Construction of a Predictive Model Based on Logistic Regression Analysis for CT Artificial
Intelligence Technology and Tumor Markers in Malignant Pulmonary Nodules
FAN Guangming
Department of Radiology , The Second Affiliated Hospital of Guizhou University of Traditional Chinese Medicine,
Guiyang Guizhou 550000, China
Abstract：Objective To investigate the effectiveness of CT artificial intelligence technology combined with tumor markers for the
differential diagnosis of malignant pulmonary nodules. Methods A total of 453 patients with pulmonary nodules admitted to our
hospital from January 2018 to January 2021 were selected and divided into benign group (n=317) and malignant group (n=136)
according to the pathological findings. The clinical data, CT artificial intelligence parameters, serum gastrin-releasing peptide (Pro-
GRP), neuron-specific enolase (NSE), carcinoembryonic antigen (CEA), cytokeratin fragment 19 (CYFRA21-1), squamous cell
carcinoma antigen (SCC) were compared, multifactorial logistic regression equation was used to analyze the influencing factors
related to malignant pulmonary nodules, and R language was used to draw a column line graph for predicting malignant nodules.
Results The number of patients with smoking history and previous history of extrapulmonary malignant tumor in the malignant
group was more than that in the benign group (P<0.05). The nodule diameter and malignant probability in the malignant group
were higher than those in the benign group, and the appearance of burr sign, lesions located in the upper lobe, with irregular shape
and vacuole sign were more than those in benign group (P<0.05). The value of each individual index to evaluate the malignant
pulmonary nodules: the AUC of the malignant probability was greater than the diameter of the nodule, the burr sign, the upper lobe,
the irregular shape of the lesion, and the vacuole sign (P<0.05). The Pro-GRP, NSE, CEA, CYFRA21-1, SCC in the malignant
group were higher than those in the benign group (P<0.05). Multifactorial analysis showed that smoking history, previous history of
extra-pulmonary malignancy, probability of malignancy, Pro-GRP, NSE, CEA, CYFRA21-1 and SCC were risk factors associated
with malignant nodules (P<0.05). Based on the risk factors screened by the above multifactorial analysis, the prediction model of
malignant pulmonary nodules was drawn. Bootstrap internal validation showed that the calibration curve fitted well with the ideal
curve, the C-index was 0.984, the AUC was 0.925, the sensitivity was 85.29%, and the specificity was 83.91%, indicating that the
nomogram model in this study had good predictive ability. Conclusion The visualization prediction model of malignant pulmonary
nodules constructed based on CT artificial intelligence technology and Pro-GRP, NSE, CEA, CYFRA21-1 and SCC can accurately
and conveniently identify the nature of pulmonary nodules, and provide a reliable reference for clinical diagnosis and treatment.
范光明. 基于Logistic回归分析构建恶性肺结节CT人工智能技术和肿瘤标志物的预测模型[J]. 中国医疗设备, 2023, 38(1): 71-76.
FAN Guangming. Construction of a Predictive Model Based on Logistic Regression Analysis for CT Artificial
Intelligence Technology and Tumor Markers in Malignant Pulmonary Nodules. China Medical Devices, 2023, 38(1): 71-76.
