No abstract

A study of 57 patients admitted to the Department of Emergency Surgery at the Ospedale Maggiore in Milan between 1980 and 1992 following the recent ingestion of a caustic substance is presented herein. Through this study, an aggressive diagnostic and therapeutic approach has been employed, including early surgery which plays a fundamental role in the prevention of acute hemorrhagic or perforative complications as well as in the development of scar tissue and neoplastic strictures over time. The criteria for early emergency surgery were the presence of endoscopic grade 3 and 4 lesions as well as those on the borderline between grades 2 and 3 with clinical symptoms. In 11 patients with lesions of moderate severity, the treatment of choice was medical therapy, which required subsequent surgical intervention for strictures in 5 patients. In 13 patients with severe lesions, an early surgical approach was performed with a mortality rate of 23%.

No abstract

Gangliocytic paragangliomas (GPs) are exceedingly rare tumors that arise in close proximity of the papilla of Vater. Nevertheless, jaundice is an uncommon presenting symptom, reported in only 3 of 125 cases described in the literature to date, with gastrointestinal bleeding being more common. Association between GPs and neurofibromatosis 1 (NF-1), described in two patients, may be more than casual, being the frequency in normal population 1:3,000. We report an additional case of this association, presenting for the first time with simultaneous obstruction of the common biliary duct, and of the main pancreatic duct. Previous cases of GPs with jaundice and/or NF-1 are reviewed, and updating of the histogenesis presented.

To retrospectively evaluate the frequency and type of findings that were missed in the original reports of multi-detector CT angiography (MDCTA) in patients with suspected acute bowel ischemia. From January 2007 to March 2011, a series of 35 patients who underwent MDCTA of the abdomen and pelvis and had surgery were included. The reports of the initial CT were retrospectively compared with the discharge diagnosis and surgical reports. Discrepant or missing findings were re-evaluated and divided into relevant or not relevant regarding the diagnosis. In 23 of the 35 patients (66 %), all findings were correctly diagnosed in the initial MDCTA report. In the remaining 12 of the 35 patients (34 %), lesions that were not reported were present at surgery. In 10 of the 12 (83 %) patients, the overlooked findings were relevant and subtle: gas in the portal vein (n = 3), gas in the bowel wall (n = 3), gas in the portal vein and bowel wall (n = 2), thrombotic occlusion of the superior mesenteric artery (n = 1), and thrombotic occlusion of the inferior mesenteric artery (n = 1). In 2 of the 12 (17 %) patients in whom the MDCTA-overlooked findings were classified as non-relevant, bowel ischemia was found at surgery. With retrospective image interpretation, 83 % of the patients with occlusive mesenteric ischemia at surgery were correctly identified, whereas the remaining 17 % with non-occlusive mesenteric ischemia at surgery showed non-relevant findings at MDCTA. About 33 % of relevant findings of bowel ischemia were overlooked by the initial MDCTA interpretation, most were subtle findings. However, secondary reading revealed most of these findings and can serve to improve diagnostic performance.

In our experience, a long time variability for spontaneous liver repair after blunt trauma and non-operative treatment was found, but a parenchymal US normalization was evidenced in a median time shorter than that usually reported in the literature.

No abstract

Due to their prevalence, time series forecasting is crucial in multiple domains. We seek to make state-of-the-art forecasting fast, accessible, and generalizable. ES-RNN is a hybrid between classical state space forecasting models and modern RNNs that achieved a 9.4% sMAPE improvement in the M4 competition. Crucially, ES-RNN implementation requires per-time series parameters. By vectorizing the original implementation and porting the algorithm to a GPU, we achieve up to 322x training speedup depending on batch size with similar results as those reported in the original submission. Our code can be found at: https://github.com/damitkwr/ESRNN-GPU