Mary-Anne Hartley

Deep learning diagnostic and severity-stratification for interstitial lung diseases and chronic obstructive pulmonary disease in digital lung auscultations and ultrasonography: clinical protocol for an observational case-control study

Mary-Anne Hartley

BackgroundInterstitial lung diseases (ILD), such as idiopathic pulmonary fibrosis (IPF) and non-specific interstitial pneumonia (NSIP), and chronic obstructive pulmonary disease (COPD) are severe, progressive pulmonary disorders with a poor prognosis. Prompt and accurate diagnosis is important to enable patients to receive appropriate care at the earliest possible stage to delay disease progression and prolong survival. Artificial intelligence-assisted lung auscultation and ultrasound (LUS) could constitute an alternative to conventional, subjective, operator-related methods for the accurate and earlier diagnosis of these diseases. This protocol describes the standardised collection of digitally-acquired lung sounds and LUS images of adult outpatients with IPF, NSIP or COPD and a deep learning diagnostic and severity-stratification approach.MethodsA total of 120 consecutive patients (>= 18 years) meeting international criteria for IPF, NSIP or COPD and 40 age-matched controls will be recruited in a Swiss pulmonology outpatient clinic, starting from August 2022. At inclusion, demographic and clinical data will be collected. Lung auscultation will be recorded with a digital stethoscope at 10 thoracic sites in each patient and LUS images using a standard point-of-care device will be acquired at the same sites. A deep learning algorithm (DeepBreath) using convolutional neural networks, long short-term memory models, and transformer architectures will be trained on these audio recordings and LUS images to derive an automated diagnostic tool. The primary outcome is the diagnosis of ILD versus control subjects or COPD. Secondary outcomes are the clinical, functional and radiological characteristics of IPF, NSIP and COPD diagnosis. Quality of life will be measured with dedicated questionnaires. Based on previous work to distinguish normal and pathological lung sounds, we estimate to achieve convergence with an area under the receiver operating characteristic curve of > 80% using 40 patients in each category, yielding a sample size calculation of 80 ILD (40 IPF, 40 NSIP), 40 COPD, and 40 controls.DiscussionThis approach has a broad potential to better guide care management by exploring the synergistic value of several point-of-care-tests for the automated detection and differential diagnosis of ILD and COPD and to estimate severity.Trial registration Registration: August 8, 2022. ClinicalTrials.gov Identifier: NCT05318599.

BMC2023

DeepBreath-automated detection of respiratory pathology from lung auscultation in 572 pediatric outpatients across 5 countries

Tatjana Chavdarova, Juliane Dervaux, Mary-Anne Hartley, Julien Niklas Heitmann, Daniel Hinjos García, Martin Jaggi, Daniel Mueller, Alexandre Perez

The interpretation of lung auscultation is highly subjective and relies on non-specific nomenclature. Computer-aided analysis has the potential to better standardize and automate evaluation. We used 35.9 hours of auscultation audio from 572 pediatric outpatients to develop DeepBreath : a deep learning model identifying the audible signatures of acute respiratory illness in children. It comprises a convolutional neural network followed by a logistic regression classifier, aggregating estimates on recordings from eight thoracic sites into a single prediction at the patient-level. Patients were either healthy controls (29%) or had one of three acute respiratory illnesses (71%) including pneumonia, wheezing disorders (bronchitis/asthma), and bronchiolitis). To ensure objective estimates on model generalisability, DeepBreath is trained on patients from two countries (Switzerland, Brazil), and results are reported on an internal 5-fold cross-validation as well as externally validated (extval) on three other countries (Senegal, Cameroon, Morocco). DeepBreath differentiated healthy and pathological breathing with an Area Under the Receiver-Operator Characteristic (AUROC) of 0.93 (standard deviation [SD] +/- 0.01 on internal validation). Similarly promising results were obtained for pneumonia (AUROC 0.75 +/- 0.10), wheezing disorders (AUROC 0.91 +/- 0.03), and bronchiolitis (AUROC 0.94 +/- 0.02). Extval AUROCs were 0.89, 0.74, 0.74 and 0.87 respectively. All either matched or were significant improvements on a clinical baseline model using age and respiratory rate. Temporal attention showed clear alignment between model prediction and independently annotated respiratory cycles, providing evidence that DeepBreath extracts physiologically meaningful representations. DeepBreath provides a framework for interpretable deep learning to identify the objective audio signatures of respiratory pathology.

NATURE PORTFOLIO2023

Point-of-care lung ultrasonography for early identification of mild COVID-19: a prospective cohort of outpatients in a Swiss screening center

Luca Bosso, Jean-Baptiste Francis Marie Juliette Cordonnier, Tanguy Espejo, Mary-Anne Hartley, Jean-Yves Meuwly, Siméon Schaad

Objectives Early identification of SARS-CoV-2 infection is important to guide quarantine and reduce transmission. This study evaluates the diagnostic performance of lung ultrasound (LUS), an affordable, consumable-free point-of-care tool, for COVID-19 screening. Design, setting and participants This prospective observational cohort included adults presenting with cough and/or dyspnoea at a SARS-CoV-2 screening centre of Lausanne University Hospital between 31 March and 8 May 2020. Interventions Investigators recorded standardised LUS images and videos in 10 lung zones per patient. Two blinded independent experts reviewed LUS recording and classified abnormal findings according to prespecified criteria to investigate their predictive value to diagnose SARS-CoV-2 infection according to PCR on nasopharyngeal swabs (COVID-19 positive vs COVID-19 negative). Primary and secondary outcome measures We finally combined LUS and clinical findings to derive a multivariate logistic regression diagnostic score. Results Of 134 included patients, 23% (n=30/134) were COVID-19 positive and 77% (n=103/134) were COVID-19 negative; 85%, (n=114/134) cases were previously healthy healthcare workers presenting within 2-5 days of symptom onset (IQR). Abnormal LUS findings were significantly more frequent in COVID-19 positive compared with COVID-19 negative (45% vs 26%, p=0.045) and mostly consisted of focal pathologic B lines. Combining clinical findings in a multivariate logistic regression score had an area under the receiver operating curve of 80.3% to detect COVID-19, and slightly improved to 84.5% with the addition of LUS features. Conclusions COVID-19-positive patients are significantly more likely to have lung pathology by LUS. However, LUS has an insufficient sensitivity and is not an appropriate screening tool in outpatients. LUS only adds little value to clinical features alone.

BMJ PUBLISHING GROUP2022

Point-of-care ultrasound for tuberculosis management in Sub-Saharan Africa-a balanced SWOT analysis

Mary-Anne Hartley

Point-of-care ultrasound (POCUS) is an increasingly accessible skill, allowing for the decentralization of its use to nonspecialist healthcare workers to guide routine clinical decision-making. The advent of ultrasound-on-a-chip has transformed the technology into a portable mobile health device. Because of its high sensitivity to detect small consolidations, pleural effusions, and subpleural nodules, POCUS has recently been proposed as a sputum-free likely triage tool for tuberculosis (TB). To make an objective assessment of the potential and limitations of POCUS in routine TB management, we present a Strengths, Weaknesses, Opportunities and Threats (SWOT) analysis based on a review of the relevant literature and focusing on Sub-Saharan Africa (SSA). We identified numerous strengths and opportunities of POCUS for TB management, e.g., accessible, affordable, easy to use and maintain, expedited diagnosis, extrapulmonary TB detection, safer pleural/pericardial puncture, use in children/pregnant women/people living with HIV, targeted screening of TB contacts, monitoring TB sequelae, and creating artificial intelligence decision support. Weaknesses and external threats such as operator dependency, lack of visualization of central lung pathology, poor specificity, lack of impact assessments and data from SSA must be taken into consideration to ensure that the potential of the technology can be fully realized in research as in practice. (c) 2022 Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

ELSEVIER SCI LTD2022

Lung Ultrasonography for Risk Stratification in Patients with Coronavirus Disease 2019 (COVID-19): A Prospective Observational Cohort Study

Mary-Anne Hartley, Jean-Yves Meuwly

Background. Lung ultrasonography (LUS) is a promising pragmatic risk-stratification tool in coronavirus disease 2019 (COVID-19). This study describes and compares LUS characteristics between patients with different clinical outcomes. Methods. Prospective observational study of polymerase chain reaction-confirmed adults with COVID-19 with symptoms of lower respiratory tract infection in the emergency department (ED) of Lausanne University Hospital. A trained physician recorded LUS images using a standardized protocol. Two experts reviewed images blinded to patient outcome. We describe and compare early LUS findings (

OXFORD UNIV PRESS INC2021

Clinical relevance of low-densityPlasmodium falciparumparasitemia in untreated febrile children: A cohort study

Mary-Anne Hartley

Background Low-density (LD)Plasmodiuminfections are missed by standard malaria rapid diagnostic tests (standard mRDT) when the blood antigen concentration is below the detection threshold. The clinical impact of these LD infections is unknown. This study investigates the clinical presentation and outcome of untreated febrile children with LD infections attending primary care facilities in a moderately endemic area of Tanzania. Methods/findings This cohort study includes 2,801 febrile pediatric outpatients (median age 13.5 months [range 2-59], female:male ratio 0.8:1.0) recruited in Dar es Salaam, Tanzania between 01 December 2014 and 28 February 2016. Treatment decisions were guided by a clinical decision support algorithm run on a mobile app, which also collected clinical data. Only standard mRDT+ cases received antimalarials. Outcomes (clinical failure, secondary hospitalization, and death) were collected in follow-up visits or interviews on days 3, 7, and 28. After patient recruitment had ended, frozen blood from all 2,801 patients was tested forPlasmodium falciparum(Pf) by ultrasensitive-quantitative polymerase chain reaction (qPCR), standard mRDT, and "ultrasensitive" mRDT. As the latter did not improve sensitivity beyond standard mRDT, it is hereafter excluded. Clinical features and outcomes in LD patients (standard mRDT-/ultrasensitive-qPCR+, not given antimalarials) were compared with those with no detectable (ND) parasitemia (standard mRDT-/ultrasensitive-qPCR-) or high-density (HD) infections (standard mRDT+/ultrasensitive-qPCR+, antimalarial-treated). Pfpositivity rate was 7.1% (n= 199/2,801) and 9.8% (n= 274/2,801) by standard mRDT and ultrasensitive qPCR, respectively. Thus, 28.0% (n = 76/274) of ultrasensitive qPCR+ cases were not detected by standard mRDT and labeled "LD". LD patients were, on average, 10.6 months younger than those with HD infections (95% CI 7.0-14.3 months,p< 0.001). Compared with ND, LD patients more frequently had the diagnosis of undifferentiated fever of presumed viral origin (risk ratio [RR] = 2.0, 95% CI 1.3-3.1,p= 0.003) and were more often suffering from severe malnutrition (RR = 3.2, 95% CI 1.1-7.5,p= 0.03). Despite not receiving antimalarials, outcomes for the LD group did not differ from ND regarding clinical failures (2.6% [n= 2/76] versus 4.0% [n= 101/2,527], RR = 0.7, 95% CI 0.2-3.5,p= 0.7) or secondary hospitalizations (2.6% [n= 2/76] versus 2.8% [n= 72/2,527], RR = 0.7,95% CI 0.2-3.2,p= 0.9), and no deaths were reported in anyPf-positive groups. HD patients experienced more secondary hospitalizations (10.1% [n= 20/198], RR = 0.3, 95% CI 0.1-1.0,p= 0.005) than LD patients. All the patients in this cohort were febrile children; thus, the association between parasitemia and fever cannot be investigated, nor can the conclusions be extrapolated to neonates and adults. Conclusions During a 28-day follow-up period, we did not find evidence of a difference in negative outcomes between febrile children with untreated LDPfparasitemia and those withoutPfparasitemia. These findings suggest LD parasitemia may either be a self-resolving fever or an incidental finding in children with other infections, including those of viral origin. These findings do not support a clinical benefit nor additional risk (e.g. because of missed bacterial infections) to using ultrasensitive malaria diagnostics at a primary care level. Author summaryWhy was this study done? Every infection has a symptomatic threshold, above which the magnitude of infection burden triggers fever and other clinical symptoms that distinguishes ill individuals from those who remain asymptomatic. In malaria, this is called the "pyrogenic (fever) threshold." The standard malaria rapid diagnostic tests (sd-mRDTs), which have been in use for decades, can detect parasite loads of up to 10-fold below this threshold. New ultrasensitive malaria tests are now able to detect genetic traces of parasites that go 1,000-fold further. However, the clinical relevance of these minute levels of parasitemia is unknown. Does it detect a previously unappreciated cause of clinical failure? Or does it risk reporting a clinically irrelevant signal that may distract the clinician from other causes of fever? What did the researchers do and find? We assessed the clinical consequences of untreated low-density (LD) parasitemia that is only detectable with the ultrasensitive malaria test. In a cohort of 2,801 pediatric outpatients in Tanzania, we observed that of all patients in whom parasites were detected using ultrasensitive tests, over a quarter were not detected by standard mRDTs. However, despite not receiving antimalarials, patients with LD parasitemia did not differ from those without any detectable parasites regarding clinical outcomes during a 28-day follow-up. What do these findings mean? These findings suggest LD parasitemia may either be a self-resolving fever or an incidental finding in children with other infections. These findings neither support a specific utility nor risk (other than the obvious wasted resources) to using ultrasensitive malaria diagnostics at a primary care level.

2020

Weight Erosion: An Update Aggregation Scheme for Personalized Collaborative Machine Learning

Felix Hans Michel Grimberg, Mary-Anne Hartley, Martin Jaggi, Sai Praneeth Reddy Karimireddy

Background. In medicine and other applications, the copying and sharing of data is impractical for a range of well-considered reasons. With federated learning (FL) techniques, machine learning models can be trained on data spread across several locations without such copying and sharing. While good privacy guarantees can often be made, FL does not automatically incentivize participation and the resulting model can suffer if data is non-identically distributed (non-IID) across locations. Model personalization is a way of addressing these concerns. Methods. In this study, we introduce Weight Erosion: an SGD-based gradient aggregation scheme for personalized collaborative ML. We evaluate this scheme on a binary classification task in the Titanic data set. Findings. We demonstrate that the novel Weight Erosion scheme can outperform two baseline FL aggregation schemes on a classification task, and is more resistant to over-fitting and non-IID data sets.

Springer, Cham2020

Detection of dicistroviruses RNA in blood of febrile Tanzanian children

Mary-Anne Hartley, Thomas Junier

Fever is the leading cause of paediatric outpatient consultations in Sub-Saharan Africa. Although most are suspected to be of viral origin, a putative causative pathogen is not identified in over a quarter of these febrile episodes. Using a de novo assembly sequencing approach, we report the detection (15.4%) of dicistroviruses (DicV) RNA in sera collected from 692 febrile Tanzanian children. In contrast, DicV RNA was only detected in 1/77 (1.3%) plasma samples from febrile Tanzanian adults, suggesting that children could represent the primary susceptible population. Estimated viral load by specific quantitative real-time RT-PCR assay ranged from < 1.32E3 to 1.44E7 viral RNA copies/mL serum. Three DicV full-length genomes were obtained, and a phylogenetic analyse on the capsid region showed the presence of two clusters representing tentative novel genus. Although DicV-positive cases were detected throughout the year, a significantly higher positivity rate was observed during the rainy season. This study reveals that novel DicV RNA is frequently detected in the blood of Tanzanian children, paving the way for further investigations to determine if DicV possibly represent a new agent in humans.

TAYLOR & FRANCIS LTD2019

Mary-Anne Hartley

Deep learning diagnostic and severity-stratification for interstitial lung diseases and chronic obstructive pulmonary disease in digital lung auscultations and ultrasonography: clinical protocol for an observational case-control study

DeepBreath-automated detection of respiratory pathology from lung auscultation in 572 pediatric outpatients across 5 countries

Point-of-care lung ultrasonography for early identification of mild COVID-19: a prospective cohort of outpatients in a Swiss screening center

Point-of-care ultrasound for tuberculosis management in Sub-Saharan Africa-a balanced SWOT analysis

Ultrasound for point-of-care sputum-free tuberculosis detection: Building collaborative standardized image-banks

Lung Ultrasonography for Risk Stratification in Patients with Coronavirus Disease 2019 (COVID-19): A Prospective Observational Cohort Study

Clinical relevance of low-densityPlasmodium falciparumparasitemia in untreated febrile children: A cohort study

Weight Erosion: An Update Aggregation Scheme for Personalized Collaborative Machine Learning

Detection of dicistroviruses RNA in blood of febrile Tanzanian children

Deep learning diagnostic and severity-stratification for interstitial lung diseases and chronic obstructive pulmonary disease in digital lung auscultations and ultrasonography: clinical protocol for an observational case-control study

Detection of dicistroviruses RNA in blood of febrile Tanzanian children

Ultrasound for point-of-care sputum-free tuberculosis detection: Building collaborative standardized image-banks

Point-of-care lung ultrasonography for early identification of mild COVID-19: a prospective cohort of outpatients in a Swiss screening center

DeepBreath-automated detection of respiratory pathology from lung auscultation in 572 pediatric outpatients across 5 countries

Clinical relevance of low-densityPlasmodium falciparumparasitemia in untreated febrile children: A cohort study

Point-of-care ultrasound for tuberculosis management in Sub-Saharan Africa-a balanced SWOT analysis

Lung Ultrasonography for Risk Stratification in Patients with Coronavirus Disease 2019 (COVID-19): A Prospective Observational Cohort Study

Weight Erosion: An Update Aggregation Scheme for Personalized Collaborative Machine Learning