基于影像组学的肺癌分型预测
|
摘要
目的基于影像组学特征对肺癌中的两大亚型分类(小细胞肺癌与非小细胞肺癌)进行分型预测。方法在131名小细胞肺癌与非小细胞肺癌患者中(其中训练集包含119人,测试集中包含12人),从手动分割的病灶区域提取107维组学特征,使用R统计学软件中的FSelector包对影像组学特征进行关键特征筛选,构建支持向量机模型和k折交叉验证模型对肺癌患者的病理进行表型分类和验证,通过绘制受试者工作特征曲线(ROC曲线)图和计算曲线下面积(AUC)数值来对训练集和测试集中的肺癌分型预测效果进行评估。结果挑选出20个主要的影像组学特征用于小细胞肺癌与非小细胞肺癌的分型鉴别,这些特征对于训练集和测试集中的小细胞肺癌与非小细胞肺癌均有较好的区分能力。在测试集中,预肺癌亚型分类的准确率为75%,组学特征的AUC结果为0. 69。结论通过构建独特的影像组学特征,以用作区分小细胞肺癌与非小细胞肺癌的诊断因素。这对实现非侵入性的肺癌病理有效分型预测,指导肺癌患者后续治疗方案的选择具有重要指导意义。
Objective To predict the classification of small cell lung cancer and non-small cell lung cancer based on Radiomics. Methods This study involved 131 patients with small cell lung cancer and non-small cell lung cancer(including 119 in the training cohort and 12 in the validation cohort). The 107-dimensional omics features were extracted from the manually segmented lesions. The FSelector package in R statistical software was used to screen the key features of the phenomenological features. The support vector machines model and the k-fold cross-validation model were used to classify the pathology of lung cancer patients. The effect of lung cancer typing prediction in the training cohort and validation cohort was evaluated by plotting the receiver operating characteristic curve(ROC) and calculating the area under curve(AUC) values. Results This study selected 20 major Radiomics features for the differential diagnosis of small cell lung cancer and non-small cell lung cancer. These features were well differentiated between small cell lung cancer and non-small cell lung cancer in the training cohort and validation cohort. In the validation cohort,the accuracy of the pre-lung cancer subtype classification was 75%,and the AUC result of the radiomics characteristics was 0. 69. Conclusion This paper constructs a unique radiomics feature to be used as a diagnostic factor for distinguishing between small cell lung cancer and non-small cell lung cancer,which has important guiding significance for the realization of non-invasive pathologically effective classification of lung cancer and guiding the selection of follow-up treatment options for lung cancer patients.
Objective To predict the classification of small cell lung cancer and non-small cell lung cancer based on Radiomics. Methods This study involved 131 patients with small cell lung cancer and non-small cell lung cancer(including 119 in the training cohort and 12 in the validation cohort). The 107-dimensional omics features were extracted from the manually segmented lesions. The FSelector package in R statistical software was used to screen the key features of the phenomenological features. The support vector machines model and the k-fold cross-validation model were used to classify the pathology of lung cancer patients. The effect of lung cancer typing prediction in the training cohort and validation cohort was evaluated by plotting the receiver operating characteristic curve(ROC) and calculating the area under curve(AUC) values. Results This study selected 20 major Radiomics features for the differential diagnosis of small cell lung cancer and non-small cell lung cancer. These features were well differentiated between small cell lung cancer and non-small cell lung cancer in the training cohort and validation cohort. In the validation cohort,the accuracy of the pre-lung cancer subtype classification was 75%,and the AUC result of the radiomics characteristics was 0. 69. Conclusion This paper constructs a unique radiomics feature to be used as a diagnostic factor for distinguishing between small cell lung cancer and non-small cell lung cancer,which has important guiding significance for the realization of non-invasive pathologically effective classification of lung cancer and guiding the selection of follow-up treatment options for lung cancer patients.
引文
[1]Schiepers C,Hoh CK. FDG-PET Imaging in Oncology[M].Diagnostic Nuclear Medicine,2006:185-204.
[2]Qi ShY,Shi HF,Shi MJ,et al. Establishment and identification of radioresistant non-small cell lung cancer model[J]. Acta Anatiomica Sinica,2017,48(2):156-159.(in Chinese)齐淑雅,施汉飞,石梦婕,等.非小细胞肺癌放疗抵抗细胞模型的建立与鉴定[J].解剖学报,2017,48(2):156-159.
[3]Siegel RL,Miller KD,Jemal A. Cancer statistics,2015[J]. CA:Cancer J Clin,2015,65(1):5-29.
[4]Lambin P,Rios-Velazquez E,Leijenaar R,et al. Radiomics:extracting more information from medical images using advanced feature analysis[J]. Eur J Cancer,2012,48(4):441-446.
[5]Justino EJR,Bortolozzi F,Sabourin R. A comparison of SVM and HMM classifiers in the off-line signature verification[J]. Pattern Recognition Letters,2005,26(9):1377-1385.
[6]Huang J,Ling CX. Using AUC and accuracy in evaluating learning algorithms[J]. IEEE Transactions on Knowledge and Data Engineering,2005,17(3):299-310.
[7]Al Badawy EA, Saha A, Mazurowski MA. Deep learning for segmentation of brain tumors:Impact of cross-institutional training and testing[J]. Med Phys,2018,45(3):1150-1158.
[8]Hanley JA,Mc Neil BJ. The meaning and use of the area under a receiver operating characteristic(ROC)curve[J]. Radiology,1982,54:212-220.
[9]Lee MC,Boroczky L,Sungur-Stasik K,et al. Computer-aided diagnosis of pulmonary nodules using a two-step approach for feature selection and classifier ensemble construction[J]. Artificial Intelligence in Medicine,2010,50(1):43-53.
[10]Zhu Y,Tan Y,Hua Y,et al. Feature Selection and Performance Evaluation of Support Vector Machine(SVM)-Based Classifier for Differentiating Benign and Malignant Pulmonary Nodules by Computed Tomography[J]. J Digit Imaging,2010,23(1):51-65.
[11]Lundberg J, Halldin C, Farde L. Measurement of serotonin transporter binding with PET and[11C]MADAM:A test-retest reproducibility study[J]. Synapse,2010,60(3):256-263.
[12]Basu S. Developing predictive models for lung tumor analysis[D].Dissertations and Theses-Gradworks,2012.
[2]Qi ShY,Shi HF,Shi MJ,et al. Establishment and identification of radioresistant non-small cell lung cancer model[J]. Acta Anatiomica Sinica,2017,48(2):156-159.(in Chinese)齐淑雅,施汉飞,石梦婕,等.非小细胞肺癌放疗抵抗细胞模型的建立与鉴定[J].解剖学报,2017,48(2):156-159.
[3]Siegel RL,Miller KD,Jemal A. Cancer statistics,2015[J]. CA:Cancer J Clin,2015,65(1):5-29.
[4]Lambin P,Rios-Velazquez E,Leijenaar R,et al. Radiomics:extracting more information from medical images using advanced feature analysis[J]. Eur J Cancer,2012,48(4):441-446.
[5]Justino EJR,Bortolozzi F,Sabourin R. A comparison of SVM and HMM classifiers in the off-line signature verification[J]. Pattern Recognition Letters,2005,26(9):1377-1385.
[6]Huang J,Ling CX. Using AUC and accuracy in evaluating learning algorithms[J]. IEEE Transactions on Knowledge and Data Engineering,2005,17(3):299-310.
[7]Al Badawy EA, Saha A, Mazurowski MA. Deep learning for segmentation of brain tumors:Impact of cross-institutional training and testing[J]. Med Phys,2018,45(3):1150-1158.
[8]Hanley JA,Mc Neil BJ. The meaning and use of the area under a receiver operating characteristic(ROC)curve[J]. Radiology,1982,54:212-220.
[9]Lee MC,Boroczky L,Sungur-Stasik K,et al. Computer-aided diagnosis of pulmonary nodules using a two-step approach for feature selection and classifier ensemble construction[J]. Artificial Intelligence in Medicine,2010,50(1):43-53.
[10]Zhu Y,Tan Y,Hua Y,et al. Feature Selection and Performance Evaluation of Support Vector Machine(SVM)-Based Classifier for Differentiating Benign and Malignant Pulmonary Nodules by Computed Tomography[J]. J Digit Imaging,2010,23(1):51-65.
[11]Lundberg J, Halldin C, Farde L. Measurement of serotonin transporter binding with PET and[11C]MADAM:A test-retest reproducibility study[J]. Synapse,2010,60(3):256-263.
[12]Basu S. Developing predictive models for lung tumor analysis[D].Dissertations and Theses-Gradworks,2012.