Federated Learning for Multi-Center Imaging Diagnostics: A Study in Cardiovascular Disease

Deep learning models can enable accurate and efficient disease diagnosis, but have thus far been hampered by the data scarcity present in the medical world. Automated diagnosis studies have been constrained by underpowered single-center datasets, and although some results have shown promise, their generalizability to other institutions remains questionable as the data heterogeneity between institutions is not taken into account. By allowing models to be trained in a distributed manner that preserves patients’ privacy, federated learning promises to alleviate these issues, by enabling diligent multi-center studies. We present the first federated learning study on the modality of cardiovascular magnetic resonance (CMR) and use four centers derived from subsets of the M&M and ACDC datasets, focusing on the diagnosis of hypertrophic cardiomyopathy (HCM). We adapt a 3D-CNN network pretrained on action recognition and explore two different ways of incorporating shape prior information to the model, and four different data augmentation setups , systematically analyzing their impact on the different collaborative learning choices. We show that despite the small size of data (180 subjects derived from four centers), the privacy preserving federated learning achieves promising results that are competitive with traditional centralized learning. We further find that federatively trained models exhibit increased robustness and are more sensitive to domain shift effects.

ST-V-Net: incorporating shape prior into convolutional neural networks for proximal femur segmentation

We aim to develop a deep-learning-based method for automatic proximal femur segmentation in quantitative computed tomography (QCT) images. We proposed a spatial transformation V-Net (ST-V-Net), which contains a V-Net and a spatial transform network (STN) to extract the proximal femur from QCT images. The STN incorporates a shape prior into the segmentation network as a constraint and guidance for model training, which improves model performance and accelerates model convergence. Meanwhile, a multi-stage training strategy is adopted to fine-tune the weights of the ST-V-Net. We performed experiments using a QCT dataset which included 397 QCT subjects. During the experiments for the entire cohort and then for male and female subjects separately, 90% of the subjects were used in ten-fold stratified cross-validation for training and the rest of the subjects were used to evaluate the performance of models. In the entire cohort, the proposed model achieved a Dice similarity coefficient (DSC) of 0.9888, a sensitivity of 0.9966 and a specificity of 0.9988. Compared with V-Net, the Hausdorff distance was reduced from 9.144 to 5.917 mm, and the average surface distance was reduced from 0.012 to 0.009 mm using the proposed ST-V-Net. Quantitative evaluation demonstrated excellent performance of the proposed ST-V-Net for automatic proximal femur segmentation in QCT images. In addition, the proposed ST-V-Net sheds light on incorporating shape prior to segmentation to further improve the model performance.

Task set and instructions influence the weight of figural priors: A psychophysical study with extremal edges and familiar configuration

In figure–ground organization, the figure is defined as a region that is both “shaped” and “nearer.” Here we test whether changes in task set and instructions can alter the outcome of the cross-border competition between figural priors that underlies figure assignment. Extremal edge (EE), a relative distance prior, has been established as a strong figural prior when the task is to report “which side is nearer?” In three experiments using bipartite stimuli, EEs competed and cooperated with familiar configuration, a shape prior for figure assignment in a “which side is shaped?” task.” Experiment 1 showed small but significant effects of familiar configuration for displays sketching upright familiar objects, although “shaped-side” responses were predominantly determined by EEs. In Experiment 2, instructions regarding the possibility of perceiving familiar shapes were added. Now, although EE remained the dominant prior, the figure was perceived on the familiar-configuration side of the border on a significantly larger percentage of trials across all display types. In Experiment 3, both task set (nearer/shaped) and the presence versus absence of instructions emphasizing that familiar objects might be present were manipulated within subjects. With familiarity thus “primed,” effects of task set emerged when EE and familiar configuration favored opposite sides as figure. Thus, changing instructions can modulate the weighing of figural priors for shape versus distance in figure assignment in a manner that interacts with task set. Moreover, we show that the influence of familiar parts emerges in participants without medial temporal lobe/ perirhinal cortex brain damage when instructions emphasize that familiar objects might be present.