patient outcome prediction. Strikingly, we found that these regions are mainly located in the stroma and are histological features associated with inflammation, cellular diversity and vacuolization. These findings suggest that deep learning models can identify new features predictive of patient survival and potentially lead to new biomarker discoveries.

Comparative performances of machine learning methods for classifying Crohn Disease patients using genome-wide genotyping data.

Crohn Disease (CD) is a complex genetic disorder for which more than 140 genes have been identified using genome wide association studies (GWAS). However, the genetic architecture of the trait remains largely unknown. The recent development of machine learning (ML) approaches incited us to apply them to classify healthy and diseased people according to their genomic information. The Immunochip dataset containing 18,227 CD patients and 34,050 healthy controls enrolled and genotyped by the international Inflammatory Bowel Disease genetic consortium (IIBDGC) has been re-analyzed using a set of ML methods: penalized logistic regression (LR), gradient boosted trees (GBT) and artificial neural networks (NN). The main score used to compare the methods was the Area Under the ROC Curve (AUC) statistics. The impact of quality control (QC), imputing and coding methods on LR results showed that QC methods and imputation of missing genotypes may artificially increase the scores. At the opposite, neither the patient/control ratio nor marker preselection or coding strategies significantly affected the results. LR methods, including Lasso, Ridge and ElasticNet provided similar results with a maximum AUC of 0.80. GBT methods like XGBoost, LightGBM and CatBoost, together with dense NN with one or more hidden layers, provided similar AUC values, suggesting limited epistatic effects in the genetic architecture of the trait. ML methods detected near all the genetic variants previously identified by GWAS among the best predictors plus additional predictors with lower effects. The robustness and complementarity of the different methods are also studied. Compared to LR, non-linear models such as GBT or NN may provide robust complementary approaches to identify and classify genetic markers.

Colonic MicroRNA Profiles, Identified by a Deep Learning Algorithm, That Predict Responses to Therapy of Patients With Acute Severe Ulcerative Colitis.

BACKGROUND & AIMS: Acute severe ulcerative colitis (ASUC) is a life-threatening condition managed with intravenous steroids followed by infliximab, cyclosporine, or colectomy (for patients with steroid resistance). There are no biomarkers to identify patients most likely to respond to therapy; ineffective medical treatment can delay colectomy and increase morbidity and mortality. We aimed to identify biomarkers of response to medical therapy for patients with ASUC. METHODS: We performed a retrospective analysis of 47 patients with ASUC, well characterized for their responses to steroids, cyclosporine, or infliximab, therapy at 2 centers in France. Fixed colonic biopsies, collected before or within the first 3 days of treatment, were used for microarray analysis of microRNA expression profiles. Deep neural network-based classifiers were used to derive candidate biomarkers for discriminating responders from non-responders to each treatment and to predict which patients would require colectomy. Levels of identified microRNAs were then measured by quantitative PCR analysis in a validation cohort of 29 independent patients-the effectiveness of the classification algorithm was tested on this cohort. RESULTS: A deep neural network-based classifier identified 9 microRNAs plus 5 clinical factors, routinely recorded at time of hospital admission, that associated with responses of patients to treatment. This panel discriminated responders to steroids from non-responders with 93% accuracy (area under the curve, 0.91). We identified 3 algorithms, based on microRNA levels, that identified responders to infliximab vs non-responders (84% accuracy, AUC = 0.82) and responders to cyclosporine vs non-responders (80% accuracy, AUC = 0.79). CONCLUSION: We developed an algorithm that identifies patients with ASUC who respond vs do not respond to first- and second-line treatments, based on microRNA expression profiles in colon tissues.

Multi-type Galton-Watson Processes with Affinity-Dependent Selection Applied to Antibody Affinity Maturation.

We analyze the interactions between division, mutation and selection in a simplified evolutionary model, assuming that the population observed can be classified into fitness levels. The construction of our mathematical framework is motivated by the modeling of antibody affinity maturation of B-cells in germinal centers during an immune response. This is a key process in adaptive immunity leading to the production of high-affinity antibodies against a presented antigen. Our aim is to understand how the different biological parameters affect the system's functionality. We identify the existence of an optimal value of the selection rate, able to maximize the number of selected B-cells for a given generation.

A Deep Learning Architecture for Temporal Sleep Stage Classification Using Multivariate and Multimodal Time Series.

Sleep stage classification constitutes an important preliminary exam in the diagnosis of sleep disorders. It is traditionally performed by a sleep expert who assigns to each 30 s of the signal of a sleep stage, based on the visual inspection of signals such as electroencephalograms (EEGs), electrooculograms (EOGs), electrocardiograms, and electromyograms (EMGs). We introduce here the first deep learning approach for sleep stage classification that learns end-to-end without computing spectrograms or extracting handcrafted features, that exploits all multivariate and multimodal polysomnography (PSG) signals (EEG, EMG, and EOG), and that can exploit the temporal context of each 30-s window of data. For each modality, the first layer learns linear spatial filters that exploit the array of sensors to increase the signal-to-noise ratio, and the last layer feeds the learnt representation to a softmax classifier. Our model is compared to alternative automatic approaches based on convolutional networks or decisions trees. Results obtained on 61 publicly available PSG records with up to 20 EEG channels demonstrate that our network architecture yields the state-of-the-art performance. Our study reveals a number of insights on the spatiotemporal distribution of the signal of interest: a good tradeoff for optimal classification performance measured with balanced accuracy is to use 6 EEG with 2 EOG (left and right) and 3 EMG chin channels. Also exploiting 1 min of data before and after each data segment offers the strongest improvement when a limited number of channels are available. As sleep experts, our system exploits the multivariate and multimodal nature of PSG signals in order to deliver the state-of-the-art classification performance with a small computational cost.

Random walks on binary strings applied to the somatic hypermutation of B-cells.

Within the germinal center in follicles, B-cells proliferate, mutate and differentiate, while being submitted to a powerful selection: a micro-evolutionary mechanism at the heart of adaptive immunity. A new foreign pathogen is confronted to our immune system, the mutation mechanism that allows B-cells to adapt to it is called somatic hypermutation: a programmed process of mutation affecting B-cell receptors at extremely high rate. By considering random walks on graphs, we introduce and analyze a simplified mathematical model in order to understand this extremely efficient learning process. The structure of the graph reflects the choice of the mutation rule. We focus on the impact of this choice on typical time-scales of the graphs' exploration. We derive explicit formulas to evaluate the expected hitting time to cover a given Hamming distance on the graphs under consideration. This characterizes the efficiency of these processes in driving antibody affinity maturation. In a further step we present a biologically more involved model and discuss its numerical outputs within our mathematical framework. We provide as well limitations and possible extensions of our approach.

Integrative Network-based Analysis of Colonic Detoxification Gene Expression in Ulcerative Colitis According to Smoking Status.

BACKGROUNDS AND AIMS: The effect of cigarette smoking [CS] is ambivalent since smoking improves ulcerative colitis [UC] while it worsens Crohn's disease [CD]. Although this clinical relationship between inflammatory bowel disease [IBD] and tobacco is well established, only a few experimental works have investigated the effect of smoking on the colonic barrier homeostasis focusing on xenobiotic detoxification genes. METHODS: A comprehensive and integrated comparative analysis of the global xenobiotic detoxification capacity of the normal colonic mucosa of healthy smokers [n = 8] and non-smokers [n = 9] versus the non-affected colonic mucosa of UC patients [n = 19] was performed by quantitative real-time polymerase chain reaction [qRT PCR]. The detoxification gene expression profile was analysed in CD patients [n = 18], in smoking UC patients [n = 5], and in biopsies from non-smoking UC patients cultured or not with cigarette smoke extract [n = 8]. RESULTS: Of the 244 detoxification genes investigated, 65 were dysregulated in UC patients in comparison with healthy controls or CD patients. The expression of ≥ 45/65 genes was inversed by CS in biopsies of smoking UC patients in remission and in colonic explants of UC patients exposed to cigarette smoke extract. We devised a network-based data analysis approach for differentially assessing changes in genetic interactions, allowing identification of unexpected regulatory detoxification genes that may play a major role in the beneficial effect of smoking on UC. CONCLUSIONS: Non-inflamed colonic mucosa in UC is characterised by a specifically altered detoxification gene network, which is partially restored by tobacco. These mucosal signatures could be useful for developing new therapeutic strategies and biomarkers of drug response in UC.

Mathematical modeling of lymphocytes selection in the germinal center.

Lymphocyte selection is a fundamental process of adaptive immunity. In order to produce B-lymphocytes with a target antigenic profile, mutation selection and division occur in the germinal center, a specific part of lymph nodes. We introduce in this article a simplified mathematical model of this phenomenon, taking into account the main mechanisms. This model is written as a non-linear, non-local, inhomogeneous second order partial differential equation, for which we develop a mathematical analysis. We assess, mathematically and numerically, in the case of piecewise-constant coefficients, the performance of the biological function by evaluating the duration of this production process as a function of several parameters such as the mutation rate or the selection profile, in various asymptotic regimes.

Network Modeling of Crohn's Disease Incidence.

BACKGROUND: Numerous genetic and environmental risk factors play a role in human complex genetic disorders (CGD). However, their complex interplay remains to be modelled and explained in terms of disease mechanisms. METHODS AND FINDINGS: Crohn's Disease (CD) was modeled as a modular network of patho-physiological functions, each summarizing multiple gene-gene and gene-environment interactions. The disease resulted from one or few specific combinations of module functional states. Network aging dynamics was able to reproduce age-specific CD incidence curves as well as their variations over the past century in Western countries. Within the model, we translated the odds ratios (OR) associated to at-risk alleles in terms of disease propensities of the functional modules. Finally, the model was successfully applied to other CGD including ulcerative colitis, ankylosing spondylitis, multiple sclerosis and schizophrenia. CONCLUSION: Modeling disease incidence may help to understand disease causative chains, to delineate the potential of personalized medicine, and to monitor epidemiological changes in CGD.

A local Echo State Property through the largest Lyapunov exponent.

Echo State Networks are efficient time-series predictors, which highly depend on the value of the spectral radius of the reservoir connectivity matrix. Based on recent results on the mean field theory of driven random recurrent neural networks, enabling the computation of the largest Lyapunov exponent of an ESN, we develop a cheap algorithm to establish a local and operational version of the Echo State Property.

Regular graphs maximize the variability of random neural networks.

In this work we study the dynamics of systems composed of numerous interacting elements interconnected through a random weighted directed graph, such as models of random neural networks. We develop an original theoretical approach based on a combination of a classical mean-field theory originally developed in the context of dynamical spin-glass models, and the heterogeneous mean-field theory developed to study epidemic propagation on graphs. Our main result is that, surprisingly, increasing the variance of the in-degree distribution does not result in a more variable dynamical behavior, but on the contrary that the most variable behaviors are obtained in the regular graph setting. We further study how the dynamical complexity of the attractors is influenced by the statistical properties of the in-degree distribution.

Relative entropy minimizing noisy non-linear neural network to approximate stochastic processes.

A method is provided for designing and training noise-driven recurrent neural networks as models of stochastic processes. The method unifies and generalizes two known separate modeling approaches, Echo State Networks (ESN) and Linear Inverse Modeling (LIM), under the common principle of relative entropy minimization. The power of the new method is demonstrated on a stochastic approximation of the El Niño phenomenon studied in climate research.