Victor Manuel Alves, Jaime Dos Santos Cardoso, João Gama
{"title":"2-[18F]FDG PET/CT图像中肺结节的三维卷积神经网络分类","authors":"Victor Manuel Alves, Jaime Dos Santos Cardoso, João Gama","doi":"10.1007/s13139-023-00821-6","DOIUrl":null,"url":null,"abstract":"<p><strong>Purpose: </strong>2-[<sup>18</sup>F]FDG PET/CT plays an important role in the management of pulmonary nodules. Convolutional neural networks (CNNs) automatically learn features from images and have the potential to improve the discrimination between malignant and benign pulmonary nodules. The purpose of this study was to develop and validate a CNN model for classification of pulmonary nodules from 2-[<sup>18</sup>F]FDG PET images.</p><p><strong>Methods: </strong>One hundred thirteen participants were retrospectively selected. One nodule per participant. The 2-[<sup>18</sup>F]FDG PET images were preprocessed and annotated with the reference standard. The deep learning experiment entailed random data splitting in five sets. A test set was held out for evaluation of the final model. Four-fold cross-validation was performed from the remaining sets for training and evaluating a set of candidate models and for selecting the final model. Models of three types of 3D CNNs architectures were trained from random weight initialization (Stacked 3D CNN, VGG-like and Inception-v2-like models) both in original and augmented datasets. Transfer learning, from ImageNet with ResNet-50, was also used.</p><p><strong>Results: </strong>The final model (Stacked 3D CNN model) obtained an area under the ROC curve of 0.8385 (95% CI: 0.6455-1.0000) in the test set. The model had a sensibility of 80.00%, a specificity of 69.23% and an accuracy of 73.91%, in the test set, for an optimised decision threshold that assigns a higher cost to false negatives.</p><p><strong>Conclusion: </strong>A 3D CNN model was effective at distinguishing benign from malignant pulmonary nodules in 2-[<sup>18</sup>F]FDG PET images.</p><p><strong>Supplementary information: </strong>The online version contains supplementary material available at 10.1007/s13139-023-00821-6.</p>","PeriodicalId":19384,"journal":{"name":"Nuclear Medicine and Molecular Imaging","volume":null,"pages":null},"PeriodicalIF":1.3000,"publicationDate":"2024-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10796312/pdf/","citationCount":"0","resultStr":"{\"title\":\"Classification of Pulmonary Nodules in 2-[<sup>18</sup>F]FDG PET/CT Images with a 3D Convolutional Neural Network.\",\"authors\":\"Victor Manuel Alves, Jaime Dos Santos Cardoso, João Gama\",\"doi\":\"10.1007/s13139-023-00821-6\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Purpose: </strong>2-[<sup>18</sup>F]FDG PET/CT plays an important role in the management of pulmonary nodules. Convolutional neural networks (CNNs) automatically learn features from images and have the potential to improve the discrimination between malignant and benign pulmonary nodules. The purpose of this study was to develop and validate a CNN model for classification of pulmonary nodules from 2-[<sup>18</sup>F]FDG PET images.</p><p><strong>Methods: </strong>One hundred thirteen participants were retrospectively selected. One nodule per participant. The 2-[<sup>18</sup>F]FDG PET images were preprocessed and annotated with the reference standard. The deep learning experiment entailed random data splitting in five sets. A test set was held out for evaluation of the final model. Four-fold cross-validation was performed from the remaining sets for training and evaluating a set of candidate models and for selecting the final model. Models of three types of 3D CNNs architectures were trained from random weight initialization (Stacked 3D CNN, VGG-like and Inception-v2-like models) both in original and augmented datasets. Transfer learning, from ImageNet with ResNet-50, was also used.</p><p><strong>Results: </strong>The final model (Stacked 3D CNN model) obtained an area under the ROC curve of 0.8385 (95% CI: 0.6455-1.0000) in the test set. The model had a sensibility of 80.00%, a specificity of 69.23% and an accuracy of 73.91%, in the test set, for an optimised decision threshold that assigns a higher cost to false negatives.</p><p><strong>Conclusion: </strong>A 3D CNN model was effective at distinguishing benign from malignant pulmonary nodules in 2-[<sup>18</sup>F]FDG PET images.</p><p><strong>Supplementary information: </strong>The online version contains supplementary material available at 10.1007/s13139-023-00821-6.</p>\",\"PeriodicalId\":19384,\"journal\":{\"name\":\"Nuclear Medicine and Molecular Imaging\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":1.3000,\"publicationDate\":\"2024-02-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10796312/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Nuclear Medicine and Molecular Imaging\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1007/s13139-023-00821-6\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"2023/8/30 0:00:00\",\"PubModel\":\"Epub\",\"JCR\":\"Q3\",\"JCRName\":\"RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Nuclear Medicine and Molecular Imaging","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1007/s13139-023-00821-6","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2023/8/30 0:00:00","PubModel":"Epub","JCR":"Q3","JCRName":"RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING","Score":null,"Total":0}
引用次数: 0
摘要
目的:2-[18F]FDG PET/CT 在肺结节的治疗中发挥着重要作用。卷积神经网络(CNN)能自动学习图像中的特征,并有可能提高肺结节恶性和良性之间的鉴别能力。本研究的目的是开发并验证一个 CNN 模型,用于从 2-[18F]FDG PET 图像中对肺结节进行分类:方法:回顾性筛选出 113 名参与者。每位参与者一个结节。2-[18F]FDG PET 图像经过预处理并标注了参考标准。深度学习实验需要将数据随机分成五组。测试集用于评估最终模型。其余数据集进行四倍交叉验证,以训练和评估一组候选模型,并选出最终模型。在原始数据集和增强数据集中,通过随机权重初始化训练了三种 3D CNN 架构的模型(堆叠 3D CNN、类 VGG 模型和类 Inception-v2 模型)。此外,还使用了来自 ImageNet 和 ResNet-50 的迁移学习:最终模型(堆叠三维 CNN 模型)在测试集中的 ROC 曲线下面积为 0.8385(95% CI:0.6455-1.0000)。在测试集中,该模型的灵敏度为 80.00%,特异度为 69.23%,准确度为 73.91%:结论:三维 CNN 模型能有效区分 2-[18F]FDG PET 图像中的良性和恶性肺结节:在线版本包含补充材料,可在 10.1007/s13139-023-00821-6。
Classification of Pulmonary Nodules in 2-[18F]FDG PET/CT Images with a 3D Convolutional Neural Network.
Purpose: 2-[18F]FDG PET/CT plays an important role in the management of pulmonary nodules. Convolutional neural networks (CNNs) automatically learn features from images and have the potential to improve the discrimination between malignant and benign pulmonary nodules. The purpose of this study was to develop and validate a CNN model for classification of pulmonary nodules from 2-[18F]FDG PET images.
Methods: One hundred thirteen participants were retrospectively selected. One nodule per participant. The 2-[18F]FDG PET images were preprocessed and annotated with the reference standard. The deep learning experiment entailed random data splitting in five sets. A test set was held out for evaluation of the final model. Four-fold cross-validation was performed from the remaining sets for training and evaluating a set of candidate models and for selecting the final model. Models of three types of 3D CNNs architectures were trained from random weight initialization (Stacked 3D CNN, VGG-like and Inception-v2-like models) both in original and augmented datasets. Transfer learning, from ImageNet with ResNet-50, was also used.
Results: The final model (Stacked 3D CNN model) obtained an area under the ROC curve of 0.8385 (95% CI: 0.6455-1.0000) in the test set. The model had a sensibility of 80.00%, a specificity of 69.23% and an accuracy of 73.91%, in the test set, for an optimised decision threshold that assigns a higher cost to false negatives.
Conclusion: A 3D CNN model was effective at distinguishing benign from malignant pulmonary nodules in 2-[18F]FDG PET images.
Supplementary information: The online version contains supplementary material available at 10.1007/s13139-023-00821-6.
期刊介绍:
Nuclear Medicine and Molecular Imaging (Nucl Med Mol Imaging) is an official journal of the Korean Society of Nuclear Medicine, which bimonthly publishes papers on February, April, June, August, October, and December about nuclear medicine and related sciences such as radiochemistry, radiopharmacy, dosimetry and pharmacokinetics / pharmacodynamics of radiopharmaceuticals, nuclear and molecular imaging analysis, nuclear and molecular imaging instrumentation, radiation biology and radionuclide therapy. The journal specially welcomes works of artificial intelligence applied to nuclear medicine. The journal will also welcome original works relating to molecular imaging research such as the development of molecular imaging probes, reporter imaging assays, imaging cell trafficking, imaging endo(exo)genous gene expression, and imaging signal transduction. Nucl Med Mol Imaging publishes the following types of papers: original articles, reviews, case reports, editorials, interesting images, and letters to the editor.
The Korean Society of Nuclear Medicine (KSNM)
KSNM is a scientific and professional organization founded in 1961 and a member of the Korean Academy of Medical Sciences of the Korean Medical Association which was established by The Medical Services Law. The aims of KSNM are the promotion of nuclear medicine and cooperation of each member. The business of KSNM includes holding academic meetings and symposia, the publication of journals and books, planning and research of promoting science and health, and training and qualification of nuclear medicine specialists.