R2D2: A scalable deep learning toolkit for medical imaging segmentation

Deep learning has gained a significant popularity in recent years thanks to its tremendous success across a wide range of relevant fields of applications, including medical image analysis domain in particular. Although convolutional neural networks (CNNs) based medical applications have been providing powerful solutions and revolutionizing medicine, efficiently training of CNNs models is a tedious and challenging task. It is a computationally intensive process taking long time and rare system resources, which represents a significant hindrance to scientific research progress. In order to address this challenge, we propose in this article, R2D2, a scalable intuitive deep learning toolkit for medical imaging semantic segmentation. To the best of our knowledge, the present work is the first that aims to tackle this issue by offering a novel distributed versions of two well-known and widely used CNN segmentation architectures [ie, fully convolutional network (FCN) and U-Net]. We introduce the design and the core building blocks of R2D2. We further present and analyze its experimental evaluation results on two different concrete medical imaging segmentation use cases. R2D2 achieves up to 17.5× and 10.4× speedup than single-node based training of U-Net and FCN, respectively, with a negligible, though still unexpected segmentation accuracy loss. R2D2 offers not only an empirical evidence and investigates in-depth the latest published works but also it facilitates and significantly reduces the effort required by researchers to quickly prototype and easily discover cutting-edge CNN configurations and architectures.     

Comparative genomic hybridization analysis of lobular carcinoma in situ and atypical lobular hyperplasia and potential roles for gains and losses of genetic material in breast neoplasia

Lobular carcinoma in situ (LCIS) and atypical lobular hyperplasia (ALH) of the breast are cytologically similar breast lesions that reportedly carry different relative risks of subsequent development of invasive carcinoma. They are frequently multifocal and bilateral. We have identified the chromosomal copy number changes in 31 LCIS and 14 ALH lesions from 28 cases and also the 7 invasive carcinomas that subsequently developed in 6 of these cases. This was achieved by comparative genomic hybridization analysis of microdissected formalin-fixed, paraffin-embedded material. There was no significant difference between the aberrations found in the unilateral versus the bilateral cases of LCIS. Loss of material from 16p, 16q, 17p, and 22q and also gain of material from 6q were found at a similar high frequency in LCIS and ALH. Loss of these genomic regions may indicate the locations of genes that predispose to the development of the lesions, and the results are consistent with LCIS and ALH representing the same genetic stage of development. Comparison of the comparative genomic hybridization results from LCIS/ALH with those from ductal carcinoma in situ and invasive cancer showed some similarities at the chromosomal level, but it also showed significant differences, including gain of 1q and 8q and evidence for genomic amplification, which were not found in LCIS/ALH. A genetic model is postulated for the possible relationships between noninvasive lobular lesions and invasive breast carcinoma, delineating potential roles for specific chromosome copy number changes.     

Unusual breast lumps

Two new Italian cases of breast infection by Dirofilaria Repens are presented. In one case the correct diagnosis was clinically achieved through needle aspiration. This case is documented by a surprising macrophotography.     

Test-execution-based reliability measurement and modeling for largecommercial software

The paper studies practical reliability measurement and modeling for large commercial software systems based on test execution data collected during system testing. The application environment and the goals of reliability assessment were analyzed to identify appropriate measurement data. Various reliability growth models were used on failure data normalized by test case executions to track testing progress and provide reliability assessment. Practical problems in data collection, reliability measurement and modeling, and modeling result analysis were also examined. The results demonstrated the feasibility of reliability measurement in a large commercial software development environment and provided a practical comparison of various reliability measurements and models under such an environment     

A multiscale expectation-maximization semisupervised classifier suitable for badly posed image classification.

This paper deals with the problem of badly posed image classification. Although underestimated in practice, bad-posedness is likely to affect many real-world image classification tasks, where reference samples are difficult to collect (e.g., in remote sensing (RS) image mapping) and/or spatial autocorrelation is relevant. In an image classification context affected by a lack of reference samples, an original inductive learning multiscale image classifier, termed multiscale semisupervised expectation maximization (MSEM), is proposed. The rationale behind MSEM is to combine useful complementary properties of two alternative data mapping procedures recently published outside of image processing literature, namely, the multiscale modified Pappas adaptive clustering (MPAC) algorithm and the sample-based semisupervised expectation maximization (SEM) classifier. To demonstrate its potential utility, MSEM is compared against nonstandard classifiers, such as MPAC, SEM and the single-scale contextual SEM (CSEM) classifier, besides against well-known standard classifiers in two RS image classification problems featuring few reference samples and modestly useful texture information. These experiments yield weak (subjective) but numerous quantitative map quality indexes that are consistent with both theoretical considerations and qualitative evaluations by expert photointerpreters. According to these quantitative results, MSEM is competitive in terms of overall image mapping performance at the cost of a computational overhead three to six times superior to that of its most interesting rival, SEM. More in general, our experiments confirm that, even if they rely on heavy class-conditional normal distribution assumptions that may not be true in many real-world problems (e.g., in highly textured images), semisupervised classifiers based on the iterative expectation maximization Gaussian mixture model solution can be very powerful in practice when: 1) there is a lack of reference samples with respect to the problem/model complexity and 2) texture information is considered negligible (i.e., a piecewise constant image model holds).     

Clinico-radiological correlation in thalassemia intermedia

Fifteen thalassemia intermedia patients were considered, whose clinical and radiological findings were examined and compared. Eight patients underwent regular transfusion therapy. All patients underwent total body CT: the volume of ectopic erythropoiesis foci was calculated by a digital calculation algorithm (ROI volume). This work was aimed at correlating the quantitative measures of ectopic erythropoiesis assessed by CT with serum level of erythropoietin (EPO) and of trasferrin-free receptors (TfR) in both transfused and non-transfused patients, also considering the volume changes of ectopic erythropoiesis and bone changes over 36 months' follow-up. A direct correlation was demonstrated between serum transferrin and ectopic erythropoietic masses in transfusion-dependent patients: in fact, increased values of serum transferrin correspond to the enlargement of these masses and to bone lesion worsening.     

Progesterone metabolites and bile acids in serum of patients with intrahepatic cholestasis of pregnancy: effect of ursodeoxycholic acid therapy

The concentrations in serum of sulfated metabolites of progesterone are known to be elevated in patients with intrahepatic cholestasis of pregnancy (ICP). The profiles of these metabolites and conjugated bile acids were analyzed in serum from 11 patients with ICP before and during administration of ursodeoxycholic acid (UDCA) (8 patients) or placebo (3 patients). The clinical condition of 7 of the patients given UDCA improved markedly, and 1 patient given placebo had a spontaneous remission of the disease. The total concentration of conjugated bile acids in the 11 patients was 25 +/- 6 micromol/L (mean +/- SEM) and decreased to 6.3 +/- 3.5 micromol/L in the 7 patients responding to treatment with UDCA. The level of 7alpha-hydroxy-4-cholesten-3-one was significantly lower (7.2 +/- 2.2 ng/mL) in patients with ICP than in healthy pregnancy (18 +/- 4.6 ng/mL) (P < .05). The concentrations of 5alpha-pregnane-3alpha,20alpha-diol mono- and disulfates decreased by 52% +/- 7.9% and 68% +/- 5.5%, respectively, in the patients responding to treatment. Similar decreases were observed for the mono- and disulfates of 5alpha-pregnane-3alpha,20alpha,21-triol and 5beta-pregnane-3alpha,20alpha-diol. The disulfate of 5alpha-pregnane-3beta,20alpha-diol showed a smaller decrease, while glucuronidated steroids were not affected. The 3alpha-/3beta-hydroxysteroid ratio and di-/monosulfate ratio decreased significantly during UDCA. The magnitudes of the changes of bile acid and steroid concentrations during UDCA were not correlated to each other. The results suggest that UDCA stimulates the biliary excretion of steroids with a 3alpha-sulfoxy group and disulfates. This effect seems to be independent of the effect on bile acid excretion, indicating the use of different transport proteins. The possibility of an effect of UDCA on the formation of the steroid sulfates cannot be excluded.