A Deep Learning Algorithm for Automatic 3D Segmentation of Rotator Cuff Muscle and Fat from Clinical MRI Scans
A Deep Learning Algorithm for Automatic 3D Segmentation of Rotator Cuff Muscle and Fat from Clinical MRI Scans
复制标题
DOI：
10.1148/ryai.220132
复制
发表时间：
2023-03-01
期刊：
影响因子：
--
通讯作者：
Blemker, Silvia S.
中科院分区：
文献类型：
--
作者：
Riem, Lara;Feng, Xue;Blemker, Silvia S.
作者旨在开发和验证一种自动化人工智能 (AI) 算法，用于对所有四个肩袖 (RC) 肌肉进行三维 (3D) 分割，以量化肌内脂肪浸润 (FI) 和个体肌肉体积。该数据集包括回顾性收集的 232 名患者的 RC MRI 扫描（63 名患者 RC 正常，169 名 RC 撕裂）。开发了一个两阶段 AI 模型来分割每个阶段的所有 RC 肌肉及其 FI。为了进行比较，创建了单级和 Otsu 过滤模型。使用两阶段模型，在 30 次扫描中进行验证时，分割性能表现出高 Dice 分数（平均 0.92 +/- 0.14 [SD]）、低体积误差（平均 5.72% +/- 9.23）和低 FI 误差（平均 1.54% +/- 2.79）。 RC 撕裂扫描中的 3D FI 与 Goutallier 等级（rho = 0.53，P < .001）和单个二维 (2D) 切片（所有肌肉，rho > 0.70；P < .001）中发现的 FI 之间存在显着相关性。然而，与 FI 的 2D 分析相比，3D 的 Bland-Altman 分析显示出比例偏差（所有肌肉，P < .001）。与 Goutallier 分类或单图像量化相比，AI 方法允许图像具有更大的可变性，并可对所有 RC 肌肉中的肌肉体积和 FI 进行客观的单独量化。
The authors aimed to develop and validate an automated artificial intelligence (AI) algorithm for three-dimensional (3D) segmentation of all four rotator cuff (RC) muscles to quantify intramuscular fat infiltration (FI) and individual muscle volume. The dataset included retrospectively collected RC MRI scans in 232 patients (63 with normal RCs, 169 with RC tears). A two-stage AI model was developed to segment all RC muscles and their FI in each stage. For comparison, single-stage and Otsu filtering models were created. Using the two-stage model, segmenta-tion performance demonstrated high Dice scores (mean, 0.92 +/- 0.14 [SD]), low volume errors (mean, 5.72% +/- 9.23), and low FI errors (mean, 1.54% +/- 2.79) when validated in 30 scans. There was a significant correlation between the 3D FI in the RC tear scans with a Goutallier grade (rho = 0.53, P < .001) and FI found from a single two-dimensional (2D) section (all muscles, rho > 0.70; P < .001). However, Bland-Altman analysis of the 3D compared with the 2D analyses of FI demonstrated a proportional bias (all muscles, P < .001). Compared with Goutallier classification or single-image quantification, the AI method allowed for more variability in images and led to objective separate quantifications of muscle volume and FI in all RC muscles.