Department of Respiratory Medicine, Tokyo Medical and Dental University, 1-5-45, Yushima, Tokyo, Bunkyo-ku, 113-8510, Japan.
Department of Biostatistics, M&D Data Science Center, Tokyo Medical and Dental University, Tokyo, Japan.
Eur Radiol. 2024 Jan;34(1):374-383. doi: 10.1007/s00330-023-09937-x. Epub 2023 Aug 3.
To compare the [F]FDG PET/CT findings of untreated sarcoidosis and malignant lymphoma (ML) and develop convolutional neural network (CNN) models to differentiate between these diseases using maximum intensity projection (MIP) [F]FDG PET images.
We retrospectively collected data on consecutive patients newly diagnosed with sarcoidosis and ML who underwent [F]FDG PET/CT before treatment. Two nuclear radiologists reviewed the images. CNN models were created using MIP PET images and evaluated with k-fold cross-validation. The points of interest were visualized using gradient-weighted class activation mapping (Grad-CAM).
A total of 56 patients with sarcoidosis and 62 patients with ML were included. Patients with sarcoidosis had more prominent FDG accumulation in the mediastinal lymph nodes and lung lesions, while those with ML had more prominent accumulation in the cervical lymph nodes (all p < 0.001). For the mediastinal lymph nodes, sarcoidosis patients had significant FDG accumulation in the level 2, 4, 7, and 10 lymph nodes (all p < 0.01). Otherwise, the accumulation in ML patients tended to be in the level 1 lymph nodes (p = 0.08). The CNN model using frontal and lateral MIP images achieved an average accuracy of 0.890 (95% CI: 0.804-0.977), a sensitivity of 0.898 (95% CI: 0.782-1.000), a specificity of 0.907 (95% CI: 0.799-1.000), and an area under the curve of 0.963 (95% CI: 0.899-1.000). Grad-CAM showed that the model focused on the sites of abnormal FDG accumulation.
CNN models based on differences in FDG accumulation sites archive high performance in differentiating between sarcoidosis and ML.
We developed a CNN model using MIP images of [F]FDG PET/CT to distinguish between sarcoidosis and malignant lymphoma. It achieved high performance and could be useful in diagnosing diseases with involvement across organs and lymph nodes.
• There are differences in FDG distribution when comparing whole-body [F]FDG PET/CT findings in patients with sarcoidosis and malignant lymphoma before treatment. • Convolutional neural networks, a type of deep learning technique, trained with maximum-intensity projection PET images from two angles showed high performance. • A deep learning model that utilizes differences in FDG distribution may be helpful in differentiating between diseases with lesions that are characteristically widespread among organs and lymph nodes.
比较未经治疗的结节病和恶性淋巴瘤（ML）的[F]FDG PET/CT 表现，并开发卷积神经网络（CNN）模型，使用最大强度投影（MIP）[F]FDG PET 图像来区分这些疾病。
我们回顾性收集了 56 例新诊断为结节病和 62 例 ML 的连续患者的资料，这些患者在治疗前均进行了[F]FDG PET/CT 检查。两名核医学医师对图像进行了评估。使用 MIP PET 图像创建 CNN 模型，并使用 k 折交叉验证进行评估。使用梯度加权类激活映射（Grad-CAM）可视化感兴趣的区域。
共纳入 56 例结节病患者和 62 例 ML 患者。与 ML 患者相比，结节病患者纵隔淋巴结和肺部病变的 FDG 摄取更明显（均 P < 0.001）。对于纵隔淋巴结，结节病患者在第 2、4、7 和 10 水平淋巴结有明显的 FDG 摄取（均 P < 0.01）。而 ML 患者的摄取倾向于第 1 水平淋巴结（P = 0.08）。使用前后位和侧位 MIP 图像的 CNN 模型平均准确率为 0.890（95%CI：0.804-0.977），敏感度为 0.898（95%CI：0.782-1.000），特异度为 0.907（95%CI：0.799-1.000），曲线下面积为 0.963（95%CI：0.899-1.000）。Grad-CAM 显示模型关注的是异常 FDG 摄取部位。
基于 FDG 摄取部位差异的 CNN 模型在区分结节病和 ML 方面具有较高的性能。
我们使用[F]FDG PET/CT 的 MIP 图像开发了一个 CNN 模型来区分结节病和恶性淋巴瘤。它具有较高的性能，可用于诊断涉及器官和淋巴结的疾病。
比较未经治疗的结节病和恶性淋巴瘤患者的全身[F]FDG PET/CT 表现，发现 FDG 分布存在差异。
利用前后位和侧位 MIP 图像训练的卷积神经网络，性能较高。
利用 FDG 分布差异的深度学习模型，可能有助于区分特征性地广泛累及器官和淋巴结的疾病。
