Differentiating brain metastases from different pathological types of lung cancers using texture analysis of T1 postcontrast MR

• Published in: Magnetic resonance in medicine Vol. 76; no. 5; pp. 1410 - 1419
• Main Authors: Li, Zhenjiang, Mao, Yu, Li, Hongsheng, Yu, Gang, Wan, Honglin, Li, Baosheng
• Format: Journal Article
• Language: English
• Published: United States Blackwell Publishing Ltd 01.11.2016; Wiley Subscription Services, Inc
• Subjects: brain metastases; Brain Neoplasms - diagnostic imaging; Brain Neoplasms - pathology; Brain Neoplasms - secondary; Contrast Media; Diagnosis, Differential; Discriminant Analysis; Feasibility Studies; Humans; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; lung cancer; Lung Neoplasms - diagnostic imaging; Lung Neoplasms - pathology; magnetic resonance imaging; Magnetic Resonance Imaging - methods; Pattern Recognition, Automated - methods; Reproducibility of Results; Sensitivity and Specificity; texture analysis; lung cancer; magnetic resonance imaging; brain metastases; texture analysis
• ISSN: 0740-3194; 1522-2594
• DOI: 10.1002/mrm.26029

Purpose The goal of this study was to investigate the feasibility of differentiating brain metastases from different types of lung cancers using texture analysis (TA) of T1 postcontrast MR images. Methods TA was performed, and four subset textures were extracted and calculated separately. The capability of each texture to classify the different types of lung carcinoma was investigated using the Kruskal‐Wallis test and receiver operating characteristic analysis. K‐nearest neighbor (KNN) classifier model and back‐propagation artificial neural network (BP‐ANN) classifier model were used to build models and improve the predictive ability of TA. Results Texture‐based lesion classification was highly specific in differentiating brain metastases originated from different types of lung cancers, with misclassification rates of 3.1%, 4.3%, 5.8%, and 8.1%, respectively, for small cell lung carcinoma, squamous cell carcinoma, adenocarcinoma, and large cell lung carcinoma. The BP‐ANN model had a better predictive ability than the KNN model. No texture feature could distinguish between all four types of lung cancer. Conclusions TA may predict the differences among various pathological types of lung cancer with brain metastases. The texture parameters, which reflect the tumor histopathology structure, may serve as an adjunct tool for clinically accurate diagnoses and deserves further investigation. Magn Reson Med 76:1410–1419, 2016. © 2015 International Society for Magnetic Resonance in Medicine