Logistic Regression Analysis: Application In Qualitative Diagnosis Of Solitary Pulmonary Nodules

Objective: The purpose of this study was to use logistic regression to differentiatebenign from malignant pulmonary nodules on High-resolution CT findings and to evaluate the value of this method in the differentiating benign from malignant pulmonary nodules.Materials and Methods: 149 solitary pulmonary nodules with diameter<3cmwere performed HRCT scan after common CT scan and were analyzed the configuration, sharp margin,spiculation, serrated margin, halo sign, vascular convergence and lymphadenopathy et al 8 radio logic findings that were extracted by two attending radiologists using subjective rating and were recorded respectively. We designed the logistic regression with the 8 input units for 8 HRCT findings and one output unit corresponding to the likelihood of malignancy.Results: 46 cases of lung benign nodules included inflammatory pseudotumor(n=31,34%), hamartoma(n=6,4.03%),bronchogenic cyst(n=4,2.68%), hemangioma(n=2.1.34%) .tuberculosis (n=l ,0.67%), lung leiomyoma (n=l ,0.67%) ,organizinghematoma (n=1.0.67%) ;103 cases included adenocarcinoma (n=36,24.16%) ,squamous cell carcinomas ( n=29,19.46% ) , bronchioalveolar carcinoma( n=l4,9.4%), adenosquamous carcinoma (n=9,6.04%), oats cell carcinoma( n=6,4.03% ) , small cell lung carcinoma ( n=4,2.68% ) carcinoid tumor(n=3,2.01%), large cell carcinoma (n=l,0.67%) , lung blastoma (n=l, 0. 67%) .The lobulation,spiculation,and vascular convergence of SPN on HRCT wereobserved with significantly greater frequency among malignant(P<0. 05).HRCT ofthe SPN accurately reflects a number of margin and internal characteristics, butnone of them was found only in lung cancer.The output showed that number of cases indicated equal to or more than 80% of likelihood of maligmancy was 0.7. Ninty-three cases (95.9%)were malignant, and only four cases (4.1%) were benign. The logistic regression showed a high performance in differentiating benign from malignant pulmonary nodules.Conclusion: The logistic regression was a valuable method in diagnosis anddistinguishing diagnosis of SPN.