Deep residual networks for crystallography trained on synthetic data

The use of artificial intelligence to process diffraction images is challenged by the need to assemble large and precisely designed training data sets. To address this, a codebase called Resonet was developed for synthesizing diffraction data and training residual neural networks on these data. Here, two per-pattern capabilities of Resonet are demonstrated: (i) interpretation of crystal resolution and (ii) identification of overlapping lattices. Resonet was tested across a compilation of diffraction images from synchrotron experiments and X-ray free-electron laser experiments. Crucially, these models readily execute on graphics processing units and can thus significantly outperform conventional algorithms. While Resonet is currently utilized to provide real-time feedback for macromolecular crystallography users at the Stanford Synchrotron Radiation Lightsource, its simple Python-based interface makes it easy to embed in other processing frameworks. This work highlights the utility of physics-based simulation for training deep neural networks and lays the groundwork for the development of additional models to enhance diffraction collection and analysis.

Use of the intra-aortic balloon pump after valve replacement. Predictive indices, correlative parameters, and patient survival

Intra-aortic balloon pump counterpulsation required as an adjunct during weaning from cardiopulmonary bypass or for circulatory support in the immediate postoperative period was analyzed in 2,498 patients undergoing valve replacement between December, 1972, and September, 1981. A total of 140 successful insertions were performed in 155 attempts. Ninety-five of these patients were from a homogeneous cohort of 1,908 patients undergoing valve replacement with porcine xenografts and were analyzed for factors that might be useful predictors of the need for balloon pump support. Univariate analysis of individual factors delineated preoperative characteristics in patients having mitral valve replacement and intraoperative factors in all patients that correlated with use of the balloon pump. Multivariate analysis revealed a subset of male patients with mitral valve and coronary disease most likely to require counterpulsation. Overall survival rate was markedly reduced at 30 days (balloon counterpulsation plus valve replacement, 50% +/- 5%; valve replacement only, 96% +/- 5%; p less than 0.001) and at 1 year (balloon counterpulsation plus valve replacement, 38% +/- 5%; valve replacement only, 89% +/- 1%, p less than 0.001) if balloon pumping was required. The entire group of 140 patients were retrospectively analyzed for factors predictive of survival. Patients requiring balloon pumping who had a preoperative diagnosis of aortic regurgitation had a lower 1 year survival rate (13% +/- 9%) than the total subgroup undergoing balloon counterpulsation (36% +/- 4.0%) (p = 0.002). Similarly patients treated by balloon counterpulsation who had postoperative renal failure had a significantly lower 1 year survival rate (17% +/- 5%) than those without renal failure (66% +/- 6%) (p = 0.003). The survival rate of patients who required this therapeutic modality after valve replacement is poor. Other methods of hemodynamic support are necessary.