Auto-FedRL: Federated Hyperparameter Optimization for Multi-institutional Medical Image Segmentation (preprint)

Federated learning (FL) is a distributed machine learning technique that enables collaborative model training while avoiding explicit data sharing. The inherent privacy-preserving property of FL algorithms makes them especially attractive to the medical field. However, in case of heterogeneous client data distributions, standard FL methods are unstable and require intensive hyperparameter tuning to achieve optimal performance. Conventional hyperparameter optimization algorithms are impractical in real-world FL applications as they involve numerous training trials, which are often not affordable with limited compute budgets. In this work, we propose an efficient reinforcement learning (RL)-based federated hyperparameter optimization algorithm, termed Auto-FedRL, in which an online RL agent can dynamically adjust hyperparameters of each client based on the current training progress. Extensive experiments are conducted to investigate different search strategies and RL agents. The effectiveness of the proposed method is validated on a heterogeneous data split of the CIFAR-10 dataset as well as two real-world medical image segmentation datasets for COVID-19 lesion segmentation in chest CT and pancreas segmentation in abdominal CT.

Auto-FedAvg: Learnable Federated Averaging for Multi-Institutional Medical Image Segmentation (preprint)

Federated learning (FL) enables collaborative model training while preserving each participant's privacy, which is particularly beneficial to the medical field. FedAvg is a standard algorithm that uses fixed weights, often originating from the dataset sizes at each client, to aggregate the distributed learned models on a server during the FL process. However, non-identical data distribution across clients, known as the non-i.i.d problem in FL, could make this assumption for setting fixed aggregation weights sub-optimal. In this work, we design a new data-driven approach, namely Auto-FedAvg, where aggregation weights are dynamically adjusted, depending on data distributions across data silos and the current training progress of the models. We disentangle the parameter set into two parts, local model parameters and global aggregation parameters, and update them iteratively with a communication-efficient algorithm. We first show the validity of our approach by outperforming state-of-the-art FL methods for image recognition on a heterogeneous data split of CIFAR-10. Furthermore, we demonstrate our algorithm's effectiveness on two multi-institutional medical image analysis tasks, i.e., COVID-19 lesion segmentation in chest CT and pancreas segmentation in abdominal CT.

AIforCOVID: predicting the clinical outcomes in patients with COVID-19 applying AI to chest-X-rays. An Italian multicentre study (preprint)

Recent epidemiological data report that worldwide more than 53 million people have been infected by SARS-CoV-2, resulting in 1.3 million deaths. The disease has been spreading very rapidly and few months after the identification of the first infected, shortage of hospital resources quickly became a problem. In this work we investigate whether chest X-ray (CXR) can be used as a possible tool for the early identification of patients at risk of severe outcome, like intensive care or death. CXR is a radiological technique that compared to computed tomography (CT) it is simpler, faster, more widespread and it induces lower radiation dose. We present a dataset including data collected from 820 patients by six Italian hospitals in spring 2020 during the first COVID-19 emergency. The dataset includes CXR images, several clinical attributes and clinical outcomes. We investigate the potential of artificial intelligence to predict the prognosis of such patients, distinguishing between severe and mild cases, thus offering a baseline reference for other researchers and practitioners. To this goal, we present three approaches that use features extracted from CXR images, either handcrafted or automatically by convolutional neuronal networks, which are then integrated with the clinical data. Exhaustive evaluation shows promising performance both in 10-fold and leave-one-centre-out cross-validation, implying that clinical data and images have the potential to provide useful information for the management of patients and hospital resources.

Early Detection Of Deep Vein Thrombosis In Patients With Coronavirus Disease 2019: Who to Screen and Who not to with Doppler Ultrasound? (preprint)

BACKGROUND. Aim of the study is to evaluate the incidence of DVT in COVID-19 patients and its correlation with the severity of the disease and with clinical and laboratory findings.METHODS. 234 symptomatic patients with COVID-19, diagnosed according to the World Health Organization guidelines, were included in the study. The severity of the disease was classified as moderate, severe and critical. Doppler ultrasound (DUS) was performed in all patients. DUS findings, clinical, laboratory’s and therapeutic variables were investigated by contingency tables, Pearson chi square test and by Student T test and Fisher's exact test. ROC curve analysis was applied to study significant continuous variables.RESULTS. Overall incidence of DVT was 10.7% (25/234): 1.6% (1/60) among moderate cases, 13.8% (24/174) in severely and critically ill patients. Prolonged bedrest and intensive care unit admission were significantly associated with the presence of DVT (19.7%). Fraction of inspired oxygen, P/F ratio, respiratory rate, heparin administration, D-dimer, IL-6, ferritin and CRP showed correlation with DVT. CONCLUSIONS. DUS may be considered a useful and valid tool for early identification of DVT. In less severely affected patients, DUS as screening of DVT might be unnecessary. High rate of DVT found in severe patients and its correlation with respiratory parameters and some significant laboratory findings suggests that these can be used as a screening tool for patients who should be getting DUS.

Analysis of the role of Chest X-Ray in a large cohort of 1117 patients with SARS-CoV-2 infection: an Italian multicenter study (preprint)

Objective. To describe the radiographic key patterns on CXR in patients with SARS-CoV-2 infection, assessing the prevalence of radiographic signs of interstitial pneumonia. To evaluate pattern variation between a baseline and a follow-up CXR.Materials and methods. 1117 patients tested positive for SARS-CoV-2 infection were retrospectively enrolled from four centers in Lombardy region. All patients underwent a CXR at presentation. Follow-up CXR was performed when clinically indicated.Two radiologists in each center reviewed CXR images and classified them as suggestive or not for interstitial pneumonia, recording the presence of ground-glass opacity (GGO), reticular pattern or consolidation and their distribution.Pearson’s chi-square test for categorical variables and McNemar test (chi-square for paired data) were performed.Results. Patients mean age 63.3 years, 767 were males (65.5%). The main result is the large proportion of positive CXR in COVID-19 patients.Baseline CXR was positive in 940 patients (80.3%), with significant differences in age and sex distribution between patients with positive and negative CXR. 382 patients underwent a follow-up CXR. The most frequent pattern on baseline CXR was the GGO (66.1%), on follow-up was consolidation (53.4%). The most common distributions were peripheral and middle-lower lung zone.Conclusions. We described key-patterns and their distribution on CXR in a large cohort of COVID-19 patients: GGO was the most frequent finding on baseline CXR, while we found an increase in the proportion of lung consolidation on follow-up CXR. CXR proved to be a reliable tool in our cohort obtaining positive results in 80.3% of the baseline cases.