Prediction of pathological classification of ground glass nodules based on artificial intelligence CT quantitative parameters and histogram parameters
10.3760/cma.j.cn115455-20230925-00301
- VernacularTitle:基于人工智能CT定量参数联合直方图参数预测磨玻璃结节病理分型
- Author:
Jie XU
1
;
Ruibin YANG
1
;
Lihua ZHAO
1
;
Lizhen LUO
1
;
Xiuqin GUO
1
Author Information
1. 佛山复星禅诚医院医学影像科，佛山 528000
- Publication Type:Journal Article
- Keywords:
Artificial intelligence;
Tomography scanners, X-ray computed;
Ground glass nodules;
Quantitative parameters;
Histogram parameters
- From:
Chinese Journal of Postgraduates of Medicine
2025;48(4):318-321
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze the prediction of pathological classification of ground glass nodules based on artificial intelligence computed tomography (CT) quantitative parameters combined with histogram parameters.Methods:The clinical data of 268 suspected patients with ground glass nodules admitted to Foshan Fosun Chancheng Hospital from June 2021 to June 2023 were retrospectively selected as the research subjects. They were divided into pre invasive lesions group (100 cases) and invasive lesions group(168 cases) according to pathological classification. Basic data of patients with different pathological classifications and the CT characteristics were compared, the prediction of pathological classification of ground glass nodules based on CT quantitative parameters combined and histogram parameters were analyzed by receiver operating characteristic (ROC) curve.Results:The edge and boundary of the tumor, shape of the lesion, the peripheral signs of the lesion and the boundary between the two groups had statistical differences ( P<0.05). The CT quantitative parameters of maximum diameter, lesion volume, average CT value in the invasive lesions group and pre invasive lesions group had statistical differences: (15.29 ± 3.20) cm vs. (9.75 ± 2.14) cm, (1.54 ± 0.31) cm 3 vs. (0.51 ± 0.10) cm 3, (- 328.16 ± 46.35) HU vs. (-541.25 ± 100.30) HU, P<0.05. The CT histogram parameters of inproportion of solid components, entropy and maximum CT value in the invasive lesions group and pre invasive lesions group had statistical differences: (66.39 ± 13.25)% vs. (42.65 ± 11.20)%, 4.31 ± 0.52 vs. 3.32 ± 0.39, (-75.34 ± 21.27) HU vs. (-141.72 ± 32.43)HU, P<0.05. Compared with the single prediction of CT quantitative parameters and CT histogram parameters, the combined prediction of the two parameters had higher value in predicting different pathological subtypes of ground glass nodules (the area under the curve was 0.877, P = 0.001). Conclusions:The combined detection of CT quantitative parameters and histogram parameters based on artificial intelligence can effectively evaluate the invasion status of ground glass nodules, which is beneficial for improving the detection of different pathological types of ground glass nodules.
