Development of a Food Composition Database for Assessing Nitrate and Nitrite Intake from Animal-based Foods

To investigate multi-site CNN segmentation of 4D flow MRI for pediatric blood flow measurement.
Retrospective.
A total of 174 subjects across two sites (female: 46%; N = 38 healthy controls, N = 136 CHD patients). Participants from site 1 (N = 100), site 2 (N = 74), and both sites (N = 174) were divided into subgroups to conduct 10-fold cross validation (10% for testing, 90% for training).
3 T/1.5 T; retrospectively gated gradient recalled echo-based 4D flow MRI.
Accuracy of the 3D CNN segmentations trained on data from single site (single-site CNNs) and data across both sites (multi-site CNN) were evaluated by geometrical similarity (Dice score, human segmentation as ground truth) and net flow quantification at the ascending aorta (Qs), main pulmonary artery (Qp), and their balance (Qp/Qs), between human observers, single-site and multi-site CNNs.
Kruskal-Wallis test, Wilcoxon rank-sum test, and Bland-Altman analysis. A P-value <0.05 was considered statistically significant.
No difference existed between single-site and multi-site CNNs for geometrical similarity in the aorta by Dice score (site 1: 0.916 vs. 0.915, P = 0.55; site 2: 0.906 vs. 0.904, P = 0.69) and for the pulmonary arteries (site 1: 0.894 vs. 0.895, P = 0.64; site 2: 0.870 vs. 0.869, P = 0.96). Qs site-1 medians were 51.0-51.3 mL/cycle (P = 0.81) and site-2 medians were 66.7-69.4 mL/cycle (P = 0.84). Qp site-1 medians were 46.8-48.0 mL/cycle (P = 0.97) and site-2 medians were 76.0-77.4 mL/cycle (P = 0.98). Qp/Qs site-1 medians were 0.87-0.88 (P = 0.97) and site-2 medians were 1.01-1.03 (P = 0.43). Bland-Altman analysis for flow quantification found equivalent performance.
Multi-site CNN-based segmentation and blood flow measurement are feasible for pediatric 4D flow MRI and maintain performance of single-site CNNs.
3 TECHNICAL EFFICACY: Stage 2.

摘要：
研究用于儿科血流测量的 4D 流 MRI 的多站点 CNN 分割。
回顾性的。
两个地点共有 174 名受试者（女性：46%；N = 38 名健康对照者，N = 136 名冠心病患者）。来自站点 1（N = 100）、站点 2（N = 74）和两个站点（N = 174）的参与者被分成小组进行 10 折交叉验证（10% 用于测试，90% 用于培训）。
3⟩T/1.5T；回顾性门控梯度召回基于回声的 4D 流动 MRI。
通过几何相似性（Dice 分数，人类分割作为地面实况）和净流量量化评估了在单个站点（单站点 CNN）和跨两个站点（多站点 CNN）的数据上训练的 3D CNN 分割的准确性升主动脉 (Qs)、主肺动脉 (Qp) 及其在人类观察者、单站点和多站点 CNN 之间的平衡 (Qp/Qs)。
Kruskal-Wallis 检验、Wilcoxon 秩和检验和 Bland-Altman 分析。 P值<0.05被认为具有统计学意义。
通过 Dice 评分（站点 1：0.916 与 0.915，P = 0.55；站点 2：0.906 与 0.904，P = 0.69）和肺部几何相似性，单站点和多站点 CNN 之间没有差异动脉（站点 1：0.894 与 0.895，P = 0.64；站点 2：0.870 与 0.869，P = 0.96）。 Qs site-1 中位数为 51.0-51.3 mL/周期 (P = 0.81)，site-2 中位数为 66.7-69.4 mL/周期 (P = 0.84)。 Qp site-1 中位数为 46.8-48.0mL/周期 (P = 0.97)，site-2 中位数为 76.0-77.4mL/周期 (P = 0.98)。 Qp/Qs 站点 1 中位数为 0.87-0.88 (P = 0.97)，站点 2 中位数为 1.01-1.03 (P = 0.43)。用于流量量化的 Bland-Altman 分析发现等效性能。
基于 CNN 的多站点分割和血流测量对于儿科 4D 流动 MRI 是可行的，并保持单站点 CNN 的性能。
3 技术效能：第 2 阶段。