The accurate diagnosis and precise prediction of tumor radiation sensitivity and normal tissue radiation-induced injury are the preconditions of precise radiotherapy for lung cancer.Radiomics is defined as a set of milestone,assistive tools in the develop-ment of precise treatment for lung cancer,which can extract many quantitative features from medical images by applying automatic or semi-automatic methods and determine the deep relationship between clinical diagnosis and treatment data.Thus,the occurrence, development,and clinical outcome of lung cancer may be revealed.Radiomics holds immense potential in the classification of benign and malignant lung nodules,prediction of lung cancer genetic phenotypes,and treatment response to radiation therapy,because it can obtain information regarding the global heterogeneity of tumors via a noninvasive approach.In the present review,we summarize the latest process of CT-based radiomics in precise radiotherapy for lung cancer.

Primary lung cancer is the first malignant tumor in our country and in the world with an increasing mortality trend, which seriously endangers the human health. By digging the deep relationship between high-dimensional imaging features and pathophysiological features, radiomics can establish a predictive model to distinguish pathological types, tumor stages, distant metastases and survival, guide individualized diagnosis and treatment strategies, and improve prognosis. PET/CT has higher diagnostic accuracy and specificity by reflecting tumor tissue metabolism. This article reviews the application of PET/CT radiomics in the treatment of non-small-cell lung cancer (NSCLC).

Objective To investigate the impact of multi-b-value on texture features of DWI in liver cirrhosis.Methods DWI manifestations of liver cirrhosis in 37 patients were analyzed retrospectively,and DWI of 27 healthy volunteers (control group) were enrolled as controls.The b values were set as 0,20,50,100,200,400,800,1 000,1 200 and 1500 s/mm2,respectively.Three ROIs at different levels of every set image were selected,and 37 texture features within these ROIs were extracted.Unstable texture features affected by different b-values were screened with the percent coefficient of variation (％COV),and the fitting degree between the unstable texture features and b values were analyzed with exponential fitting.Results Among 37 texture features,20 (20/37,54.05 ％) were unstable.With the increase of b values,exponential upward trend was found in 10 texture features,exponential downward trend was found in 4 texture features,and the relative trends could not be defined in other 6 unstable texture features.Conclusion The b values of DWI impact the texture features in liver cirrhosis.Correlations exist among some texture features and b values.

Objective:To provide a feasible method for the evaluation of cardiac function based on cardiac gated 4DCT, the radiomics technology combined with enhanced ECG gated 4DCT images were used to quantitatively analyze the changes of left ventricular CT radiomics characteristics in cardiac cycle.Methods:The enhanced ECG 4DCT images of 14 patients were reconstructed at intervals of 5% of cardiac cycle. The left ventricular muscle (LVM) and the contrast agent well filled area of left ventricular were delineated with a 13 mm diameter sphere (Cardiac Region of Interest, cardiac ROI) in a single phase. 3Dslicer software was used to extract 92 features of all the sketches, analyze the distribution of CT values on the cardiac ROI and LVM, and preliminarily screen the stable features based on the cardiac ROI (one-way ANOVA). The stable features were used to further screen LVM (one-way ANOVA) to get the difference features. Wilcoxon rank sum test was used to analyze the change of characteristics with heartbeat in the heartbeat cycle.Results:In the heartbeat cycle the mean CT values of cardiac cavity ROI in cardiac cavity changed less than that in LVM, with the change rates of 9.23% and 17.88%, respectively. There were 36 stable features with no significant difference in cardiac cavity ROI ( P>0.05). 20 of them were statistically significant ( F=1.641-6.206, P<0.05), and the average change rate was 98.63%, such as median (-103.96%) and mean (123.67%) of the first order matrix, gray level non uniformity (99.81%) of GLDM matrix and other changes reached more than 99%. The differences between the maximum and minimum values in different cardiac cycles were statistically significant ( Z=-3.921--3.173, P<0.05). Conclusions:With the combination of radiomics and enhanced ECG 4DCT image, the microscopic changes of CT image features in the cardiac cycle can be amplifed. A new method for the assessment of left ventricular function changes was provided. The features such as median, mean may have more application potential.

Radiomics has played an irreplaceable role along with the development of precision medicine. In the field of radiomics researches,the stability of imaging features is of vital significance,which is directly linked to the modeling analysis. In this review,we summarized the recent research progress on the reproducibility problems in four crucial steps of the standard workflow of radiomics including imaging acquisition and reconstruction, region of interest(ROI)segmentation, imaging feature extraction and modeling establishment. In addition,the commonly used software related to radiomics was briefly introduced.

Objective To investigate the changes of the parameters related to planning and re-planning CT imaging features in lung cancer patients presenting with radiation pneumonitis ( RP) by using radiomics technique,and identify the parameters intimately related to the incidence of RP. Methods A total of 31 lung cancer patients who were diagnosed with grade ≥ 2 RP after receiving radiation therapy were selected in this study. For each patient, planning CT images before radiation therapy and re-planning CT images after 40 Gy radiation therapy were obtained. The affected and contralateral lungs were considered as the region of interest (ROI).After the automatic segmentation of normal lung tissues,the parameters related to radiomics features were extracted from ROI by using radiomics software. The differences of these parameters between planning and re-planning CT images were statistically compared. Results ( 1 ) For unilateral lung within each time interval,86 parameters related to radiomics features were extracted; ( 2) Twenty-two parameters significantly differed between the affected and contralateral lungs prior to radiotherapy;(3) Twelve parameters significantly differed between the affected and contralateral lungs on re-planning CT images;(4) Twenty-eight parameters significantly differed in the affected lung before and after radiation therapy;(5) Twenty-eight parameters significantly differed in the contralateral lung before and after radiation therapy. Conclusions The CT imaging radiomics features significantly differ between planning and re-planning CT scan in partial lung cancer patients presenting with RP.Monitoring the dynamic changes of these parameters plays a potential role in predicting the incidence of RP.

Objective:To construct machine learning models based on CT imaging and clinical parameters for predicting progression-free survival (PFS) of locally advanced cervical cancer (LACC) patients after concurrent chemoradiotherapy (CCRT).Methods:Clinical data of 167 LACC patients treated with CCRT at Shandong Cancer Hospital from September 2015 to October 2021 were retrospectively analyzed. All patients were randomly divided into the training and validation cohorts according to the ratio of 7 vs. 3. Clinical features were selected by univariate and multivariate Cox proportional hazards model ( P<0.1). Radiomics models and nomograms were constructed by radiomics features which were selected by least absolute shrinkage and selection operator (LASSO) Cox regression model to predict the 1-, 3- and 5-year PFS. Combined models and nomogram models were developed by selected clinical and radiomics features. The Kaplan Meier-curve, receiver operating characteristic (ROC) curve, C-index and calibration curve were used to evaluate the model performance. Results:A total of 1 409 radiomics features were extracted based on the region of interest (ROI) in CT images. CT radiomics models showed better performance for predicting 1-, 3-and 5-year PFS than the clinical model in the training and validation cohorts. The combined model displayed the optimal performance in predicting 1-, 3-and 5-year PFS in the training cohort [area under the curve (AUC): 0.760, 0.648, 0.661, C-index: 0.740, 0.667, 0.709] and verification cohort (AUC: 0.763, 0.677, 0.648, C-index: 0.748, 0.668, 0.678).Conclusions:Combined model constructed based on CT radiomics and clinical features yield better prediction performance than that based on radiomics or clinical features alone. As an objective image analysis approach, it possesses high prediction efficiency for PFS of LACC patients after CCRT, which can provide reference for clinical decision-making.

Objective: To explore the ability of computed-tomography (CT) radiomic features to predict the Epidermal growth factor receptor (EGFR) mutation status and the therapeutic response of advanced lung adenocarcinoma to EGFR- Tyrosine kinase inhibitors (TKIs) treatment. Methods: A retrospective analysis was performed on 253 patients diagnosed as advanced lung adenocarcinoma, who underwent EGFR mutation detection, and those with EGFR sensitive mutation were treated with TKIs. Using the Lasso regression model and the 10 fold cross-validation method, the radiomic features of predicted EGFR mutation status and the screening of TKIs for sensitive populations were obtained. 715 radiomic features were extracted from unenhanced, arterial phase and venous phase, respectively. Results: The area under curve (AUC) values of the multi-phases including unenhanced, arterial phase and venous phase of the EGFR mutation status validation group were 0.763, 0.807 and 0.808, respectively. The number of radiomic features extracted from the multi-phases were 5, 18 and 23, respectively, which could distinguish the EGFR mutation status. The AUC values of the multi-phases of the EGFR-TKIs sensitive validation group were 0.730, 0.833 and 0.895, respectively. The number of radiomic features extracted from the multi-phases were 3, 7 and 22, respectively, which can be used to screen the superior population for TKIs treatment. The efficiency of radiomic features extracted from venous phase in predicting EGFR mutant status and EGFR-TKIs sensitivity was significantly superior than those of unenhanced and arterial phase. Conclusions The radiomic features of CT scanning can be used as the radiomics biomarker to predict the EGFR mutation status of lung adenocarcinoma and to further screen the dominant population in TKIs therapy, which provides the basis for targeted therapy.

Objective: To establish radiomics models based on different CT scaning phases to distinguish mediastinal metastatic lymph nodes in NSCLC and to explore the diagnostic efficacy of these models. Methods: The CT images of 86 preoperative patients with NSCLC who were performed both plain and enhanced CT scans were analyzed retrospectively. The 231 mediastinal lymph nodes were enrolled in this study which were divided into two independent cohorts: 163 lymph nodes enrolled from January 2015 to June 2017 constituted the training cohort, and 68 lymph nodes enrolled from July 2017 to June 2018 constituted the validation cohort. The regions of interest (ROIs) were delineated on plain scan phase, arterial phase and venous phase CT images respectively, and 841 features were extracted from each ROI. LASSO-logistic regression analysis was used to select features and develop models. The area under the ROC curve (AUC value), sensitivity, specificity, accuracy, positive predictive value and negative predictive value of different models for distinguishing metastatic lymph nodes were compared. Results: A total of 6 models were established, and the AUC values were all greater than 0.800. The plain CT model yielded the highest AUC, specificity, accuracy and positive predictive value with 0.926, 0.860, 0.871, 0.906 in the training cohort and 0.925, 0.769, 0.882, 0.870 in the validation cohort. When plain and venous phase CT images were combined with arterial phase CT images, the sensitivity and negative predictive value of the models increased from 0.879, 0.821 and 0.919, 0.789 to 0.949, 0.878 and 0.979, 0.900 respectively. Conclusions The CT radiomics model could be used to assist the clinical diagnosis of lymph nodes. The AUC value of the model based on plain scanning was the highest, while the sensitivity and negative predictive value of the model could be improved by combining the arterial phase CT images.