PurposeDeep learning (DL) techniques are widely used in medical imaging and in particular for segmentation. Indeed, manual segmentation of organs at risk (OARs) is time-consuming and suffers from inter- and intra-observer segmentation variability. Image segmentation using DL has given very promising results. In this work, we present and compare the results of segmentation of OARs and a clinical target volume (CTV) in thoracic CT images using three DL models.Materials and methodsWe used CT images of 52 patients with breast cancer from a public dataset. Automatic segmentation of the lungs, the heart and a CTV was performed using three models based on the U-Net architecture. Three metrics were used to quantify and compare the segmentation results obtained with these models: the Dice similarity coefficient (DSC), the Jaccard coefficient (J) and the Hausdorff distance (HD).ResultsThe obtained values of DSC, J and HD were presented for each segmented organ and for the three models. Examples of automatic segmentation were presented and compared to the corresponding ground truth delineations. Our values were also compared to recent results obtained by other authors.ConclusionThe performance of three DL models was evaluated for the delineation of the lungs, the heart and a CTV. This study showed clearly that these 2D models based on the U-Net architecture can be used to delineate organs in CT images with a good performance compared to other models. Generally, the three models present similar performances. Using a dataset with more CT images, the three models should give better results. 相似文献
Noninvasive imaging of cardiac fibrosis is important for early diagnosis and intervention in chronic heart diseases. Here, we investigated whether noninvasive, contrast agent-free MRI T2-mapping can quantify myocardial fibrosis in preclinical models of aging and pressure overload. Myocardial fibrosis and remodeling were analyzed in two animal models: (i) aging (15-month-old male CF-1 mice vs. young 6- to 8-week-old mice), and (ii) pressure overload (PO; by transverse aortic constriction in 4- to 5-month-old male C57BL/6 mice vs. sham-operated for 14 days). In vivo T2-mapping was performed by acquiring data during the isovolumic and early diastolic phases, with a modified respiratory and ECG-triggered multiecho TurboRARE sequence on a 7-T MRI. Cine MRI provided cardiac morphology and function. A quantitative segmentation method was developed to analyze the in vivo T2-maps of hearts at midventricle, apex, and basal regions. The cardiac fibrosis area was analyzed ex vivo by picro sirius red (PSR) staining. Both aged and pressure-overloaded hearts developed significant myocardial contractile dysfunction, cardiac hypertrophy, and interstitial fibrosis. The aged mice had two phenotypes, fibrotic and mild-fibrotic. Notably, the aged fibrotic subgroup and the PO mice showed a marked decrease in T2 relaxation times (25.3 ± 0.6 in aged vs. 29.9 ± 0.7 ms in young mice, p = 0.002; and 24.3 ± 1.7 in PO vs. 28.7 ± 0.7 ms in shams, p = 0.05). However, no significant difference in T2 was detected between the aged mild-fibrotic subgroup and the young mice. Accordingly, an inverse correlation between myocardial fibrosis percentage (FP) and T2 relaxation time was derived (R2 = 0.98): T2 (ms) = 30.45 – 1.05 × FP. Thus, these results demonstrate a statistical agreement between T2-map–quantified fibrosis and PSR staining in two different clinically relevant animal models. In conclusion, T2-mapping MRI is a promising noninvasive contrast agent-free quantitative technique to characterize myocardial fibrosis. 相似文献
ObjectivesTo examine iron stores, hemoglobin mass, and performance before, during and after intermittent altitude exposure in a professional male rugby player experiencing iron overload following blood transfusions for treatment for acute myeloid leukemia.DesignLongitudinal, repeated measures, single case-study.MethodsThe player was followed prior to (control), and during (study), an in-season block of altitude training. During the control period two venesections were performed for a total of 750 mL of blood removal. Internal and external training load, match statistics, blood volume, plasma volume, haemoglobin mass, serum ferritin and reticulocyte count were monitored throughout.ResultsDuring the control period serum ferritin declined following the two venesections (∼51%) as did haemoglobin mass (∼2%), reticulocyte count remained stable. During the study period serum ferritin further declined (∼30%), however haemoglobin mass and reticulocyte count increased (∼4% and ∼14% respectively). Internal training load for the control and study period was similar, however external training load was lower in the study period. Match statistics were not favourable for the player during the control period, however they improved during the study period.ConclusionsThis case supports the theory that individuals with elevated iron availability are well placed to achieve increases in haemoglobin mass. Furthermore, although therapeutic venesections may still be required to manage iron overload, the addition of altitude exposure may be a method to assist in reducing total body iron by means of mobilising available (excessive) iron to incorporate into haemoglobin. Altitude exposure did not hinder the players’ performance. Further research is encouraged. 相似文献
BackgroundThe Agatston Calcium Score is a predictor of major adverse cardiovascular events but it is unable to identify high-risk lesions. Recent research suggests that examining calcification phenotype could be more indicative of plaque stability.ObjectiveTo examine the Agatston score's ability to determine atherosclerotic calcification phenotype.MethodsMicro-Computed Tomography was performed on 20 carotid and 20 peripheral lower limb lesions. ImageJ pixel histogram analysis quantified the non-calcified (≥30HU, <130HU) and calcified (≥130HU) tissue volumes. ImageJ ‘3D Objects Counter’ plugin determined the calcified particle count, volumes and maximum attenuation density of each particle. Image stacks were subsequently downsampled to a resolution of 0.7 × 0.7 × 3 mm and an approximation for the Extra-Coronary Calcium Scores (ECCS) were calculated. Spearman's correlation examined the relationships between ECCS approximations and calcification parameters.ResultsECCS has a strong positive correlation with the Calcified Volume Fraction (CVF) (rs = 0.865, p < 0.0005), weak positive correlations with Calcified Particle Fraction (CPF) (rs = 0.422, p = 0.007) and Microcalcification Fraction (micro-CF) (rs = 0.361, p = 0.022). There is no correlation evident between ECCS and Calcified Particle Index (CPI) (rs = −0.162, p = 0.318). It is apparent that there is a high prevalence of microcalcifications in both carotid and peripheral lower limb lesions. Additionally, an inverse relationship exists between calcified particle volume and maximum-recorded attenuation density.ConclusionThe density-weighted Agatston calcium scoring methodology needs to be reviewed. Calcium scoring which differentiates between critical calcification morphologies, rather than presenting a density-weighted score, is required to direct high-risk plaques towards tailored treatment. 相似文献