Title of article :
Distinguishing Lymphomatous and Cancerous Lymph Nodes in 18F-Fluorodeoxyglucose Positron Emission Tomography/ Computed Tomography by Radiomics Analysis
Author/Authors :
Zheng, Bo Department of Biotherapy - Department of Oncology - West China Hospital - Sichuan University - Chengdu, China , Wu, Jiayi West China School of Medicine - Sichuan University - Chengdu, China , Zhao, Zixuan West China School of Medicine - Sichuan University - Chengdu, China , Ou, Xuejin West China School of Medicine - Sichuan University - Chengdu, China , Cao, Peng Department of Oncology - West China Hospital - Sichuan University - Chengdu, China , Ma, Xuelei Department of Biotherapy - Department of Oncology - West China Hospital - Sichuan University - Chengdu, China
Abstract :
The National Comprehensive Cancer Network guidelines recommend excisional biopsies for the diagnosis of
lymphomas. However, resection biopsies in all patients who are suspected of having malignant lymph nodes may cause unnecessary injury and increase medical costs. We investigated the usefulness of 18F-fluorodeoxyglucose positron emission/
computed tomography- (18F-FDG-PET/CT-) based radiomics analysis for differentiating between lymphomatous lymph nodes
(LLNs) and cancerous lymph nodes (CLNs). Methods. Using texture analysis, radiomic parameters from the 18F-FDG-PET/CT
images of 492 lymph nodes (373 lymphomatous lymph nodes and 119 cancerous lymph nodes) were extracted with the LIFEx
package. Predictive models were generated from the six parameters with the largest area under the receiver operating characteristics curve (AUC) in PET or CT images in the training set (70% of the data), using binary logistic regression. These models
were applied to the test set to calculate predictive variables, including the combination of PET and CT predictive variables
(PREcombination). The AUC, sensitivity, specificity, and accuracy were used to compare the differentiating ability of the
predictive variables. Results. Compared with the pathological diagnosis of the patient’s primary tumor, the AUC, sensitivity,
specificity, and accuracy of PREcombination in differentiating between LLNs and CLNs were 0.95, 91.67%, 94.29%, and 92.96%,
respectively. Moreover, PREcombination could effectively distinguish LLNs caused by various lymphoma subtypes (Hodgkin’s
lymphoma and non-Hodgkin’s lymphoma) from CLNs, with the AUC, sensitivity, specificity, and accuracy being 0.85 and 0.90,
77.78% and 77.14%, 97.22% and 88.89%, and 90.74% and 83.10%, respectively. Conclusions. Radiomics analysis of 18F-FDG-PET/
CT images may provide a noninvasive, effective method to distinguish LLN and CLN and inform the choice between fine-needle
aspiration and excision biopsy for sampling suspected lymphomatous lymph nodes.
Keywords :
18F , Tomography , CT , 18F-FDG-PET/CT
Journal title :
Contrast Media and Molecular Imaging