Simulating federated learning for steatosis detection using ultrasound images.

We aimed to implement four data partitioning strategies evaluated with four federated learning (FL) algorithms and investigate the impact of data distribution on FL model performance in detecting steatosis using B-mode US images. A private dataset (153 patients; 1530 images) and a public dataset (55 patient; 550 images) were included in this retrospective study. The datasets contained patients with metabolic dysfunction-associated fatty liver disease (MAFLD) with biopsy-proven steatosis grades and control individuals without steatosis. We employed four data partitioning strategies to simulate FL scenarios and we assessed four FL algorithms. We investigated the impact of class imbalance and the mismatch between the global and local data distributions on the learning outcome. Classification performance was assessed with area under the receiver operating characteristic curve (AUC) on a separate test set. AUCs were 0.93 (95% CI 0.92, 0.94) for source-based partitioning scenario with FedAvg, 0.90 (95% CI 0.89, 0.91) for a centralized model, and 0.83 (95% CI 0.81, 0.85) for a model trained in a single-center scenario. When data was perfectly balanced on the global level and each site had an identical data distribution, the model yielded an AUC of 0.90 (95% CI 0.88, 0.92). When each site contained data exclusively from one single class, irrespective of the global data distribution, the AUC fell in the range of 0.34-0.70. FL applied to B-mode US images provide performance comparable to a centralized model and higher than single-center scenario. Global data imbalance and local data heterogeneity influenced the learning outcome.

CODA: an open-source platform for federated analysis and machine learning on distributed healthcare data.

OBJECTIVES: Distributed computations facilitate multi-institutional data analysis while avoiding the costs and complexity of data pooling. Existing approaches lack crucial features, such as built-in medical standards and terminologies, no-code data visualizations, explicit disclosure control mechanisms, and support for basic statistical computations, in addition to gradient-based optimization capabilities. MATERIALS AND METHODS: We describe the development of the Collaborative Data Analysis (CODA) platform, and the design choices undertaken to address the key needs identified during our survey of stakeholders. We use a public dataset (MIMIC-IV) to demonstrate end-to-end multi-modal FL using CODA. We assessed the technical feasibility of deploying the CODA platform at 9 hospitals in Canada, describe implementation challenges, and evaluate its scalability on large patient populations. RESULTS: The CODA platform was designed, developed, and deployed between January 2020 and January 2023. Software code, documentation, and technical documents were released under an open-source license. Multi-modal federated averaging is illustrated using the MIMIC-IV and MIMIC-CXR datasets. To date, 8 out of the 9 participating sites have successfully deployed the platform, with a total enrolment of >1M patients. Mapping data from legacy systems to FHIR was the biggest barrier to implementation. DISCUSSION AND CONCLUSION: The CODA platform was developed and successfully deployed in a public healthcare setting in Canada, with heterogeneous information technology systems and capabilities. Ongoing efforts will use the platform to develop and prospectively validate models for risk assessment, proactive monitoring, and resource usage. Further work will also make tools available to facilitate migration from legacy formats to FHIR and DICOM.

Automated screening of potential organ donors using a temporal machine learning model.

Organ donation is not meeting demand, and yet 30-60% of potential donors are potentially not identified. Current systems rely on manual identification and referral to an Organ Donation Organization (ODO). We hypothesized that developing an automated screening system based on machine learning could reduce the proportion of missed potentially eligible organ donors. Using routine clinical data and laboratory time-series, we retrospectively developed and tested a neural network model to automatically identify potential organ donors. We first trained a convolutive autoencoder that learned from the longitudinal changes of over 100 types of laboratory results. We then added a deep neural network classifier. This model was compared to a simpler logistic regression model. We observed an AUROC of 0.966 (CI 0.949-0.981) for the neural network and 0.940 (0.908-0.969) for the logistic regression model. At a prespecified cutoff, sensitivity and specificity were similar between both models at 84% and 93%. Accuracy of the neural network model was robust across donor subgroups and remained stable in a prospective simulation, while the logistic regression model performance declined when applied to rarer subgroups and in the prospective simulation. Our findings support using machine learning models to help with the identification of potential organ donors using routinely collected clinical and laboratory data.

Interpretable clinical phenotypes among patients hospitalized with COVID-19 using cluster analysis.

Background: Multiple clinical phenotypes have been proposed for coronavirus disease (COVID-19), but few have used multimodal data. Using clinical and imaging data, we aimed to identify distinct clinical phenotypes in patients admitted with COVID-19 and to assess their clinical outcomes. Our secondary objective was to demonstrate the clinical applicability of this method by developing an interpretable model for phenotype assignment. Methods: We analyzed data from 547 patients hospitalized with COVID-19 at a Canadian academic hospital. We processed the data by applying a factor analysis of mixed data (FAMD) and compared four clustering algorithms: k-means, partitioning around medoids (PAM), and divisive and agglomerative hierarchical clustering. We used imaging data and 34 clinical variables collected within the first 24 h of admission to train our algorithm. We conducted a survival analysis to compare the clinical outcomes across phenotypes. With the data split into training and validation sets (75/25 ratio), we developed a decision-tree-based model to facilitate the interpretation and assignment of the observed phenotypes. Results: Agglomerative hierarchical clustering was the most robust algorithm. We identified three clinical phenotypes: 79 patients (14%) in Cluster 1, 275 patients (50%) in Cluster 2, and 203 (37%) in Cluster 3. Cluster 2 and Cluster 3 were both characterized by a low-risk respiratory and inflammatory profile but differed in terms of demographics. Compared with Cluster 3, Cluster 2 comprised older patients with more comorbidities. Cluster 1 represented the group with the most severe clinical presentation, as inferred by the highest rate of hypoxemia and the highest radiological burden. Intensive care unit (ICU) admission and mechanical ventilation risks were the highest in Cluster 1. Using only two to four decision rules, the classification and regression tree (CART) phenotype assignment model achieved an AUC of 84% (81.5-86.5%, 95 CI) on the validation set. Conclusions: We conducted a multidimensional phenotypic analysis of adult inpatients with COVID-19 and identified three distinct phenotypes associated with different clinical outcomes. We also demonstrated the clinical usability of this approach, as phenotypes can be accurately assigned using a simple decision tree. Further research is still needed to properly incorporate these phenotypes in the management of patients with COVID-19.

Association Between Functional Tricuspid Regurgitation and Mortality Following Cardiac Surgery.

Background: Current guidelines recommend concomitant repair of certain non-severe cases of tricuspid regurgitation (TR) in patients undergoing cardiac surgery, but the prognostic relevance and postsurgical impact of the TR remain uncertain. Objectives: The purpose of this study was to determine the prognostic impact of functional TR in patients undergoing diverse cardiac surgeries and to examine the effect-modifying role of patient characteristics in patients in whom TR confers a greater risk of adverse outcomes. Methods: Patients undergoing coronary artery bypass, aortic, and mitral valve surgery were included. Patients with severe TR, organic tricuspid valve pathology, undergoing tricuspid valve surgery or without a recent preoperative echocardiogram were excluded. Clinical variables were extracted from the Society of Thoracic Surgeons Adult Cardiac Surgery Database. An independent cohort was used for external validation. Results: Of 2,119 patients (mean age 67.4 years; 29% females), TR severity was moderate in 185 (9%), mild in 636 (30%), trivial in 1,126 (53%), and absent in 172 (8%). There were 238 deaths during the median follow-up period of 2.6 years. After adjusting for relevant factors, moderate TR was found to be independently associated with mid-term mortality (HR: 2.58; 95% CI: 1.22-5.47) and with in-hospital mortality or major morbidity (OR: 3.18; 95% CI: 1.37-7.42). The association between TR and mortality was apparent when preoperative pulmonary artery systolic pressure was <40 mm Hg but not ≥40 mm Hg (P for interaction = 0.036). Conclusions: In this diverse cohort of contemporary cardiac surgery patients, moderate functional TR was associated with increased mortality and major morbidity, particularly in the absence of pulmonary hypertension.

Association between obesity and hospital mortality in critical COVID-19: a retrospective cohort study.

BACKGROUND: The impact of obesity on outcomes in acute respiratory distress syndrome (ARDS) is not well understood and remains controversial. Recent studies suggest that obesity might be associated with higher morbidity and mortality in respiratory disease caused by SARS-CoV-2 (COVID-19 disease). Our objective was to evaluate the association between obesity and hospital mortality in critical COVID-19 patients. METHODS: We conducted a retrospective cohort study in a tertiary academic center located in Montréal between March and August 2020. We included all consecutive adult patients admitted to the ICU for COVID-19-confirmed respiratory disease. Our main outcome was hospital mortality. We estimated the association between obesity, using the body mass index as a continuous variable, and hospital survival by fitting a multivariable Cox proportional hazards model. RESULTS: We included 94 patients. Median [q1, q3] body mass index (BMI) was 29 [26-32] kg/m2 and 37% of patients were obese (defined as BMI > 30 kg/m2). Hospital mortality for the entire cohort was 33%. BMI was significantly associated with hospital mortality (hazard ratio [HR] = 2.49 per 10 units BMI; 95% CI, from 1.69 to 3.70; p < 0.001) even after adjustment for sex, age and obesity-related comorbidities (adjusted HR = 3.50; 95% CI from 2.03 to 6.02; p < 0.001). CONCLUSIONS: Obesity was prevalent in hospitalized patients with critical illness secondary to COVID-19 disease and a higher BMI was associated with higher hospital mortality. Further studies are needed to validate this association and to better understand its underlying mechanisms.

CoreSlicer: a web toolkit for analytic morphomics.

BACKGROUND: Analytic morphomics, or more simply, "morphomics," refers to the measurement of specific biomarkers of body composition from medical imaging, most commonly computed tomography (CT) images. An emerging body of literature supports the use of morphomic markers measured on single-slice CT images for risk prediction in a range of clinical populations. However, uptake by healthcare providers been limited due to the lack of clinician-friendly software to facilitate measurements. The objectives of this study were to describe the interface and functionality of CoreSlicer- a free and open-source web-based interface aiming to facilitate measurement of analytic morphomics by clinicians - and to validate muscle and fat measurements performed in CoreSlicer against reference software. RESULTS: Measurements of muscle and fat obtained in CoreSlicer show high agreement with established reference software. CoreSlicer features a full set of DICOM viewing tools and extensible plugin interface to facilitate rapid prototyping and validation of new morphomic markers by researchers. We present published studies illustrating the use of CoreSlicer by clinicians with no prior knowledge of medical image segmentation techniques and no formal training in radiology, where CoreSlicer was successfully used to predict operative risk in three distinct populations of cardiovascular patients. CONCLUSIONS: CoreSlicer enables extraction of morphomic markers from CT images by non-technically skilled clinicians. Measurements were reproducible and accurate in relation to reference software.

Phase Angle as a Biomarker for Frailty and Postoperative Mortality: The BICS Study.

Background Phase angle (PA) is a bioimpedance measurement that is determined lean body mass and hydration status. Patients with low PA values are more likely to be frail, sarcopenic, or malnourished. Previous work has shown that low PA predicts adverse outcomes after cardiac surgery, but the effect of PA on survival has not previously been assessed in this setting. Methods and Results The BICS (Bioimpedance in Cardiac Surgery) study recruited 277 patients undergoing major cardiac surgery at 2 university-affiliated hospitals in Montreal, QC, Canada. Bioimpedance measurements as well as frailty and nutritional assessments were performed preoperatively. The primary outcome was all-cause mortality. Secondary outcomes were 30-day mortality, postoperative morbidity, and hospital length of stay. There were 10 deaths at 1 month of follow-up and 16 deaths at 12 months of follow-up. PA was associated with age, sex, body mass index, comorbidities, and frailty, as measured by the Short Physical Performance Battery and Fried scales. After adjusting for Society of Thoracic Surgeons-predicted mortality, lower PA was associated with higher mortality at 1 month (adjusted odds ratio, 3.57 per 1° decrease in PA ; 95% confidence interval, 1.35-9.47) and at 12 months (adjusted odds ratio, 3.03 per 1° decrease in PA ; 95% confidence interval, 1.30-7.09), a higher risk of overall morbidity (adjusted hazard ratio, 2.51 per 1° decrease in PA ; 95% confidence interval, 1.32-4.75), and a longer hospital length of stay (adjusted ß, 4.8 days per 1° decrease in PA ; 95% confidence interval, 1.3-8.2 days). Conclusions Low PA is associated with frailty and is predictive of mortality, morbidity, and length of stay after major cardiac surgery. Further work is needed to determine the responsiveness of PA to interventions aimed at reversing frailty.

Muscle Mass and Mortality After Cardiac Transplantation.

BACKGROUND: Frailty assessment is recommended to evaluate the candidacy of adults referred for orthotopic heart transplantation (OHT). Psoas muscle area (PMA) is an easily measured biomarker for frailty. There has yet to be a study examining the prognostic impact of PMA in OHT patients. METHODS: In this retrospective study, preoperative and postoperative computed tomography (CT) scans were retrieved for adults transplanted between 2000 and 2015 at a tertiary care hospital. Psoas muscle area was measured on a single axial image. Outcomes of interest were all-cause mortality over 6 years and a composite of in-hospital mortality or major morbidity (prolonged ventilation, stroke, dialysis, mediastinitis, or reoperation). RESULTS: Of 161 adult patients transplanted, 82 had at least 1 abdominal CT scan. At baseline, mean PMA was 25.7 ± 5.8 cm in men and 16.0 ± 3.6 cm in women, and decreased by 8% from the first to the last available CT scan. Adjusting for age, sex, body mass index, and cardiomyopathy etiology, every 1-cm increase in PMA was found to be associated with a 9% reduction in long-term mortality (hazard ratio, 0.91; 95% confidence interval [CI], 0.83-0.99; P = 0.031) and a 17% reduction in in-hospital mortality or major morbidity (odds ratio, 0.83; 95% CI, 0.72-0.96; P = 0.014). When PMA was smaller than the sex-specific median, the risk of mortality or major morbidity increased fourfold (odds ratio, 4.29; 95% CI, 1.19-15.46; P = 0.026). CONCLUSIONS: Muscle mass is an independent predictor of mortality and major morbidity after OHT. Further research is needed to determine whether frail OHT patients with low PMA may benefit from muscle-building interventions to improve outcomes.

Psoas Muscle Area and Length of Stay in Older Adults Undergoing Cardiac Operations.

BACKGROUND: Frailty assessment can help predict which older adults will experience adverse events after cardiac surgical procedures. Low muscle mass is a core component of frailty that is suboptimally captured by self-reported weight loss; refined measures using computed tomographic (CT) images have emerged and are predictive of outcomes in noncardiac surgical procedures. The objective of this study was to evaluate the association between CT muscle area and length of stay (LOS) after cardiac surgical procedures. METHODS: Frail patients who had a perioperative abdominal or thoracic CT scan were identified. The CT scans were analyzed to measure cross-sectional lean muscle area at the L4 vertebra (psoas muscle area [PMA], lumbar muscle area [LMA]) and the T4 vertebra (thoracic muscle area [TMA]). The associations of PMA, LMA, and TMA with frailty markers and postoperative LOS were investigated. RESULTS: Eighty-two patients were included; the mean age was 69.2 ± 9.97 years. Low muscle area was correlated with lower handgrip strength and short physical performance battery (SPPB) scores indicative of physical frailty. Postoperative LOS was correlated with PMA (R = -0.47, p = 0.004), LMA (R = -0.41, p = 0.01), and TMA (R = -0.29, p = 0.03). After adjustment for the predicted risk of prolonged LOS, age, sex, and body surface area, PMA remained significantly associated with LOS (ß = -2.35, 95% CI -4.48 to -0.22). The combination of low PMA and handgrip strength, indicative of sarcopenia, yielded the greatest incremental value in predicting LOS. CONCLUSIONS: Low PMA is a marker of physical frailty associated with increased LOS in older adults undergoing cardiac surgical procedures. Further research is necessary to validate PMA as a prognostic marker and therapeutic target in this vulnerable population.

Psoas Muscle Area and All-Cause Mortality After Transcatheter Aortic Valve Replacement: The Montreal-Munich Study.

BACKGROUND: Psoas muscle area (PMA) is a novel measure of frailty that can be efficiently measured from computed tomography images to help predict risk in older adults referred for transcatheter aortic valve replacement (TAVR). The objective of this study was to determine if PMA would be incrementally predictive of mortality and morbidity after TAVR. METHODS: The pre-TAVR computed tomography scans of 208 consecutive patients at 2 hospitals in Montreal and Munich were analyzed to measure the cross-sectional area of the left and right psoas muscles on a single axial slice at the level of L4. The primary outcome was all-cause mortality assessed according to sex-stratified Cox regression models adjusted for the Society of Thoracic Surgeons predicted risk of mortality. RESULTS: The mean age was 80.7 ± 6.8 years with 55% women and a total of 57 deaths over a mean follow-up of 504 days. PMA was lower in nonsurvivors compared with survivors among women (12.9 vs 14.5 cm(2); P = 0.047) but not men (21.7 vs 22.4 cm(2); P = 0.50). The association between PMA and all-cause mortality in women persisted after adjustment for Society of Thoracic Surgeons risk (hazard ratio, 0.88 per cm(2); 95% confidence interval, 0.78-0.99). An association between PMA and bleeding complications was seen in men (odds ratio, 0.78; 95% confidence interval, 0.62-0.97). Sensitivity analyses with PMA normalized to body mass index yielded similar results. CONCLUSIONS: This study has shown that PMA is a marker of frailty associated with midterm survival in women who undergo TAVR. Further research is warranted to pursue PMA as a prognostic marker and therapeutic target in this vulnerable population.

Therapeutic interventions for frail elderly patients: part I. Published randomized trials.

The body of literature for frailty as a prognostic marker continues to grow, yet the evidence for frailty as a therapeutic target is less well defined. In the setting of cardiovascular disease, the prevalence of frailty is elevated and its impact on mortality and major morbidity is substantial. Therapeutic interventions aimed at improving frailty may impart gains in functional status and survival. Randomized clinical trials that tested one or more therapeutic interventions in a population of frail older adults were reviewed. The interventions studied were exercise training in 13 trials, nutritional supplementation in 4 trials, combined exercise plus nutritional supplementation in 7 trials, pharmaceutical agents in 8 trials, multi-dimensional programs in 5 trials, and home-based services in 1 trial. The main findings of these trials are explored along with a discussion of their relative merits and limitations.