Dataset Information

Introduction of a cascaded segmentation pipeline for parametric T1 mapping in cardiovascular magnetic resonance to improve segmentation performance.

ABSTRACT: The manual and often time-consuming segmentation of the myocardium in cardiovascular magnetic resonance is increasingly automated using convolutional neural networks (CNNs). This study proposes a cascaded segmentation (CASEG) approach to improve automatic image segmentation quality. First, an object detection algorithm predicts a bounding box (BB) for the left ventricular myocardium whose 1.5 times enlargement defines the region of interest (ROI). Then, the ROI image section is fed into a U-Net based segmentation. Two CASEG variants were evaluated: one using the ROI cropped image solely (cropU) and the other using a 2-channel-image additionally containing the original BB image section (crinU). Both were compared to a classical U-Net segmentation (refU). All networks share the same hyperparameters and were tested on basal and midventricular slices of native and contrast enhanced (CE) MOLLI T1 maps. Dice Similarity Coefficient improved significantly (p < 0.05) in cropU and crinU compared to refU (81.06%, 81.22%, 72.79% for native and 80.70%, 79.18%, 71.41% for CE data), while no significant improvement (p < 0.05) was achieved in the mean absolute error of the T1 time (11.94 ms, 12.45 ms, 14.22 ms for native and 5.32 ms, 6.07 ms, 5.89 ms for CE data). In conclusion, CASEG provides an improved geometric concordance but needs further improvement in the quantitative outcome.

SUBMITTER: Viezzer D

PROVIDER: S-EPMC9902617 | biostudies-literature | 2023 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Introduction of a cascaded segmentation pipeline for parametric T1 mapping in cardiovascular magnetic resonance to improve segmentation performance.

Viezzer Darian D Hadler Thomas T Ammann Clemens C Blaszczyk Edyta E Fenski Maximilian M Grandy Thomas Hiroshi TH Wetzl Jens J Lange Steffen S Schulz-Menger Jeanette J

Scientific reports 20230206 1

The manual and often time-consuming segmentation of the myocardium in cardiovascular magnetic resonance is increasingly automated using convolutional neural networks (CNNs). This study proposes a cascaded segmentation (CASEG) approach to improve automatic image segmentation quality. First, an object detection algorithm predicts a bounding box (BB) for the left ventricular myocardium whose 1.5 times enlargement defines the region of interest (ROI). Then, the ROI image section is fed into a U-Net ...[more]

PMID: 36746989

Similar Datasets

Project description:AimTo evaluate the ability of single heartbeat fast-strain encoded (SENC) cardiovascular magnetic resonance (CMR) derived myocardial strain to discriminate between different forms of left ventricular (LV) hypertrophy (LVH).Methods314 patients (228 with hypertensive heart disease (HHD), 45 with hypertrophic cardiomyopathy (HCM), 41 with amyloidosis, 22 competitive athletes, and 33 healthy controls) were systematically analysed. LV ejection fraction (LVEF), LV mass index and interventricular septal (IVS) thickness, T1 mapping and atypical late gadolinium enhancement (LGE) were assessed. In addition, the percentage of LV myocardial segments with strain ≤ - 17% (%normal myocardium) was determined.ResultsPatients with amyloidosis and HCM exhibited the highest IVS thickness (17.4 ± 3.3 mm and 17.4 ± 6 mm, respectively, p < 0.05 vs. all other groups), whereas patients with amyloidosis showed the highest LV mass index (95.1 ± 20.1 g/m2, p < 0.05 vs all others) and lower LVEF compared to controls (50.5 ± 9.8% vs 59.2 ± 5.5%, p < 0.05). Analysing subjects with mild to moderate hypertrophy (IVS 11-15 mm), %normal myocardium exhibited excellent and high precision, respectively for the differentiation between athletes vs. HCM (sensitivity and specificity = 100%, Area under the curve; AUC%normalmyocardium = 1.0, 95%CI = 0.85-1.0) and athletes vs. HHD (sensitivity = 83%, specificity = 75%, AUC%normalmyocardium = 0.85, 95%CI = 0.78-0.90). Combining %normal myocardial strain with atypical LGE provided high accuracy also for the differentiation of HHD vs. HCM (sensitivity = 82%, specificity = 100%, AUCcombination = 0.92, 95%CI = 0.88-0.95) and HCM vs. amyloidosis (sensitivity = 83%, specificity = 100%, AUCcombination = 0.83, 95%CI = 0.60-0.96).ConclusionFast-SENC derived myocardial strain is a valuable tool for differentiating between athletes vs. HCM and athletes vs. HHD. Combining strain and LGE data is useful for differentiating between HHD vs. HCM and HCM vs. cardiac amyloidosis.

Project description:PurposeTo characterize T1 relaxation times of the pancreas, liver, and spleen in children with and without abdominal pathology.MethodsThis retrospective study included pediatric patients (< 18-years-old). T1 mapping was performed with a Modified Look-Locker Inversion Recovery sequence. Patients were grouped based on review of imaging reports and electronic medical records. The Kruskal-Wallis test with Dunn's multiple comparison was used to compare groups.Results220 participants were included (mean age: 11.4 ± 4.2 years (1.5 T); 10.9 ± 4.5 years (3 T)). Pancreas T1 (msec) was significantly different between subgroups at 1.5 T (p < 0.0001). Significant pairwise differences included: normal (median: 583; IQR: 561-654) vs. acute pancreatitis (731; 632-945; p = 0.0024), normal vs. chronic pancreatitis (700; 643-863; p = 0.0013), and normal vs. acute + chronic pancreatitis (1020; 897-1099; p < 0.0001). Pancreas T1 was also significantly different between subgroups at 3 T (p < 0.0001). Significant pairwise differences included: normal (779; 753-851) vs. acute pancreatitis (1087; 910-1259; p = 0.0012), and normal vs. acute + chronic pancreatitis (1226; 1025-1367; p < 0.0001). Liver T1 was significantly different between subgroups only at 3 T (p = 0.0011) with pairwise differences between normal (818, 788-819) vs. steatotic (959; 848-997; p = 0.0017) and normal vs. other liver disease (882; 831-904; p = 0.0455). Liver T1 was weakly correlated with liver fat fraction at 1.5 T (r = 0.39; 0.24-0.52; p < 0.0001) and moderately correlated at 3 T (r = 0.64; 0.49-0.76; p < 0.0001). There were no significant differences in splenic T1 relaxation times between subgroups.ConclusionPancreas T1 relaxation times are higher at 1.5 T and 3 T in children with pancreatitis and liver T1 relaxation times are higher in children with steatotic and non-steatotic chronic liver disease at 3 T.

Dataset Information

Introduction of a cascaded segmentation pipeline for parametric T1 mapping in cardiovascular magnetic resonance to improve segmentation performance.

Publications

Introduction of a cascaded segmentation pipeline for parametric T1 mapping in cardiovascular magnetic resonance to improve segmentation performance.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets