Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several common benchmarks.

We present a modeling framework designed for patient-specific computational hemodynamics to be performed in the context of large-scale studies. The framework takes advantage of the integration of image processing, geometric analysis and mesh generation techniques, with an accent on full automation and high-level interaction. Image segmentation is performed using implicit deformable models taking advantage of a novel approach for selective initialization of vascular branches, as well as of a strategy for the segmentation of small vessels. A robust definition of centerlines provides objective geometric criteria for the automation of surface editing and mesh generation. The framework is available as part of an open-source effort, the Vascular Modeling Toolkit, a first step towards the sharing of tools and data which will be necessary for computational hemodynamics to play a role in evidence-based medicine.

Background and Purpose-That certain vessels might be at so-called geometric risk of atherosclerosis rests on assumptions of wide interindividual variations in disturbed flow and of a direct relationship between disturbed flow and lumen geometry. In testing these often-implicit assumptions, the present study aimed to determine whether investigations of local risk factors in atherosclerosis can indeed rely on surrogate geometric markers of disturbed flow. Methods-Computational fluid dynamics simulations were performed on carotid bifurcation geometries derived from MRI of 25 young adults. Disturbed flow was quantified as the surface area exposed to low and oscillatory shear beyond objectively-defined thresholds. Interindividual variations in disturbed flow were contextualized with respect to effects of uncertainties in imaging and geometric reconstruction. Relationships between disturbed flow and various geometric factors were tested via multiple regression. Results-Relatively wide variations in disturbed flow were observed among the 50 vessels. Multiple regression revealed a significant (PϽ0.002) relationship between disturbed flow and both proximal area ratio (␤Ϸ0.5) and bifurcation tortuosity (␤ϷϪ0.4), but not bifurcation angle, planarity, or distal area ratio. These findings were shown to be insensitive to assumptions about the flow conditions and to the choice of disturbed flow indicator and threshold. Conclusions-Certain geometric features of the young adult carotid bifurcation are robust surrogate markers of its exposure to disturbed flow. It may therefore be reasonable to consider large-scale retrospective or prospective imaging studies of local risk factors for atherosclerosis without the need for time-consuming and expensive flow imaging or CFD studies.

In ADPKD patients, 6-month somatostatin therapy is safe and may slow renal volume expansion. This may reflect an inhibited growth in particular of smallest cysts beyond the detection threshold of CT scan evaluation. Whether this effect may prove renoprotective in the long term should be tested in additional trials of longer duration.