Back to Search Start Over

A diagnosis model in nasopharyngeal carcinoma based on PET/MRI radiomics and semiquantitative parameters.

Authors :
Feng, Qi
Liang, Jiangtao
Wang, Luoyu
Ge, Xiuhong
Ding, Zhongxiang
Wu, Haihong
Source :
BMC Medical Imaging; 8/29/2022, Vol. 22 Issue 1, p1-9, 9p
Publication Year :
2022

Abstract

Background: The staging of nasopharyngeal carcinoma (NPC) is of great value in treatment and prognosis. We explored whether a positron emission tomography/ magnetic resonance imaging (PET/MRI) based comprehensive model of radiomics features and semiquantitative parameters was useful for clinical evaluation of NPC staging. Materials and methods: A total of 100 NPC patients diagnosed with non-keratinized undifferentiated carcinoma were divided into early-stage group (I—II) and advanced-stage group (III—IV) and divided into the training set (n = 70) and the testing set (n = 30). Radiomics features (n = 396 × 2) of the primary site of NPC were extracted from MRI and PET images, respectively. Three major semiquantitative parameters of primary sites including maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) in all NPC patients were measured. After feature selection, three diagnostic models including the radiomics model, the metabolic parameter model, and the combined model were established using logistic regression model. Finally, internal validation was performed, and a nomogram for NPC comprehensive diagnosis has been made. Results: The radiomics model and metabolic parameter model showed an area under the curve (AUC) of 0.83 and 0.80 in the testing set, respectively. The combined model based on radiomics and semiquantitative parameters showed an AUC of 0.90 in the testing set, with the best performance among the three models. Conclusion: The combined model based on PET/MRI radiomics and semiquantitative parameters is of great value in the evaluation of clinical stage (early-stage group and advanced-stage group) of NPC. [ABSTRACT FROM AUTHOR]

Details

Language :
English
ISSN :
14712342
Volume :
22
Issue :
1
Database :
Complementary Index
Journal :
BMC Medical Imaging
Publication Type :
Academic Journal
Accession number :
158782946
Full Text :
https://doi.org/10.1186/s12880-022-00883-6