The Jostent coronary stent graft (CSG) is composed of a PTFE layer sandwiched between two stainless steel stents, initially introduced for the treatment of coronary perforations and aneurysms with excellent results. By providing a mechanical barrier, this stent design also may be beneficial in the treatment of complex ulcerated lesions and in-stent restenosis by preventing debris protrusion and neointimal proliferation through the stent struts. To evaluate the safety and efficacy of this stent graft, we implanted 78 CSGs in 70 patients for a broad range of indications, including coronary perforations, aneurysms, degenerated saphenous vein grafts, complex lesions, and in-stent restenosis. The primary angiographic success rate (95.9%) was high, and using intravascular ultrasound (IVUS) guidance during stent implantation and high inflation pressures (19.3 +/- 3.2 atm), stent expansion with optimal symmetry was achieved in 94.7%. One limitation of the Jostent CSG was the side-branch occlusion rate (18.6%) and the resulting non-Q-wave infarction rate in seven cases (mean CK elevation, 238 U/l), acute Q-wave MI in two cases, and transient ventricular fibrillation in one patient after occlusion of the proximal RCA side branch without further complications. Subacute stent thrombosis occurred in four cases (5.7%) 7 to 70 days after stent implantation, despite using combined antiplatelet therapy with aspirin (ASA), ticlopidine, and/or clopidogrel for 30 days. Angiographic follow-up was available in 56 patients (80.0%) after a mean of 159 +/- 49 days, and follow-up IVUS was available in 38 cases. The overall restenosis rate (> 50% diameter stenosis) was 31.6% manifest primarily as edge restenosis (29.8% stent edge vs. 8.8% stent center; P < 0.001). IVUS examinations showed a minimal late lumen loss of 0.4 +/- 2.2 mm(2) within the center of the stent graft vs. 3.2 +/- 2.3 mm(2) at the stent edges (P < 0.001). The restenosis rate in the prespecified subgroups was 33.3% for saphenous vein grafts (2/6 lesions), 30.0% in complex lesions (6/20 lesions), and 38.5% (10/26 lesions) for the treatment of in-stent restenosis. Implantation of the Jostent CSG is feasible and safe, even in complex lesion subsets, and is associated with high primary success rates provided major side branches are avoided. The use of this stent may require an extended time course of antiplatelet therapy. Frequent focal stent edge renarrowing influences the overall restenosis rate. However, in treatment of complex in-stent restenosis and vein graft lesions, stent grafts may offer benefit over conventional therapies. Covered stents such as the JoMed coronary stent graft may become essential for bailout treatment of coronary perforations.

In WRIST, not only was (192)Ir therapy effective in reducing restenosis, but it also reduced the lesion length of treatment failures by 50%, and it was not associated with edge proliferation. The restenosis rate obtained from the vessel segment inclusive of the dose fall-off zones was the best correlate of TVR and should become a standard analysis site in all vascular brachytherapy trials.

The de-facto approach to many vision tasks is to start from pretrained visual representations, typically learned via supervised training on ImageNet. Recent methods have explored unsupervised pretraining to scale to vast quantities of unlabeled images. In contrast, we aim to learn high-quality visual representations from fewer images. To this end we revisit supervised pretraining, and seek data-efficient alternatives to classification-based pretraining. We propose VirTex -a pretraining approach using semantically dense captions to learn visual representations. We train convolutional networks from scratch on COCO Captions, and transfer them to downstream recognition tasks including image classification, object detection, and instance segmentation. On all tasks, VirTex yields features that match or exceed those learned on ImageNet -supervised or unsupervised -despite using up to ten times fewer images.

Patients undergoing CAWR with BMIs greater than 30 kg/m have significantly higher rates of skin necrosis, hernia recurrence, and reoperation compared with subgroups of lower BMI. Rates of overall complications among all BMI groups are similar, although trended up with BMI. Surgeons should weight the risks and benefits of CAWR in patients with higher BMIs to reduce specific postoperative complications.

Type of reconstruction, surgeon, presence of complications, and need for postoperative chemotherapy or radiation therapy all affect timing to completion of breast reconstruction. Patients should be counseled as to these factors at the initial consultation to set appropriate expectations.

Percutaneous coronary intervention (PCI) devices are much more expensive in Japan than in the United States, but their prices were reduced in April 2002. This study evaluated the impact of that change in the price of PCI devices on medical expenses. In-hospital costs of 22 consecutive patients who underwent elective single-vessel PCI without a debulking procedure before April 2002 were collected and the in-hospital cost of each patient was recalculated by applying the current prices of the PCI devices and those in the USA. For patients treated with PCI before April 2002, the in-hospital cost was 1,456,375+/-358,781 yen, but when the current price is used, the in-hospital cost is estimated to be 1,355,812 +/-313,237 yen (7% reduction). If the prices of the devices were reduced to those in USA, there would be a 53% reduction (689,417 +/-99,139 yen). Although the change in the price of PCI devices in April 2002 has reduced in-hospital costs, the devices are still much more expensive in Japan than in the USA. Further reduction of the price is required to make PCI more cost-effective.

No abstract

High-dimensional always-changing environments constitute a hard challenge for current reinforcement learning techniques. Artificial agents, nowadays, are often trained off-line in very static and controlled conditions in simulation such that training observations can be thought as sampled i.i.d. from the entire observations space. However, in real world settings, the environment is often non-stationary and subject to unpredictable, frequent changes. In this paper we propose and openly release CRLMaze, a new benchmark for learning continually through reinforcement in a complex 3D non-stationary task based on ViZDoom and subject to several environmental changes. Then, we introduce an end-to-end model-free continual reinforcement learning strategy showing competitive results with respect to four different baselines and not requiring any access to additional supervised signals, previously encountered environmental conditions or observations. Preprint. Under review.