Department of Radiation Oncology, Duke University, Durham, North Carolina, USA.
Medical Physics Graduate Program, Duke Kunshan University, Kunshan, Jiangsu, China.
Med Phys. 2024 Mar;51(3):1931-1943. doi: 10.1002/mp.16695. Epub 2023 Sep 11.
Uncertainty quantification in deep learning is an important research topic. For medical image segmentation, the uncertainty measurements are usually reported as the likelihood that each pixel belongs to the predicted segmentation region. In potential clinical applications, the uncertainty result reflects the algorithm's robustness and supports the confidence and trust of the segmentation result when the ground-truth result is absent. For commonly studied deep learning models, novel methods for quantifying segmentation uncertainty are in demand.
To develop a U-Net segmentation uncertainty quantification method based on spherical image projection of multi-parametric MRI (MP-MRI) in glioma segmentation.
The projection of planar MRI data onto a spherical surface is equivalent to a nonlinear image transformation that retains global anatomical information. By incorporating this image transformation process in our proposed spherical projection-based U-Net (SPU-Net) segmentation model design, multiple independent segmentation predictions can be obtained from a single MRI. The final segmentation is the average of all available results, and the variation can be visualized as a pixel-wise uncertainty map. An uncertainty score was introduced to evaluate and compare the performance of uncertainty measurements. The proposed SPU-Net model was implemented on the basis of 369 glioma patients with MP-MRI scans (T1, T1-Ce, T2, and FLAIR). Three SPU-Net models were trained to segment enhancing tumor (ET), tumor core (TC), and whole tumor (WT), respectively. The SPU-Net model was compared with (1) the classic U-Net model with test-time augmentation (TTA) and (2) linear scaling-based U-Net (LSU-Net) segmentation models in terms of both segmentation accuracy (Dice coefficient, sensitivity, specificity, and accuracy) and segmentation uncertainty (uncertainty map and uncertainty score).
The developed SPU-Net model successfully achieved low uncertainty for correct segmentation predictions (e.g., tumor interior or healthy tissue interior) and high uncertainty for incorrect results (e.g., tumor boundaries). This model could allow the identification of missed tumor targets or segmentation errors in U-Net. Quantitatively, the SPU-Net model achieved the highest uncertainty scores for three segmentation targets (ET/TC/WT): 0.826/0.848/0.936, compared to 0.784/0.643/0.872 using the U-Net with TTA and 0.743/0.702/0.876 with the LSU-Net (scaling factor = 2). The SPU-Net also achieved statistically significantly higher Dice coefficients, underscoring the improved segmentation accuracy.
The SPU-Net model offers a powerful tool to quantify glioma segmentation uncertainty while improving segmentation accuracy. The proposed method can be generalized to other medical image-related deep-learning applications for uncertainty evaluation.
深度学习中的不确定性量化是一个重要的研究课题。对于医学图像分割，不确定性测量通常表示为每个像素属于预测分割区域的可能性。在潜在的临床应用中，不确定性结果反映了算法的稳健性，并在没有地面实况结果时支持分割结果的置信度和信任度。对于通常研究的深度学习模型，需要新的方法来量化分割不确定性。
基于多参数 MRI（MP-MRI）的球形图像投影，开发用于胶质瘤分割的 U-Net 分割不确定性量化方法。
将平面 MRI 数据投影到球面上等效于保留全局解剖信息的非线性图像变换。通过在我们提出的基于球形投影的 U-Net（SPU-Net）分割模型设计中纳入此图像变换过程，可以从单个 MRI 获得多个独立的分割预测。最终分割是所有可用结果的平均值，并且可以将变化可视化作为像素级不确定性图。引入不确定性得分来评估和比较不确定性测量的性能。在具有 MP-MRI 扫描（T1、T1-Ce、T2 和 FLAIR）的 369 名胶质瘤患者的基础上实现了所提出的 SPU-Net 模型。分别训练了三个 SPU-Net 模型来分割增强肿瘤（ET）、肿瘤核心（TC）和整个肿瘤（WT）。SPU-Net 模型与（1）具有测试时扩充（TTA）的经典 U-Net 模型和（2）基于线性缩放的 U-Net（LSU-Net）分割模型进行了比较，分别从分割准确性（Dice 系数、灵敏度、特异性和准确性）和分割不确定性（不确定性图和不确定性得分）两个方面进行了比较。
所开发的 SPU-Net 模型成功地为正确的分割预测（例如肿瘤内部或健康组织内部）实现了低不确定性，而为不正确的结果（例如肿瘤边界）实现了高不确定性。该模型可以识别 U-Net 中的遗漏肿瘤靶标或分割错误。从数量上看，SPU-Net 模型在三个分割目标（ET/TC/WT）中实现了最高的不确定性得分：0.826/0.848/0.936，而使用具有 TTA 的 U-Net 则为 0.784/0.643/0.872，而使用 LSU-Net 则为 0.743/0.702/0.876（缩放因子=2）。SPU-Net 还实现了统计学上显著更高的 Dice 系数，这强调了改进的分割准确性。
SPU-Net 模型提供了一种强大的工具，可量化胶质瘤分割的不确定性，同时提高分割准确性。该方法可以推广到其他与医学图像相关的深度学习应用程序中，以进行不确定性评估。
