Yang Xinyi, Silosky Michael, Wehrend Jonathan, Litwiller Daniel V, Nachiappan Muthiah, Metzler Scott D, Ghosh Debashis, Xing Fuyong, Chin Bennett B
Department of Biostatistics and Informatics, University of Colorado Anschutz Medical Campus, Aurora, CO 80045, USA.
Department of Radiology, University of Colorado Anschutz Medical Campus, Aurora, CO 80045, USA.
Bioengineering (Basel). 2024 Feb 27;11(3):226. doi: 10.3390/bioengineering11030226.
Deep learning (DL) algorithms used for DOTATATE PET lesion detection typically require large, well-annotated training datasets. These are difficult to obtain due to low incidence of gastroenteropancreatic neuroendocrine tumors (GEP-NETs) and the high cost of manual annotation. Furthermore, networks trained and tested with data acquired from site specific PET/CT instrumentation, acquisition and processing protocols have reduced performance when tested with offsite data. This lack of generalizability requires even larger, more diverse training datasets. The objective of this study is to investigate the feasibility of improving DL algorithm performance by better matching the background noise in training datasets to higher noise, out-of-domain testing datasets. Ga-DOTATATE PET/CT datasets were obtained from two scanners: Scanner1, a state-of-the-art digital PET/CT (GE DMI PET/CT; n = 83 subjects), and Scanner2, an older-generation analog PET/CT (GE STE; n = 123 subjects). , the data set from Scanner1, was reconstructed with standard clinical parameters (5 min; Q.Clear) and list-mode reconstructions (VPFXS 2, 3, 4, and 5-min). , data from Scanner2 representing out-of-domain clinical scans, used standard iterative reconstruction (5 min; OSEM). A deep neural network was trained with each dataset: Network1 for Scanner1 and Network2 for Scanner2. DL performance (Network1) was tested with out-of-domain test data (). To evaluate the effect of training sample size, we tested DL model performance using a fraction (25%, 50% and 75%) of for training. Scanner1, list-mode 2-min reconstructed data demonstrated the most similar noise level compared that of , resulting in the best performance ( = 0.713). This was not significantly different compared to the highest performance, upper-bound limit using in-domain training for Network2 ( = 0.755; -value = 0.103). Regarding sample size, the 1 score significantly increased from 25% training data ( = 0.478) to 100% training data ( = 0.713; < 0.001). List-mode data from modern PET scanners can be reconstructed to better match the noise properties of older scanners. Using existing data and their associated annotations dramatically reduces the cost and effort in generating these datasets and significantly improves the performance of existing DL algorithms. List-mode reconstructions can provide an efficient, low-cost method to improve DL algorithm generalizability.
用于[¹⁸F]DOTATATE PET病变检测的深度学习（DL）算法通常需要大量经过良好注释的训练数据集。由于胃肠胰神经内分泌肿瘤（GEP-NETs）发病率低以及手动注释成本高，这些数据集很难获得。此外，使用从特定站点PET/CT仪器、采集和处理协议获取的数据进行训练和测试的网络，在使用异地数据进行测试时性能会降低。这种缺乏通用性的情况需要更大、更多样化的训练数据集。本研究的目的是探讨通过使训练数据集中的背景噪声更好地匹配更高噪声的域外测试数据集来提高DL算法性能的可行性。[⁶⁸Ga] - DOTATATE PET/CT数据集来自两台扫描仪：Scanner1，一台先进的数字PET/CT（GE DMI PET/CT；n = 83名受试者），以及Scanner2，一台旧一代模拟PET/CT（GE STE；n = 123名受试者）。Scanner1的数据集使用标准临床参数（5分钟；Q.Clear）和列表模式重建（VPFXS 2、3、4和5分钟）进行重建。Scanner2的数据代表域外临床扫描，使用标准迭代重建（5分钟；OSEM）。使用每个数据集训练一个深度神经网络：用于Scanner1的Network1和用于Scanner2的Network2。使用域外测试数据（Scanner2）测试DL性能（Network1）。为了评估训练样本大小的影响，我们使用Scanner1的一部分（25%、50%和75%）进行训练来测试DL模型性能。Scanner1的列表模式2分钟重建数据与Scanner2的数据相比显示出最相似的噪声水平，从而产生了最佳性能（Dice = 0.713）。与使用Network2的域内训练的最高性能上限（Dice = 0.755；p值 = 0.103）相比，这没有显著差异。关于样本大小，Dice分数从25%训练数据（Dice = 0.478）显著增加到100%训练数据（Dice = 0.713；p < 0.001）。现代PET扫描仪的列表模式数据可以进行重建，以更好地匹配旧扫描仪的噪声特性。使用现有数据及其相关注释可显著降低生成这些数据集的成本和工作量，并显著提高现有DL算法的性能。列表模式重建可以提供一种有效、低成本的方法来提高DL算法的通用性。
