Objective To assess the variation in patient setup corrections for three different anatomic treat-ment sites using daily pretreatment megavoltage CT(MVCT)in helical tomotherapy,and to analyze alternative refer-ence margins for specific tumor site.Methods Sixty patients treated for three anatomical sites on helical tomotherapy were analyzed.Daily MVCT was carried out for registration and setup corrections before each treatment fraction.Setup errors and rotational setup corrections from 587 head and neck,500 thoracic,371 abdomen and gynecology MCVT imaging were input to statistical analysis.Incidences of three dimensional vector error lengths were calculated for each anatomic site.Overall distribution histograms of the three -dimensionally error was presented using GraphPad Prism 5.The distributions of systematic and random setup errors were quantitative analyzed and the additional margins required were also taken into account.Results 1 458 MVCT scans were implemented for sixty patients.Head and neck had lower frequencies of translational setup errors than others.Frequency of at least 3mm three -dimensional setup errors for head and neck,thoracic,abdomen and gynecology was 55.3%,70.8%,79.8%,respectively.This fre-quency decreased to 17.5%,40.6%,47.2% if 3D vector distance ≥5mm was scored.Overall systematic errors ranged from -1.4mm to 2.7mm,abdomen and gynecology had the largest setup errors in the vertical direction which was statistically significant(χ2 =19.3,P <0.05).The suggested margins should be increased by 4 -7 mm in three -dimensional direction for head and neck,9 -14mm for thoracic,10 -17mm for abdomen and gynecology.Conclusion Differences in setup corrections are perceived between head and neck,thoracic,abdomen and gynecology.The accura-cy of patient positioning can be improved if pre -treatment daily MVCT scans are put into use.Results from setup cor-rection can provide evidence for tumor treatment margin and improve the accuracy of regular radiotherapy.

Objective To investigate the distribution rules of setup errors in different locations for tomotherapy.Methods 151 patients induding 53 head and neck tumors,45 thoracic tumors,20 abdominal tumors,and 33 pelvic tumors,who accepted tomotherapy were retrospectively analyzed in this study.The planning CT images of patients were obtained in simulation,and all patients underwent megavoltage CT (MVCT) scan before radiotherapy.And the setup errors were calculated by rigid registering MVCT images to planning CT images,and setup errors on + x(left),-x(right),+ y(in),-y(out),+z(ventral),-z (dorsal)axes were analyzed respectively.Results A total of 3 281 MVCT scans were performed on 151 patients,The setup errors on +x (left),-x(right),+y(in),-y(out),+z (ventral),-z (dorsal)axes were (1.61 ± 1.21),(1.76 ±2.11),(2.26 ± 1.74),(1.83 ± 1.47),(3.24±1.76) and (1.75 ± 1.61)mm for head and neck tumors;(2.43 ±1.88),(2.55 ± 1.92),(3.06 ±2.64),(3.90 ±2.91),(6.71 ±3.46) and (2.64 ±2.77)mm for thoracic tumors;(3.67±3.06),(2.37±1.77),(3.18±1.96),(3.98±3.01),(6.74±3.25) and (1.92±2.00) mm for abdominal tumors;(2.92 ±2.13),(2.17±1.68),(3.50±2.61),(3.72±2.66),(7.18± 3.43) and (1.92 ± 1.61)mm for pelvic tumors,respectively.The setup errors were different between +z and-z with statistically significant in all tumors (t =-4.119、-5.033、-3.763、-5.057,P ＜ 0.05).The setup errors on + z direction of patients immobilized with thermoplastic mask were smaller than those immobilized with vacuum cushions for thoracic tumors (t =-2.357,P ＜ 0.05).Conclusions The setup errors of head and neck tumors are less than other parts tumor in tomotherapy.The patients immobilized with thermoplastic mask can reduce the setup errors for thoracic tumors.The heterogeneity of setup errors on ventral-dorsal directions for the all parts of tumors should not be ignored.

Objective To study the impact of heartbeat on the left ventricular myocardial (mLV) and the left anterior descending (LAD) coronary arterial dosimetry following intensity modulated radiotherapy (IMRT) for left-sided breast carcinoma.Methods A total of 15 female patients who received electrocardiography gated 4D-CT scan in inspiratory breathing-holding were enrolled.According to cardiac cycle,20 time-phase images were reconstructed from 0 to 95％ at 5％ interval.Then,the mLV and the LAD were delineated respectively and the IMRT plans of the left breast carcinoma were generated based on 0％ phase CT images.The volume and dice similarity coefficient (DSC) of the mLV were calculated,and the variation range of the mLV and the LAD dose-volume indexes were compared among different phases.Results The average rate of DSC variation of mLV was eightfold of volume,which reached to 472.07％,and the average change rate of mLV (Dmean) was up to 41.95％.The difference of the mLV's volume and DSC was statistically significant between the maximum and minimum(t =-6.585,-28.870,P＜0.05),as well as the D V10,V20,V30 and V40(t=-5.260,-4.084,-3.592,-3.273,-2.566,P ＜0.05).The average change rate of the LAD's D was up to 130.14％.The difference of D V10,V20,V30 and V40 of LAD was statistically significant between the maximum and minimum(t =-9.758,-8.810,-8.682,-7.853,-6.205,P ＜ 0.05).Conclusions The dosimetry impact of heartbeat on mLV and LAD should not be ignored.

Objective To evaluate the dynamic variation of the dose-volume parameters of the left ventricular myocardium following heart beat in radiotherapy for esophageal cancer. Methods The left ventricular myocardium of 22 patients was contoured on 20 phases (0％-95％) of electrocardiography (ECG) gated heart 4DCT images. The radiotherapy plan was designed on the simulation CT images,and then the dose distribution of radiotherapy plan was imported into MIM Maestro system and 4D dose-volume histogram (DVH) was reconstructed. The variations of position,volume and dice similarity coefficient (DSC) of the left ventricular myocardium were analyzed. The changing ranges of Dmean ,V10,V20 ,V30 and V40 of the left ventricular myocardium during different phases were statistically compared. Results ( 1 ) The biggest displacement of the left ventricular myocardium was in Y axes. The maximum variation rate of volume and DSC of the left ventricular myocardium were (24.23±11. 35)％ and (184.33±128. 61)％ in different phases with statistical significance (both P<0. 05).(2) The maximum variation rate of Dmean of the left ventricular myocardium was (87.05± 38. 34)％ in different phases with the highest rate of 163. 52％ with statistical significance (P<0. 05).(3) The maximum variation values of V10,V20,V30 and V40 of the left ventricular myocardium were (13.64±4. 33)％,(12.84±4. 55)％,(11.62±4. 85)％ and (3.63±2. 56)％ with statistical significance (all P<0. 05). Conclusions The impact of heart beat on the dose-volume parameters of the left ventricular myocardium should be considered during esophageal cancer radiotherapy. Traditional static 3DCT-based assessment of the dose-volume parameters of the left ventricular myocardium can yield relatively large errors, which is probably reduces the prediction efficiency of the dose-volume parameters for radiation-induced heart injury.

Objective:To evaluate the feasibility and clinical value of pre-treatment non-enhanced chest CT radiomics features and machine learning algorithm to predict the mutation status and subtype (19Del/21L858R) of epidermal growth factor receptor (EGFR) for patients with non-small cell lung cancer (NSCLC).Methods:This retrospective study enrolled 280 NSCLC patients from first and second affiliated hospital of University of South China who were confirmed by biopsy pathology, gene examination, and have pre-treatment non-enhanced CT scans. There are 136 patients were confirmed EGFR mutation. Primary lung gross tumor volume was contoured by two experienced radiologists and oncologists, and 851 radiomics features were subsequently extracted. Then, spearman correlation analysis and RELIEFF algorithm were used to screen predictive features. The two hospitals were training and validation cohort, respectively. Clinical-radiomics model was constructed using selected radiomics and clinical features, and compared with models built by radiomics features or clinical features respectively. In this study, machine learning models were established using support vector machine (SVM) and a sequential modeling procedure to predict the mutation status and subtype of EGFR. The area under receiver operating curve (AUC-ROC) was employed to evaluate the performances of established models.Results:After feature selection, 21 radiomics features were found to be efffective in predicting EGFR mutation status and subtype and were used to establish radiomics models. Three types models were established, including clinical model, radiomics model, and clinical-radiomics model. The clinical-radiomics model showed the best predictive efficacy, AUCs of predicting EGFR mutation status for training dataset and validation dataset were 0.956 (95% CI: 0.952-1.000) and 0.961 (95% CI: 0.924-0.998), respectively. The AUCs of predicting 19Del/L858R mutation subtype for training dataset and validation dataset were 0.926 (95% CI: 0.893-0.959), 0.938 (95% CI: 0.876-1.000), respectively. Conclusions:The constructed sequential models based on integration of CT radiomics, clinical features and machine learning can accurately predict the mutation status and subtype of EGFR.

Objective To perform 3D dose reconstruction based on electronic portal imaging device ( EPID) of linear accelerator for the static intensity-modulated using Edose, a dose verification system, Aiming to assist the radiotherapy professionals to better understand the radiotherapy organs at risk and target dose changes. Methods CBCT image was acquired for patients with head and neck cancer and thoracic cancer once a week for a total of six times. Subsequently,CBCT images and planning CT images were subject to rigid registration and exported to the Edose software. According to the setup error, EPID-based three-dimensional dose reconstruction was performed by using Edose software. The gamma passing rate and dose of different organs at risk ( OARs ) were analyzed and statistically compared. Results For patients with nasopharyngeal carcinoma,the intra-fractional Dmax of the spinal cord was more significantly fluctuated and higher compared with the planning dose, whereas the intra-fractional Dmax of the brainstem did not significantly fluctuate. The V30 of the parotid gland significantly changed with a maximum increase of 28. 69％ per fraction. For patients with thoracic tumors,the Dmax of the spinal cord was slightly changed,and the actual doses in the lung and heart were higher than the planning doses. The average deviation of the pulmonary V5 was up to 16. 99％ between the actual and planning doses with statistical significance ( P<0. 05).According to the analysis of gamma passing rate,significant dose changes occurring in the OARs were detected in the 16th fraction for the head and neck cancer and the 24th fraction for the thoracic neoplasms. Conclusions The dose changes in the OARs can be obtained by reconstructing the EPID-based 3D dose distribution using the Edose software for each fraction, which can better protect the OAR, enhance the coverage of target dose and provide certain reference for dose-guided and self-adaptive radiotherapy.