The study on the segmentation of carotid vessel wall in multicontrast MR images based on U-Net neural network / 中华放射学杂志

Objective@#To investigate the value of automatic segmentation of carotid vessel wall in multicontrast MR images using U-Net neural network.@*Methods@#Patients were retrospectively collected from 2012 to 2015 in Carotid Atherosclerosis Risk Assessment (CARE II) study. All patients who recently suffered ischemic stroke and/or transient ischemic attack underwent identical, state-of-the-art multicontrast MRI technique. A total of 17 568 carotid vessel wall MR images from 658 subjects were included in this study after inclusion criteria and exclusion criteria. All MR images were analyzed using customized analysis platform (CASCADE). Randomly, 10 592 images were assigned into training dataset, 3 488 images were assigned into validating dataset and 3 488 images were assigned into test dataset according to a ratio of 6∶2∶2. Data augmentation was performed to avoid over fitting and improve the ability of model generalization. The fine-tuned U-Net model was utilized in the segmentation of carotid vessel wall in multicontrast MR images. The U-Net model was trained in the training dataset and validated in the validating dataset. To evaluate the accuracy of carotid vessel wall segmentation, the sensitivity, specificity and Dice coefficient were used in the testing dataset. In addition, the interclass correlation and the Bland-Altman analysis of max wall thickness and wall area were obtained to demonstrate the agreement of the U-Net segmentation and the manual segmentation.@*Results@#The sensitivity, specificity and Dice coefficient of the fine-tuned U-Net model achieved 0.878,0.986 and 0.858 in the test dataset, respectively. The interclass correlation (95% confidence interval) was 0.921 (0.915-0.925) for max wall thickness and 0.929 (0.924-0.933) for wall area. In the Bland-Altman analysis, the difference of max wall thickness was (0.037±0.316) mm and the difference of wall area was (1.182±4.953) mm2. The substantial agreement was observed between U-Net segmentation method and manual segmentation method.@*Conclusion@#Automatic segmentation of carotid vessel wall in multicontrast MR images can be achieved using fine-tuned U-Net neural network, which is trained and tested in the large scale dataset labeled by professional radiologists.

The study on the segmentation of carotid vessel wall in multicontrast MR images based on U?Net neural network / 中华放射学杂志

Objective To investigate the value of automatic segmentation of carotid vessel wall in multicontrast MR images using U?Net neural network. Methods Patients were retrospectively collected from 2012 to 2015 in Carotid Atherosclerosis Risk Assessment (CARE II) study. All patients who recently suffered ischemic stroke and/or transient ischemic attack underwent identical, state?of?the?art multicontrast MRI technique. A total of 17 568 carotid vessel wall MR images from 658 subjects were included in this study after inclusion criteria and exclusion criteria. All MR images were analyzed using customized analysis platform (CASCADE). Randomly, 10 592 images were assigned into training dataset, 3 488 images were assigned into validating dataset and 3 488 images were assigned into test dataset according to a ratio of 6∶2∶2. Data augmentation was performed to avoid over fitting and improve the ability of model generalization. The fine?tuned U?Net model was utilized in the segmentation of carotid vessel wall in multicontrast MR images. The U?Net model was trained in the training dataset and validated in the validating dataset. To evaluate the accuracy of carotid vessel wall segmentation, the sensitivity, specificity and Dice coefficient were used in the testing dataset. In addition, the interclass correlation and the Bland?Altman analysis of max wall thickness and wall area were obtained to demonstrate the agreement of the U?Net segmentation and the manual segmentation. Results The sensitivity, specificity and Dice coefficient of the fine?tuned U?Net model achieved 0.878,0.986 and 0.858 in the test dataset, respectively. The interclass correlation (95% confidence interval) was 0.921 (0.915-0.925) for max wall thickness and 0.929 (0.924-0.933) for wall area. In the Bland?Altman analysis, the difference of max wall thickness was (0.037±0.316) mm and the difference of wall area was (1.182±4.953) mm2. The substantial agreement was observed between U?Net segmentation method and manual segmentation method. Conclusion Automatic segmentation of carotid vessel wall in multicontrast MR images can be achieved using fine?tuned U?Net neural network, which is trained and tested in the large scale dataset labeled by professional radiologists.

Blood Vessel and Intraplaque Hemorrhage Simultaneous Imaging Sequence in Evaluation of the Integrity of Willis Circle / 中国医学影像学杂志

Purpose To investigate the value of simultaneousnon-contrast angiographyand intra-plaquehemorrhage (SNAP) in assessing Willis circle integrity of the brain using three-dimensional time of flight (TOF) MRA as a reference. Materials and Methods According to the inclusion criteria, a total of 62 patients with stroke symptoms within 3 months were collected. All patients underwent head SNAP and TOF on Philips 3.0T MR scanner. TOF and SNAP were treated respectively at the Philips workstation using maximum and minimum intensity projections to generate non-enhanced MRA images which were, afterwards, interpreted using blind reading to determine the presence or absence of individual blood vessels in the circle of Willis. In order to assess the consistency of the interpretation, all images were interpreted again 2 weeks later to avoid memory bias. Finally, The consistency of the two interpretations was analyzed, and the ability of SNAP and TOF in evaluating the integrity of Willis circle was compared. In addition, for 15 patients with VISTA images, the plaque of Willis circle vascular wall was evaluated using SNAP and VISTA images to compare their evaluation ability. Results The results of the first and second interpretations of SNAP and TOF were highly consistent. The arteries with exactly the same interpretation included left anterior cerebral artery A1 (LA1), right anterior cerebral artery A1 (RA1), left posterior cerebral artery P1 (LP1) and right posterior cerebral artery P1 (RP1) (Kappa=1.000). The results in terms of left posterior communicating artery (LPCoA) (Kappa=0.926 and 0.924, respectively), right posterior communicating artery (RPCoA) (Kappa=0.931 and 0.732, respectively) and integrity (Kappa=0.815 and 0.816, respectively) were relatively consistent, while there was a relatively low consistency in terms of anterior communicating artery (ACoA) (Kappa=0.640 and 0.675, respectively). In evaluating the constituent vessels of the circle of Willis and its integrity, SNAP and TOF showed good consistency. Vessels with identical SNAP and TOF interpretations include LA1, RA1, LP1, and RP1 (Kappa=1.000). The interpretation consistency of the two imaging techniques on LPCoA (Kappa=0.852 and 0.848), RPCoA (Kappa=0.796 and 0.796, respectively), and integrity (Kappa=0.701 and 0.742, respectively) was relatively high. While the consistency of SNAP and TOF in terms of ACoA was slightly lower (Kappa=0.680 and 0.714, respectively). In assessing the plaque of Willis circle vascular wall, the evaluation results of SNAP and VISTA were consistent. Conclusion As a non-enhanced angiography technique, the magnetic resonance SNAP sequence can well evaluate the integrity of Willis circle, as well as the presence or absence of plaque on Willis circle vascular wall.

MRI for characterizing progressive carotid plaques at different sites in elderly carotid atherosclerosis patients / 中华老年心脑血管病杂志

Objective To study the relationship between progression and clinical characteristics of carotid plaques at different sites in elderly carotid atherosclerosis patients.Methods Fifty-one elderly carotid atherosclerosis patients who underwent twice of high resolution MRI in our hospital were included in this study.The MRI data of 75 carotid arteries,annual progression of stenosis,maximum wall thickness,wall size,lumen size at plaques in common carotid artery (CCA),carotid bifurcation (BIF) and internal carotid artery (ICA) were analyzed.Results Of the 131 plaques in 75 carotid arteries,42 were detected in ICA,54 were detected in BIF.The annual progression rates of wall size and maximum wall thickness were higher at plaques in ICA than at those in CCA and BIF (P=0.036,P=0.028).The wall size of plaques in ICA was related with age,hypertension and diabetes (r=0.39,P=0.011;r=0.37,P=0.016;r=0.31,P=0.041).Conclusion The progression of plaques in ICA is faster than that of those in CCA and BIF,and is related with the high clinical risk factors.MRI follow-up can characterize the progressive plaques in elderly carotid atherosclerosis patients.

Relationship between carotid atherosclerosis score and its high-resolution MRI characteristics / 中华老年心脑血管病杂志

Objective To study the relationship of carotid atherosclerosis score (CAS) with carotid atherosclerotic disease and its clinical indexes in ischemic stroke patients.Methods Ninety-six patients with ischemic stroke (＜2 weeks) or transient ischemic attack underwent high-resolution MRI of bilateral carotid arteries to measure their carotid atherosclerotic luminal stenosis,maximal wall thickness (MaxWT) and plaque involvement size.The carotid plaque images with unstable components of lipid-rich core (LRC) were analyzed with CASCAD software.The CAS value was calculated and divided into low risk group and high risk group.The relationship between CAS and its clinical indexes was analyzed.Results LRC was detected in plaques of 148 carotid arteries of the 96 patients with a CAS value of 21.6％±17.5％.The CAS value was related with the carotid luminal stenosis,MaxWT and plaque involvement size (r =0.610,r=0.569,r =0.527,P＜ 0.001).A significant difference was found in carotid luminal stenosis,MaxWT and plaque involvement size with a different CAS value (P＜0.01).The CAS value was related with LDL and HDL/TC ratio (r=0.469,P＜0.01;r=-0.269,P=0.035).The SBP,DBP and LDL level were higher in high risk group than in low risk group (P＜0.05).Conclusion CAS is closely related with carotid atherosclerotic disease and lipid metabolism.The higher the CAS value is,the higher the risk of carotid plaque hemorrhage and fibrous cap rupture is.

Characteristics of Atherosclerotic Plaque in Thoracic Aorta in the Elderly Evaluated by Magnetic Resonance Vessel Wall Imaging / 中国医学影像学杂志

Purpose The thoracic aortic atherosclerotic plaque is an important source of ischemic stroke embolism in the elderly.This study aims to explore the characteristics of vulnerable plaque of atherosclerosis in the thoracic aorta in the elderly by using three-dimensional multi-contrast magnetic resonancewall imaging technique,so as to actively prevent the occurrence of cardiovascular and cerebrovascular complications.Materials and Methods Fifty-three cases of elderly subjects (＞60 years old) without serious cerebrovascular diseases were recruited in this prospective study.All subjects were divided into A and B groups (60-74 and 75-90 years old).All subjects underwent MRI of multiple contrast sequences of the aortic wall.The thoracic aorta was divided into three segments of ascending aorta,aortic arch and descending aorta,and the characteristics of the atherosclerotic plaque were evaluated.The load characteristics of thoracic aortic atherosclerotic plaques in the elderly were calculated quantitatively,and the compositional characteristics were evaluated qualitatively.Results The incidence of intraplaque hemorrhage in the thoracic aortic atherosclerotic plaque in the elderly was 26.4％ (14/53),and the incidence of lipid nucleus was 94.3％ (50/53).Meanwhile,the maximum wall thickness of three segments of ascending aorta,aortic arch and descending aorta in group B was significantly higher than that in group A [(3.1±0.6) mm vs (3.0±0.4) mm,P＜0.05;(3.2±0.7) mm vs (3.1 ±0.7) mm,P＜0.05;(3.0±0.8) mm vs (2.9±0.7) mm,P＜0.001];the normalized wall index of the three segments in group B was significantly higher than that in group A [(26.9±3.5)％ vs (26.7±2.9)％,P＜0.001;(31.9±5.1)％ vs (31.0±5.1)％,P＜0.001;(34.6±5.0)％ vs (34.1 ±4.6)％,P＜0.001)].Conclusion The incidence of hemorrhage in the atherosclerotic plaque in the thoracic aorta in the elderly is higher,and the plaque load increases with age.Therefore,early screening of vulnerable plaque in the thoracic aorta in the elderly using magnetic resonance wall imaging will be helpful for prevention and treatment of stroke.

Influences of statin treatment on MR vessel wall imaging-observed characteristics of atherosclerotic plaque in the thoracic aorta in the elderly / 中华老年医学杂志

Objective To evaluate the influences of statin treatment on MR vessel wall imagingobserved characteristics of atherosclerotic plaque in the thoracic aorta of the elderly.Methods Elderly subjects (≥ 60 years) without any serious cerebro-cardiovascular diseases were recruited.Thoracic aorta was imaged on MR scanner for all the subjects.The plaque burden was calculated quantitatively,the composition of plaque in thoracic aorta was evaluated qualitatively,and the contributions of statin treatment to these characteristics were also compared by image interpretation personals.The thoracic aorta was divided into three segments (AAO:ascending aorta;AOA:aortic arch,and DOA:descending aorta)on the imaging.Results Totally 55 recruited subjects had atherosclerotic plaque in thoracic aorta,with 24 subjects receiving statin treatment,and 50 ％ (12/24) male,aged 73.8±6.3 years.The level of LDL C[(2.4±0.7)mmol/L vs.(3.1±0.8)mmol/L(P＜ 0.01)]and total cholesterol[(4.4±0.6)mmol/L vs.(5.1 ±1.0)mmol/L(P＜0.01)]were significantly lower in statin group than in non-statin group.The lumen area,wall area,and total vessel area in all three segments of thoracic aorta were significantly smaller in statin group(all P＜0.05)than in nonstatin group.The average wall thickness in segment of AOA[(2.7±0.3)mm vs.(2.8±0.4)mm(P＜0.01)]and DAO[(2.5±0.4)mm vs.(2.6±0.5)mm(P＜0.01)]were smaller in statin group than in non-statin group.The incidence rate of intraplaque hemorrhage / mural thrombus [6 cases (25.0％) vs.8 cases(25.8 ％)]in thoracic aorta was a little lower in statin group than in non-statin group,with no significant difference(P＞0.05).Conclusions Statin treatment decreases LDL-C level,reduces the burden of atherosclerotic plaque in thoracic aorta,and maintains the atherosclerotic plaque stability.

Increased ostial pulmonary vein diameter in congestive heart failure: a multi-slice computed tomography angiography evaluation / 老年心脏病学杂志（英文版）

Background and Objectives The relationship between left atrial (LA) size and congestive heart failure (CHF) is well recognized;however, there is little information on the association of pulmonary vein (PV) diameter and CHF.The purpose of this study was to investigate the changes of PV and LA sizes in CHF patients by multislice computed tomography (MSCT) angiography using a new 64-slice scanner. Methods and Results We assessed diameters of PVs ostium and LA by 64-slice MSCT with three-dimensional reconstruction in 25 CHF patients and in 26 age- and sex-matched non-CHF controls. Compared with controls, CHF patients showed significant greater diameters of left superior pulmonary vein (LSPV) and right inferior pulmonary vein (RIPV) in both anteriorposterior(AP) and superior-inferior (SI) directions (P＜0.01), significant dilation of right superior pulmonary vein (RSPV) in AP direction (P＜0.05), as well as significant increase of LA transverse, AP, and SI diameters (P＜0.01). Conclusion Significant dilation of PVs with simultaneous LA enlargement was demonstrated in CHF patients. This anatomic and geometric changes may participate in the perpetuation of AF.

Clinical Value and Detection of Myocardial Bridge by Multi-slice Spiral CT Coronary Angiography / 实用放射学杂志

Objective To explore the characteristic and clinical meaning of myocardial bridge in patients with multi-slice spiral CT(MSCT) coronary angiography.Methods 875 patients with suspected or diagnosed coronary artery disease were studied with MSCT coronary angiography,579 cases were male and 296 cases were female,ranged from 30~87 years old in age with average of 60 years old.The heart was scanned with retrospectively ECG-gated,reconstruction from phase for all original tomographic source images at 30%~40% R-R phase interval according to heart rate ≥75 bpm and at 40%~50% R-R phase interval according to heart rate ≤75 bpm.The images of maximum intensity projection,multiplanar reconstruction and vulume reconstruction were gained to show left and right main coronary artery and their main branchs on mulliple angle.Results The myocardial bridges were detected in 89 cases,the positive rate was 10.2%,36 cases(40.5%)had different grade stenosis(≤50%).The myocardial bridges located at middle segment of anterior descending artery(79.8%),the thickness of myocardial bridge was 0.06~0.55 cm,the atherosclerosis plaques were found in 20 cases,the vascular stenosis was approximate to 50%.39 cases with simple myocardial bridges had angina,21 cases hadn’t angina.23 cases with myocardial bridges accompanied by coronary artery atherosclerosis had angina.Conclusion MSCT coronary angiography may clearly detect the myocardial bridge,which can provide more worthy information for clinical diagnosis and treatment.

Myocardial bridges and mural coronary artery:identification with multi-slice CT versus coronary angiography / 中华放射学杂志

Objective To compare the feasibility of multi-slices computed tomography(MSCT)versus coronary angiography(CAG)for identification of myocardial bridges and mural coronary artery(MB-MCA).Methods For 76 patients suspecting coronary heart disease the MSCT and CAG were performed for the coronary imaging.The MB-MCAs were identified on the CT images and angiograms by radiologists and cardiologists separately and independently before and after consulting each other.The data was statistically analyzed using ?2-squrae test.Results Before consulting each other,29 MB-MCA in 27 patients and 2 MB-MCA in two cases were detected using MSCT and CAG by radiologists and cardiologist separately and independently with significant difference statistically(?2=10.52,P