cardiac MRI

我们提出了一个使用形状和外观统计模型分析短轴心脏MRI的框架。该框架集成了时间和结构约束，避免了此类高维模型固有的常见优化问题。第一个贡献是引入了一种在短轴心脏MRI上拟合3D主动外观模型 (aam) 的算法。我们观察到拟合速度提高了44倍，分割精度与高斯-牛顿优化相当，高斯-牛顿优化是此类问题最广泛使用的优化算法之一。第二个贡献涉及对分层2D时间活动形状模型 (asm) 的研究，该模型集成了时间约束并同时改善了基于3D AAM的分割。我们在从33名受试者获得的7980短轴心脏MR图像上获得了令人鼓舞的结果 (心内膜/心外膜误差1.43 /-0.49毫米/1.51 /-0.48毫米)。

We propose a framework for analyzing short-axis cardiac MRI using statistical shape and appearance models. This framework integrates temporal and structural constraints to avoid common optimization issues inherent in such high-dimensional models. Our first contribution is the introduction of an algorithm for fitting 3D Active Appearance Models (AAMs) on short-axis cardiac MRI. We observe that the fitting speed is increased by a factor of 44, and the segmentation accuracy is comparable to that of Gauss-Newton optimization, which is one of the most widely used optimization algorithms for such problems. Our second contribution involves the study of a hierarchical 2D temporal Active Shape Model (ASM) that integrates temporal constraints and simultaneously improves segmentation based on 3D AAMs. We obtain encouraging results on 7980 short-axis cardiac MR images acquired from 33 subjects (endocardial/epicardial errors: 1.43 ± 0.49 mm / 1.51 ± 0.48 mm).

搜集汇总

背景与挑战

背景概述

该数据集包含7980张短轴心脏MR图像，来自33名受试者，用于研究心脏MRI分析框架。它提出了一个结合形状和外观统计模型的框架，集成了时间和结构约束，以提高分割精度和优化效率，例如通过3D主动外观模型算法实现44倍拟合速度提升，并报告了心内膜/心外膜分割误差数据。数据集由约克大学和IBM于2008年发布，适用于医学图像分析研究。