Fig. 4

QCIGISH diagnostic performance in benign lesions and lung cancers. A ROC chart showing the AUC performance of the QCIGISH diagnostic grading model in the model building, testing and validation sets. B Computed sensitivities and specificities of the final QCIGISH diagnostic grading model in the model building, testing and validation sets. C Heatmap analysis plot showing the elevated multiallelic expression patterns for the lung cancers as compared to the benign lesions for the cytology and small biopsy specimens from both model testing and validation sets. D Analysis showing high sensitivities of QCIGISH for the diagnosis of different lung cancer subtypes. E Analysis showing high specificities of QCIGISH for the diagnosis of different benign lung lesions. AdC, adenocarcinoma. SqCC, squamous cell carcinoma. AdSqLC, adenosquamous lung carcinoma. SCLC, small cell lung cancer. TB, pulmonary tuberculosis. COP, cryptogenic organizing pneumonia. NTBI, non-TB infections. Infl, inflammation. Gran, granuloma. Hama, hamartoma. Error bars on the bar charts showed the 95% CI