Weber Andreas, Enderle-Ammour Kathrin, Schröder Karl-Moritz, Metzger Marc C, Brandenburg Leonard Simon, Beck Jürgen, Straehle Jakob, Steybe David, Hassan Mohamed, Schmid Severin, Ohm Birte, Werner Martin, Passlick Bernward, Schmelzeisen Rainer, Le Uyen-Thao, Bronsert Peter
Institute for Surgical Pathology, Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany.
Faculty of Biology, University of Freiburg, Freiburg, Germany.
Quant Imaging Med Surg. 2025 Sep 1;15(9):7711-7726. doi: 10.21037/qims-2024-2608. Epub 2025 Aug 12.
Images obtained from stimulated Raman scattering can be used to identify histomorphologically relevant information intraoperatively. In order to leverage deep learning algorithms for distinguishing tumoral and non-tumoral tissue, data preprocessing remains a crucial task and may affect the classification performance. To date, the effect of different preprocessing techniques on deep learning algorithm performance is unclear. This study aims to make a contribution to closing this knowledge gap.
To investigate the influence of different preprocessing techniques of images obtained from stimulated Raman scattering, six deep learning architectures (VGG19, ResNet50, InceptionResNetV2, Xception, ConvNeXt and Vision Transformer) and five different preprocessing procedures were compared. For this, annotated datasets comprising 542 images of tissue samples obtained from patients with oral squamous cell carcinoma and non-small cell lung carcinoma were used for network training. Each network was trained five times for 40 epochs. Performance metrics balanced accuracy, precision, recall and F1-score were recorded. Class activation and attention maps were used to highlight on which input pixels a prediction is based.
A scaling of the original pixel values of stimulated Raman scattering images to the range [0, 1] yielded a higher and more stable overall classification performance across the neural networks when compared to more sophisticated and computationally expensive methods [ =0.8327; standard deviation (SD) =0.0622 on scaled dataset and =0.7213 (SD =0.2315) on complex preprocessed dataset; P≤0.05]. Absolute performance was best on stimulated Raman histology images ( =0.8478; SD =0.1487).
This study shows that preprocessing of pixel values of stimulated Raman scattering images can have a great impact on the performance and the stability of deep learning algorithms when applied for classification of cancer tissue.
通过受激拉曼散射获得的图像可用于术中识别组织形态学相关信息。为了利用深度学习算法区分肿瘤组织和非肿瘤组织，数据预处理仍然是一项关键任务，并且可能会影响分类性能。迄今为止，不同预处理技术对深度学习算法性能的影响尚不清楚。本研究旨在为填补这一知识空白做出贡献。
为了研究受激拉曼散射获得的图像的不同预处理技术的影响，比较了六种深度学习架构（VGG19、ResNet50、InceptionResNetV2、Xception、ConvNeXt和视觉Transformer）和五种不同的预处理程序。为此，使用包含从口腔鳞状细胞癌和非小细胞肺癌患者获得的542张组织样本图像的注释数据集进行网络训练。每个网络训练40个轮次，共训练五次。记录性能指标平衡准确率、精确率、召回率和F1分数。使用类激活和注意力图来突出预测基于哪些输入像素。
与更复杂且计算成本更高的方法相比，将受激拉曼散射图像的原始像素值缩放到[0, 1]范围在整个神经网络中产生了更高且更稳定的总体分类性能[缩放数据集上的准确率 =0.8327；标准差（SD）=0.0622，复杂预处理数据集上的准确率 =0.7213（SD =0.2315）；P≤0.05]。绝对性能在受激拉曼组织学图像上最佳（准确率 =0.8478；SD =0.1487）。
本研究表明，当应用于癌症组织分类时，受激拉曼散射图像的像素值预处理对深度学习算法的性能和稳定性有很大影响。
