In recent years, 3D convolutional neural networks have become the dominant approach for volumetric medical image segmentation. However, compared to their 2D counterparts, 3D networks introduce substantially more training parameters and higher requirement for the GPU memory. This has become a major limiting factor for designing and training 3D networks for high-resolution volumetric images. In this work, we propose a novel memory-efficient network architecture for 3D high-resolution image segmentation. The network incorporates both global and local features via a two-stage U-net-based cascaded framework and at the first stage, a memory-efficient U-net (meU-net) is developed. The features learnt at the two stages are connected via post-concatenation, which further improves the information flow. The proposed segmentation method is evaluated on an ultra high-resolution microCT dataset with typically 250 million voxels per volume. Experiments show that it outperforms state-of-the-art 3D segmentation methods in terms of both segmentation accuracy and memory efficiency.

In 2005 there was an outbreak of XDR (extensively drug resistant) TB in Tugela Ferry, which is served by the Church of Scotland Hospital (COSH), in the uMzinyathi District, KwaZulu-Natal, South Africa. An investigation was carried out to determine if XDR TB was occurring elsewhere in the province, and to develop hypotheses for the rise in drug resistance with a view to developing a strategy for the control of MDR (multi-drug resistant) and XDR TB in the province and elsewhere. TB incidence and treatment success rates, for each of the 11 districts in the province, were obtained from the provincial electronic TB register for the years 2002-2007. The results of culture and drug sensitivity tests for the years 2002 to 2007 in each of the districts were compiled and culture taking practices were compared to the number of MDR TB cases. Interviews were conducted with key personnel in affected sites. In 2007, 2799, or 2.3% of 119,218 notified TB cases in the province were multi-drug resistant (MDR), and of these 270 (9.6%) were XDR. The two worst affected districts were uMzinyathi where 226 (4.1%) of 5522 notified TB cases were MDR, and of these 120 (53%) were extensively drug resistant (XDR), and uMkhanyakude where 337 (4.8%) of 6991 notified TB cases were MDR, but of these only four or (1.2%) were XDR. The worst affected medical centre was COSH where 164 or 9.8% of notified TB cases were MDR and of these 99 (60%) were XDR. Very high rates of XDR TB in the province are only found in uMzinyathi district even though MDR TB is common in most other districts. XDR may arisen at COSH because of the early and effective integration of the TB and HIV programmes in overcrowded and poorly ventilated facilities particular to COS.H To control XDR TB better management of both susceptible and resistant forms of TB is needed including treatment supervision, infection control and HIV management.

Using administrative patient-care data such as Electronic Health Records (EHR) and medical/ pharmaceutical claims for population-based scientific research has become increasingly common. With vast sample sizes leading to very small standard errors, researchers need to pay more attention to potential biases in the estimates of association parameters of interest, specifically to biases that do not diminish with increasing sample size. Of these multiple sources of biases, in this paper, we focus on understanding selection bias. We present an analytic framework using directed acyclic graphs for guiding applied researchers to dissect how different sources of selection bias may affect estimates of the association between a binary outcome and an exposure (continuous or categorical) of interest. We consider four easy-to-implement weighting approaches to reduce selection bias with accompanying variance formulae. We demonstrate through a simulation study when they can rescue us in practice with analysis of real world data. We compare these methods using a data example where our goal is to estimate the well-known association of cancer and biological sex, using EHR from a longitudinal biorepository at the University of Michigan Healthcare system. We provide annotated R codes to implement these weighted methods with associated inference.