No abstract

CHD risk in acromegalic patients, predicted by FS as in nonacromegalic subjects, is low; AS might have adjunctive role only in a subset of patients. However, most patients have systemic complications of acromegaly, which participate in the assessment of global CHD risk.

We compared whole-lung densitometry with visual evaluation of pulmonary emphysema. Thirty patients with chronic obstructive pulmonary disease underwent multi-detector CT (150 mAs and 0.75 collimation) with double reconstruction: thick (5-mm) slices with smooth filter for whole-lung densitometry and thin (1 mm) slices with sharp filter for visual assessment (one of every ten slices). Densitometry and visual assessment were performed by three operators each, and the time required for assessment, the inter-observer agreement and the correlation with the results of the diffusion capacity of carbon monoxide (DL(CO)) in the same patients were computed. The average time for densitometry (8.49 +/- 0.13 min) was significantly longer (p < 0.0001) than that for visual evaluation (5.14 +/- 0.11 min). However, the inter-operator agreement ranged between "moderate" to "almost perfect" for densitometry (kappa range 0.58-0.87) and "slight" for visual (kappa = 0.20) assessment. The correlation coefficients of DL(CO) with relative area at -960 and -970 Hounsfield units (HU) (both r = -0.66) and of the first percentile point of lung density (r = 0.66) were slightly stronger than that of the visual score (r = -0.62). Densitometry should be preferred to visual assessment because it enables a more reproducible evaluation of the extent of pulmonary emphysema, which can be carried out on the entire lung in a reasonable amount of time.

Universidad Autónoma de Madrid, SpainKeywords Mammogram Registration Á Temporal study Á Consistent registration Á Diffeomorphism Á Hormonal replacement therapy Purpose Mammographic density is a strong risk factor for breast cancer. However, whether changes in mammographic density due to HRT are associated with risk remains unclear. The aim of this study is to provide a framework for accurate interval change analysis in temporal pairs of mammograms of patients undergoing HRT treatment using consistent and fully automatic registration framework. MethodsThe combination of B-spline based registration and consistent image registration is proposed that simplifies the search for optimal deformation and does not require specification of landmarks. KullbackLieber distance is used as a similarity metric which makes this method more robust against different image resolutions and other intensity based artifacts during temporal scan of mammograms. ResultsRegistration was performed on the pairs of mammograms from longitudinal double blind placebo controlled HRT study that includes 39 placebo and 36 HRT treated volunteers for two years. The method's performance was evaluated by an experienced radiologist. Validation study showed that method has given an optimal consistency error of 36 pixels (mm) with average computation time 45 s (Figs. 1, 2). ConclusionsThe work describes a novel approach for consistent and robust registration framework for interval change analysis in pairs of mammograms. This would lead radiologist to asses the risk of individual patient by examining the changes in tissue structure and density due to hormonal replacement therapy.Reducing false positive marks in breast mass computer-aided detection via bilateral ranklet texture analysis Keywords Mammography Á Computer-aided detection Á Ranklets Á Support vector machine Á Bilateral analysis Purpose In order to reduce subtle false positive marks, our current breast mass computer-aided detection (CAD) system makes use of a ranklet texture false positive reduction (RT-FPR, hereafter) module, where regions of interest (ROIs) are encoded by means of a set of ranklet texture features and classified by means of a support vector machine (SVM) classifier.In an attempt to reduce even further the number of false positive marks, we herein propose an alternative, bilateral RT-FPR module (B-RT-FPR, hereafter): ROIs found in either one, left or right projection of the breast are encoded and fed into the SVM classifier as vectors containing their corresponding ranklet texture features, as well as those corresponding to the regions found symmetrically on the same projection of the contralateral mammogram.In this work, we present the B-RT-FPR module, then compare the CAD system's overall performance when combined with either the RT-FPR and B-RT-FPR modules. Methods In our CAD system, ROIs surviving the detection and first-level false positive reduction (FPR) modules appear quite subtle, so the associated classification task results particularly tough. Currently, such mass candi...

No abstract