Objective

We aimed to develop deep learning classifiers for assessing therapeutic response on bone scans of patients with prostate cancer.

Methods

A set of 3791 consecutive bone scans coupled with their last previous scan (1528 patients) was evaluated. Bone scans were labeled as “progression” or “nonprogression” on the basis of clinical reports and image review. A 2D-convolutional neural network architecture was trained with three different preprocessing methods: 1) no preprocessing (Raw), 2) spatial normalization (SN), and 3) spatial and count normalization (SCN). Data were allocated into training, validation, and test sets in the ratio of 72:8:20, with the 20% independent test set rotating all scans over a five-fold testing procedure. A Grad-CAM algorithm was employed to generate class activation maps to visualize the lesions contributing to the decision. Diagnostic performance was compared using area under the receiver operating characteristics curves (AUCs).

Results

The data consisted of 791 scans labeled as “progression” and 3000 scans labeled as “nonprogression.” The AUCs of the classifiers were 0.632–0.710 on the Raw dataset, were significantly higher with the use of SN at 0.784–0.854 (p < 0.001 for Raw versus SN), and higher still with SCN at 0.954–0.979 (p < 0.001 for SN versus SCN). Class activation maps of the SCN model visualized lesions contributing to the model’s decision of progression.

Conclusion

With preprocessing of spatial and count normalization, our deep learning model achieved excellent performance in classifying the therapeutic response of bone scans in patients with prostate cancer.

中文翻译：

---

深度学习模型在前列腺癌全身骨扫描反应评估中的表现

客观的

我们的目标是开发深度学习分类器，用于评估前列腺癌患者骨扫描的治疗反应。

方法

对一组 3791 次连续骨扫描及其上次扫描（1528 名患者）进行了评估。根据临床报告和图像审查，骨扫描被标记为“进展”或“非进展”。使用三种不同的预处理方法训练 2D 卷积神经网络架构：1) 无预处理 (Raw)、2) 空间归一化 (SN) 和 3) 空间和计数归一化 (SCN)。数据按照 72:8:20 的比例分配到训练集、验证集和测试集，其中 20% 的独立测试集在五倍测试过程中轮换所有扫描。采用 Grad-CAM 算法生成类激活图，以可视化有助于决策的病变。使用受试者工作特征曲线下面积（AUC）来比较诊断性能。

结果

数据包括 791 次标记为“进展”的扫描和 3000 次标记为“非进展”的扫描。原始数据集上分类器的 AUC 为 0.632–0.710，使用 SN 时显着更高，为 0.784–0.854（ Raw 与 SN 的p  < 0.001），使用 SCN 时的 AUC 仍然更高，为 0.954–0.979（对于 SN，p  < 0.001） SN 与 SCN）。SCN 模型的类别激活图可视化有助于模型进展决策的病变。

结论

通过空间和计数标准化的预处理，我们的深度学习模型在对前列腺癌患者骨扫描的治疗反应进行分类方面取得了优异的性能。