Generalizing the Enhanced-Deep-Super-Resolution neural network to brain MR images: a retrospective study on the Cam-CAN dataset.

The Enhanced-Deep-Super-Resolution (EDSR) model is a state-of-art convolutional neural network suitable for improving image spatial resolution. It was previously trained with general-purpose pictures and then, in this work, tested on biomedical Magnetic Resonance (MR) images, comparing the network outcomes with traditional up-sampling techniques. We explored possible changes in the model response when different MR sequences were analyzed. T1w and T2w MR brain images of 70 human healthy subjects (F:M 40:30) from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) repository were down-sampled and then up-sampled using EDSR model and BiCubic (BC) interpolation. Several reference metrics were used to quantitatively assess the performance of up-sampling operations (RMSE, pSNR, SSIM and HFEN). Two-dimensional and three-dimensional reconstructions were evaluated. Different brain tissues were analyzed individually. The EDSR model was superior to BC interpolation on the selected metrics, both for two- and three- dimensional reconstructions. The reference metrics showed higher quality of EDSR over BC reconstructions for all the analyzed images, with a significant difference of all the criteria in T1w images and of the perception-based SSIM and HFEN in T2w images. The analysis per tissue highlights differences in EDSR performance related to the gray-level values, showing a relative lack of out-performance in reconstructing hyper-intense areas. The EDSR model, trained on general-purpose images, better reconstructs MR T1w and T2w images than BC, without any re-training or fine-tuning. These results highlight the excellent generalization ability of the network and lead to possible applications on other MR measurements.Significance Statement Super resolution applications in biomedical images may help in reducing acquisition scan time and concurrently improving the quality of the exam. Neural networks have been shown to work better than traditional up-sampling techniques, even though ad hoc training experiments need to be performed for specific kind of data. In this work, we used a model previously trained with general-purpose images and we directly applied to magnetic resonance human brain ones; we verified its ability of reconstructing new kind of images, comparing the results with traditional up-sampling techniques. Our analysis highlights the excellent generalization capabilities of the model over the images tested, without need of specific re-training, suggesting that such results might be reproduced on images from other acquisition systems.

Sulcal Morphometry Predicts Mild Cognitive Impairment Conversion to Alzheimer's Disease.

Background: Being able to differentiate mild cognitive impairment (MCI) patients who would eventually convert (MCIc) to Alzheimer's disease (AD) from those who would not (MCInc) is a key challenge for prognosis. Objective: This study aimed to investigate the ability of sulcal morphometry to predict MCI progression to AD, dedicating special attention to an accurate identification of sulci. Methods: Twenty-five AD patients, thirty-seven MCI and twenty-five healthy controls (HC) underwent a brain-MR protocol (1.5T scanner) including a high-resolution T1-weighted sequence. MCI patients underwent a neuropsychological assessment at baseline and were clinically re-evaluated after a mean of 2.3 years. At follow-up, 12 MCI were classified as MCInc and 25 as MCIc. Sulcal morphometry was investigated using the BrainVISA framework. Consistency of sulci across subjects was ensured by visual inspection and manual correction of the automatic labelling in each subject. Sulcal surface, depth, length, and width were retrieved from 106 sulci. Features were compared across groups and their classification accuracy in predicting MCI conversion was tested. Potential relationships between sulcal features and cognitive scores were explored using Spearman's correlation. Results: The width of sulci in the temporo-occipital region strongly differentiated between each pair of groups. Comparing MCIc and MCInc, the width of several sulci in the bilateral temporo-occipital and left frontal areas was significantly altered. Higher width of frontal sulci was associated with worse performances in short-term verbal memory and phonemic fluency. Conclusions: Sulcal morphometry emerged as a strong tool for differentiating HC, MCI, and AD, demonstrating its potential prognostic value for the MCI population.

Impact on the Transcriptome of Proton Beam Irradiation Targeted at Healthy Cardiac Tissue of Mice.

Proton beam therapy is considered a step forward with respect to electromagnetic radiation, thanks to the reduction in the dose delivered. Among unwanted effects to healthy tissue, cardiovascular complications are a known long-term radiotherapy complication. The transcriptional response of cardiac tissue from xenografted BALB/c nude mice obtained at 3 and 10 days after proton irradiation covering both the tumor region and the underlying healthy tissue was analyzed as a function of dose and time. Three doses were used: 2 Gy, 6 Gy, and 9 Gy. The intermediate dose had caused the greatest impact at 3 days after irradiation: at 2 Gy, 219 genes were differently expressed, many of them represented by zinc finger proteins; at 6 Gy, there were 1109, with a predominance of genes involved in energy metabolism and responses to stimuli; and at 9 Gy, there were 105, mainly represented by zinc finger proteins and molecules involved in the regulation of cardiac function. After 10 days, no significant effects were detected, suggesting that cellular repair mechanisms had defused the potential alterations in gene expression. The nonlinear dose-response curve indicates a need to update the models built on photons to improve accuracy in health risk prediction. Our data also suggest a possible role for zinc finger protein genes as markers of proton therapy efficacy.

Data Science for Health Image Alignment: A User-Friendly Open-Source ImageJ/Fiji Plugin for Aligning Multimodality/Immunohistochemistry/Immunofluorescence 2D Microscopy Images.

Most of the time, the deep analysis of a biological sample requires the acquisition of images at different time points, using different modalities and/or different stainings. This information gives morphological, functional, and physiological insights, but the acquired images must be aligned to be able to proceed with the co-localisation analysis. Practically speaking, according to Aristotle's principle, "The whole is greater than the sum of its parts", multi-modal image registration is a challenging task that involves fusing complementary signals. In the past few years, several methods for image registration have been described in the literature, but unfortunately, there is not one method that works for all applications. In addition, there is currently no user-friendly solution for aligning images that does not require any computer skills. In this work, DS4H Image Alignment (DS4H-IA), an open-source ImageJ/Fiji plugin for aligning multimodality, immunohistochemistry (IHC), and/or immunofluorescence (IF) 2D microscopy images, designed with the goal of being extremely easy to use, is described. All of the available solutions for aligning 2D microscopy images have also been revised. The DS4H-IA source code; standalone applications for MAC, Linux, and Windows; video tutorials; manual documentation; and sample datasets are publicly available.

Sediment core analysis using artificial intelligence.

Subsurface stratigraphic modeling is crucial for a variety of environmental, societal, and economic challenges. However, the need for specific sedimentological skills in sediment core analysis may constitute a limitation. Methods based on Machine Learning and Deep Learning can play a central role in automatizing this time-consuming procedure. In this work, using a robust dataset of high-resolution digital images from continuous sediment cores of Holocene age that reflect a wide spectrum of continental to shallow-marine depositional environments, we outline a novel deep-learning-based approach to perform automatic semantic segmentation directly on core images, leveraging the power of convolutional neural networks. To optimize the interpretation process and maximize scientific value, we use six sedimentary facies associations as target classes in lieu of ineffective classification methods based uniquely on lithology. We propose an automated model that can rapidly characterize sediment cores, allowing immediate guidance for stratigraphic correlation and subsurface reconstructions.

Modelling antimicrobial resistance transmission to guide personalized antimicrobial stewardship interventions and infection control policies in healthcare setting: a pilot study.

Infection control programs and antimicrobial stewardship have been proven effective in reducing the burden of diseases due to multidrug-resistant organisms, but quantifying the effect of each intervention is an open issue. For this aim, we propose a model to characterize the effect of interventions at single ward level. We adapted the Ross-Macdonald model to describe hospital cross-transmission dynamics of carbapenem resistant Klebsiella pneumoniae (CRKP), considering healthcare workers as the vectors transmitting susceptible and resistant pathogens among admitted patients. The model parameters were estimated from a literature review, further adjusted to reproduce observed clinical outcomes, and validated using real life data from a 2-year study in a university hospital. The model has been further explored through extensive sensitivity analysis, in order to assess the relevance of single interventions as well as their synergistic effects. Our model has been shown to be an effective tool to describe and predict the impact of interventions in reducing the prevalence of CRKP colonisation and infection, and can be extended to other specific hospital and pathological scenarios to produce tailored estimates of the most effective strategies.

Genetic Diversity and Antimicrobial Resistance of Extraintestinal E. coli Populations Pre- and Post-Antimicrobial Therapy on Broilers Affected by Colisepticemia.

The aim of the present study was to investigate the genetic diversity and antimicrobial resistance (AMR) of E. coli during enrofloxacin therapy in broilers affected by colisepticemia. Three unrelated farms with ongoing colibacillosis outbreaks were sampled at day 1 before treatment and at days 5, 10 and 24 post-treatment. A total of 179 E. coli isolates were collected from extraintestinal organs and submitted to serotyping, PFGE and the minimum inhibitory concentration (MIC) against enrofloxacin. PFGE clusters shifted from 3-6 at D1 to 10-16 at D5, D10 and D24, suggesting an increased population diversity after the treatment. The majority of strains belonged to NT or O78 and to ST117 or ST23. PFGE results were confirmed with SNP calling: no persistent isolates were identified. An increase in resistance to fluoroquinolones in E. coli isolates was observed along the treatment. Resistome analyses revealed qnrB19 and qnrS1 genes along with mutations in the gyrA, parC and parE genes. Interestingly, despite a fluoroquinolone selective pressure, qnr-carrying plasmids did not persist. On the contrary, two conjugative AMR plasmid clusters (AB233 and AA474) harboring AMR genes other than qnr were persistent since they were identified in both D1 and D10 genomes in two farms. Further studies should be performed in order to confirm plasmid persistence not associated (in vivo) to antimicrobial selective pressure.

Comparison of Source Attribution Methodologies for Human Campylobacteriosis.

Campylobacter spp. are the most common cause of bacterial gastrointestinal infection in humans both in Denmark and worldwide. Studies have found microbial subtyping to be a powerful tool for source attribution, but comparisons of different methodologies are limited. In this study, we compare three source attribution approaches (Machine Learning, Network Analysis, and Bayesian modeling) using three types of whole genome sequences (WGS) data inputs (cgMLST, 5-Mers and 7-Mers). We predicted and compared the sources of human campylobacteriosis cases in Denmark. Using 7mer as an input feature provided the best model performance. The network analysis algorithm had a CSC value of 78.99% and an F1-score value of 67%, while the machine-learning algorithm showed the highest accuracy (98%). The models attributed between 965 and all of the 1224 human cases to a source (network applying 5mer and machine learning applying 7mer, respectively). Chicken from Denmark was the primary source of human campylobacteriosis with an average percentage probability of attribution of 45.8% to 65.4%, representing Bayesian with 7mer and machine learning with cgMLST, respectively. Our results indicate that the different source attribution methodologies based on WGS have great potential for the surveillance and source tracking of Campylobacter. The results of such models may support decision makers to prioritize and target interventions.

Effectiveness of Radiomic ZOT Features in the Automated Discrimination of Oncocytoma from Clear Cell Renal Cancer.

BACKGROUND: Benign renal tumors, such as renal oncocytoma (RO), can be erroneously diagnosed as malignant renal cell carcinomas (RCC), because of their similar imaging features. Computer-aided systems leveraging radiomic features can be used to better discriminate benign renal tumors from the malignant ones. The purpose of this work was to build a machine learning model to distinguish RO from clear cell RCC (ccRCC). METHOD: We collected CT images of 77 patients, with 30 cases of RO (39%) and 47 cases of ccRCC (61%). Radiomic features were extracted both from the tumor volumes identified by the clinicians and from the tumor's zone of transition (ZOT). We used a genetic algorithm to perform feature selection, identifying the most descriptive set of features for the tumor classification. We built a decision tree classifier to distinguish between ROs and ccRCCs. We proposed two versions of the pipeline: in the first one, the feature selection was performed before the splitting of the data, while in the second one, the feature selection was performed after, i.e., on the training data only. We evaluated the efficiency of the two pipelines in cancer classification. RESULTS: The ZOT features were found to be the most predictive by the genetic algorithm. The pipeline with the feature selection performed on the whole dataset obtained an average ROC AUC score of 0.87 ± 0.09. The second pipeline, in which the feature selection was performed on the training data only, obtained an average ROC AUC score of 0.62 ± 0.17. CONCLUSIONS: The obtained results confirm the efficiency of ZOT radiomic features in capturing the renal tumor characteristics. We showed that there is a significant difference in the performances of the two proposed pipelines, highlighting how some already published radiomic analyses could be too optimistic about the real generalization capabilities of the models.

Adding pieces to the puzzle of differentiated-to-anaplastic thyroid cancer evolution: the oncogene E2F7.

Anaplastic Thyroid Cancer (ATC) is the most aggressive and de-differentiated subtype of thyroid cancer. Many studies hypothesized that ATC derives from Differentiated Thyroid Carcinoma (DTC) through a de-differentiation process triggered by specific molecular events still largely unknown. E2F7 is an atypical member of the E2F family. Known as cell cycle inhibitor and keeper of genomic stability, in specific contexts its function is oncogenic, guiding cancer progression. We performed a meta-analysis on 279 gene expression profiles, from 8 Gene Expression Omnibus patient samples datasets, to explore the causal relationship between DTC and ATC. We defined 3 specific gene signatures describing the evolution from normal thyroid tissue to DTC and ATC and validated them in a cohort of human surgically resected ATCs collected in our Institution. We identified E2F7 as a key player in the DTC-ATC transition and showed in vitro that its down-regulation reduced ATC cells' aggressiveness features. RNA-seq and ChIP-seq profiling allowed the identification of the E2F7 specific gene program, which is mainly related to cell cycle progression and DNA repair ability. Overall, this study identified a signature describing DTC de-differentiation toward ATC subtype and unveiled an E2F7-dependent transcriptional program supporting this process.

A network approach for low dimensional signatures from high throughput data.

One of the main objectives of high-throughput genomics studies is to obtain a low-dimensional set of observables-a signature-for sample classification purposes (diagnosis, prognosis, stratification). Biological data, such as gene or protein expression, are commonly characterized by an up/down regulation behavior, for which discriminant-based methods could perform with high accuracy and easy interpretability. To obtain the most out of these methods features selection is even more critical, but it is known to be a NP-hard problem, and thus most feature selection approaches focuses on one feature at the time (k-best, Sequential Feature Selection, recursive feature elimination). We propose DNetPRO, Discriminant Analysis with Network PROcessing, a supervised network-based signature identification method. This method implements a network-based heuristic to generate one or more signatures out of the best performing feature pairs. The algorithm is easily scalable, allowing efficient computing for high number of observables ([Formula: see text]-[Formula: see text]). We show applications on real high-throughput genomic datasets in which our method outperforms existing results, or is compatible with them but with a smaller number of selected features. Moreover, the geometrical simplicity of the resulting class-separation surfaces allows a clearer interpretation of the obtained signatures in comparison to nonlinear classification models.

Intraspecies characterization of bacteria via evolutionary modeling of protein domains.

The ability to detect and characterize bacteria within a biological sample is crucial for the monitoring of infections and epidemics, as well as for the study of human health and its relationship with commensal microorganisms. To this aim, a commonly used technique is the 16S rRNA gene targeted sequencing. PCR-amplified 16S sequences derived from the sample of interest are usually clustered into the so-called Operational Taxonomic Units (OTUs) based on pairwise similarities. Then, representative OTU sequences are compared with reference (human-made) databases to derive their phylogeny and taxonomic classification. Here, we propose a new reference-free approach to define the phylogenetic distance between bacteria based on protein domains, which are the evolving units of proteins. We extract the protein domain profiles of 3368 bacterial genomes and we use an ecological approach to model their Relative Species Abundance distribution. Based on the model parameters, we then derive a new measurement of phylogenetic distance. Finally, we show that such model-based distance is capable of detecting differences between bacteria in cases in which the 16S rRNA-based method fails, providing a possibly complementary approach , which is particularly promising for the analysis of bacterial populations measured by shotgun sequencing.

Identification of a Maturation Plasma Cell Index through a Highly Sensitive Droplet Digital PCR Assay Gene Expression Signature Validation in Newly Diagnosed Multiple Myeloma Patients.

DNA microarrays and RNA-based sequencing approaches are considered important discovery tools in clinical medicine. However, cross-platform reproducibility studies undertaken so far have highlighted that microarrays are not able to accurately measure gene expression, particularly when they are expressed at low levels. Here, we consider the employment of a digital PCR assay (ddPCR) to validate a gene signature previously identified by gene expression profile. This signature included ten Hedgehog (HH) pathways' genes able to stratify multiple myeloma (MM) patients according to their self-renewal status. Results show that the designed assay is able to validate gene expression data, both in a retrospective as well as in a prospective cohort. In addition, the plasma cells' differentiation status determined by ddPCR was further confirmed by other techniques, such as flow cytometry, allowing the identification of patients with immature plasma cells' phenotype (i.e., expressing CD19+/CD81+ markers) upregulating HH genes, as compared to others, whose plasma cells lose the expression of these markers and were more differentiated. To our knowledge, this is the first technical report of gene expression data validation by ddPCR instead of classical qPCR. This approach permitted the identification of a Maturation Index through the integration of molecular and phenotypic data, able to possibly define upfront the differentiation status of MM patients that would be clinically relevant in the future.

AC amplification gain in organic electrochemical transistors for impedance-based single cell sensors.

Research on electrolyte-gated and organic electrochemical transistor (OECT) architectures is motivated by the prospect of a highly biocompatible interface capable of amplifying bioelectronic signals at the site of detection. Despite many demonstrations in these directions, a quantitative model for OECTs as impedance biosensors is still lacking. We overcome this issue by introducing a model experiment where we simulate the detection of a single cell by the impedance sensing of a dielectric microparticle. The highly reproducible experiment allows us to study the impact of transistor geometry and operation conditions on device sensitivity. With the data we rationalize a mathematical model that provides clear guidelines for the optimization of OECTs as single cell sensors, and we verify the quantitative predictions in an in-vitro experiment. In the optimized geometry, the OECT-based impedance sensor allows to record single cell adhesion and detachment transients, showing a maximum gain of 20.2±0.9 dB with respect to a single electrode-based impedance sensor.

Source Attribution of Human Campylobacteriosis Using Whole-Genome Sequencing Data and Network Analysis.

Campylobacter spp. are a leading and increasing cause of gastrointestinal infections worldwide. Source attribution, which apportions human infection cases to different animal species and food reservoirs, has been instrumental in control- and evidence-based intervention efforts. The rapid increase in whole-genome sequencing data provides an opportunity for higher-resolution source attribution models. Important challenges, including the high dimension and complex structure of WGS data, have inspired concerted research efforts to develop new models. We propose network analysis models as an accurate, high-resolution source attribution approach for the sources of human campylobacteriosis. A weighted network analysis approach was used in this study for source attribution comparing different WGS data inputs. The compared model inputs consisted of cgMLST and wgMLST distance matrices from 717 human and 717 animal isolates from cattle, chickens, dogs, ducks, pigs and turkeys. SNP distance matrices from 720 human and 720 animal isolates were also used. The data were collected from 2015 to 2017 in Denmark, with the animal sources consisting of domestic and imports from 7 European countries. Clusters consisted of network nodes representing respective genomes and links representing distances between genomes. Based on the results, animal sources were the main driving factor for cluster formation, followed by type of species and sampling year. The coherence source clustering (CSC) values based on animal sources were 78%, 81% and 78% for cgMLST, wgMLST and SNP, respectively. The CSC values based on Campylobacter species were 78%, 79% and 69% for cgMLST, wgMLST and SNP, respectively. Including human isolates in the network resulted in 88%, 77% and 88% of the total human isolates being clustered with the different animal sources for cgMLST, wgMLST and SNP, respectively. Between 12% and 23% of human isolates were not attributed to any animal source. Most of the human genomes were attributed to chickens from Denmark, with an average attribution percentage of 52.8%, 52.2% and 51.2% for cgMLST, wgMLST and SNP distance matrices respectively, while ducks from Denmark showed the least attribution of 0% for all three distance matrices. The best-performing model was the one using wgMLST distance matrix as input data, which had a CSC value of 81%. Results from our study show that the weighted network-based approach for source attribution is reliable and can be used as an alternative method for source attribution considering the high performance of the model. The model is also robust across the different Campylobacter species, animal sources and WGS data types used as input.

Automated Prediction of the Response to Neoadjuvant Chemoradiotherapy in Patients Affected by Rectal Cancer.

BACKGROUND: Rectal cancer is a malignant neoplasm of the large intestine resulting from the uncontrolled proliferation of the rectal tract. Predicting the pathologic response of neoadjuvant chemoradiotherapy at an MRI primary staging scan in patients affected by locally advanced rectal cancer (LARC) could lead to significant improvement in the survival and quality of life of the patients. In this study, the possibility of automatizing this estimation from a primary staging MRI scan, using a fully automated artificial intelligence-based model for the segmentation and consequent characterization of the tumor areas using radiomic features was evaluated. The TRG score was used to evaluate the clinical outcome. METHODS: Forty-three patients under treatment in the IRCCS Sant'Orsola-Malpighi Polyclinic were retrospectively selected for the study; a U-Net model was trained for the automated segmentation of the tumor areas; the radiomic features were collected and used to predict the tumor regression grade (TRG) score. RESULTS: The segmentation of tumor areas outperformed the state-of-the-art results in terms of the Dice score coefficient or was comparable to them but with the advantage of considering mucinous cases. Analysis of the radiomic features extracted from the lesion areas allowed us to predict the TRG score, with the results agreeing with the state-of-the-art results. CONCLUSIONS: The results obtained regarding TRG prediction using the proposed fully automated pipeline prove its possible usage as a viable decision support system for radiologists in clinical practice.

Effectiveness of Biologically Inspired Neural Network Models in Learning and Patterns Memorization.

Purpose: In this work, we propose an implementation of the Bienenstock-Cooper-Munro (BCM) model, obtained by a combination of the classical framework and modern deep learning methodologies. The BCM model remains one of the most promising approaches to modeling the synaptic plasticity of neurons, but its application has remained mainly confined to neuroscience simulations and few applications in data science. Methods: To improve the convergence efficiency of the BCM model, we combine the original plasticity rule with the optimization tools of modern deep learning. By numerical simulation on standard benchmark datasets, we prove the efficiency of the BCM model in learning, memorization capacity, and feature extraction. Results: In all the numerical simulations, the visualization of neuronal synaptic weights confirms the memorization of human-interpretable subsets of patterns. We numerically prove that the selectivity obtained by BCM neurons is indicative of an internal feature extraction procedure, useful for patterns clustering and classification. The introduction of competitiveness between neurons in the same BCM network allows the network to modulate the memorization capacity of the model and the consequent model selectivity. Conclusions: The proposed improvements make the BCM model a suitable alternative to standard machine learning techniques for both feature selection and classification tasks.

Circumventing the curse of dimensionality in magnetic resonance fingerprinting through a deep learning approach.

Magnetic resonance fingerprinting (MRF) is a rapidly developing approach for fast quantitative MRI. A typical drawback of dictionary-based MRF is an explosion of the dictionary size as a function of the number of reconstructed parameters, according to the "curse of dimensionality", which determines an explosion of resource requirements. Neural networks (NNs) have been proposed as a feasible alternative, but this approach is still in its infancy. In this work, we design a deep learning approach to MRF using a fully connected network (FCN). In the first part we investigate, by means of simulations, how the NN performance scales with the number of parameters to be retrieved in comparison with the standard dictionary approach. Four MRF sequences were considered: IR-FISP, bSSFP, IR-FISP-B1 , and IR-bSSFP-B1 , the latter two designed to be more specific for B1+ parameter encoding. Estimation accuracy, memory usage, and computational time required to perform the estimation task were considered to compare the scalability capabilities of the dictionary-based and the NN approaches. In the second part we study optimal training procedures by including different data augmentation and preprocessing strategies during training to achieve better accuracy and robustness to noise and undersampling artifacts. The study is conducted using the IR-FISP MRF sequence exploiting both simulations and in vivo acquisitions. Results demonstrate that the NN approach outperforms the dictionary-based approach in terms of scalability capabilities. Results also allow us to heuristically determine the optimal training strategy to make an FCN able to predict T1 , T2 , and M0 maps that are in good agreement with those obtained with the original dictionary approach. k-SVD denoising is proposed and found to be critical as a preprocessing step to handle undersampled data.

Dynamics of social media behavior before and after SARS-CoV-2 infection.

Introduction: Online social media have been both a field of research and a source of data for research since the beginning of the COVID-19 pandemic. In this study, we aimed to determine how and whether the content of tweets by Twitter users reporting SARS-CoV-2 infections changed over time. Methods: We built a regular expression to detect users reporting being infected, and we applied several Natural Language Processing methods to assess the emotions, topics, and self-reports of symptoms present in the timelines of the users. Results: Twelve thousand one hundred and twenty-one twitter users matched the regular expression and were considered in the study. We found that the proportions of health-related, symptom-containing, and emotionally non-neutral tweets increased after users had reported their SARS-CoV-2 infection on Twitter. Our results also show that the number of weeks accounting for the increased proportion of symptoms was consistent with the duration of the symptoms in clinically confirmed COVID-19 cases. Furthermore, we observed a high temporal correlation between self-reports of SARS-CoV-2 infection and officially reported cases of the disease in the largest English-speaking countries. Discussion: This study confirms that automated methods can be used to find digital users publicly sharing information about their health status on social media, and that the associated data analysis may supplement clinical assessments made in the early phases of the spread of emerging diseases. Such automated methods may prove particularly useful for newly emerging health conditions that are not rapidly captured in the traditional health systems, such as the long term sequalae of SARS-CoV-2 infections.