Somatom definition as 40

We selected 250 pN2-NSCLC patients who received PORT in our department between 2012 and 2016 to build the deep learning model, with 200 patients randomly assigned to a training set and 50 to the validation set. We then selected 19 patients treated between 2016 and 2018 in our department as the test cohort to generate the automatic delineation of the CTV. All patients were simulated in the supine position with both arms raised above the head. The thickness of all scanned slices was 5 mm, and all computed tomography (CT) images were transferred using the Digital Imaging and Communications in Medicine (DICOM) format. The CT data were acquired on a Somatom Definition AS 40 (Siemens Healthcare, Forchheim, Germany) or on Brilliance CT Big Bore (Philips Healthcare, Best, the Netherlands) systems.

We retrospectively selected 150 NPC patients treated in our hospital between January 2016 and May 2019. The patient demographics are shown in Table 1. The patients with locally advanced cancer (N = 53) were treated with induction chemotherapy followed by concurrent chemoradiotherapy, and the remaining patients were treated with either radiotherapy or concurrent chemoradiotherapy. During CT simulation, patients were immobilized in supine position with a thermoplastic mask and underwent contrast-enhanced CT scan on a Somatom Definition AS 40 (Siemens Healthcare, Forchheim, Germany) system. The dimension, resolution, and thickness of CT images were 512 × 512, 0.98, and 2.5 mm, respectively. To better delineate the tumor region, T1-weighted MR images were also acquired and fused with CT images. The CTVp1 was delineated by experienced radiation oncologists on the CT images in a Pinnacle TPS (Philips Radiation Oncology System, Fitchburg, WI, USA) following the international guideline for NPC CTVp1 delineation (5 (link)).

Data of 91 patients with prostate cancer were collected in this study. The planning CT images and daily CBCT images were acquired and registered. The planning CT images were acquired with a CT simulator (SOMATOM Definition AS 40, Siemens or Brilliance CT Big Bore, Philips) with the following parameters: voltage: 120 kV; exposure: 280 (Siemens) or 240 (Philips) mAs; image resolution: 512 × 512; pixel size: 1.27 × 1.27 mm²; slice thickness: 3 mm. The CBCT images were scanned on a Varian On-board Imager with the following parameters: voltage: 125 kV; exposure: 1,080 mAs; rotation range: 360°; projections: 900 frames; image resolution: 512 × 512; pixel size: 0.91 × 0.91 mm²; slice thickness: 1.91 mm. The radiotherapy was implemented on a Varian Edge radiosurgery system. Deformable registration was implemented using the MIM software (v.7.0.1, MIM Software Inc., Cleveland, OH, USA) to make the planning CT images paired to the CBCT images. The deformed CT images were resampled to have the same spatial resolution and pixel size as the reference CBCT images. The gray value of pixels outside the patient body was set to zero to avoid background influence. Institutional Review Board approval was obtained for this retrospective analysis.

Ethical approval for this retrospective study was granted by the local institutional review board (EK 028-19). A total of 34 abdominal CT scans of patients with age between 20 and 80 years, acquired in the time between 01/2018 and 03/2018, were randomly selected from the hospital’s picture archiving and communication system (PACS). Inclusion criteria were venous contrast phase with a slice thickness of 5 mm. All CT examinations were performed in the Department for diagnostic and interventional Radiology, University Hospital RWTH Aachen, using either a 128-row multidetector CT scanner (Somatom Definition Flash; Siemens Medical Systems, Erlangen, Germany) or a 40-row multidetector CT scanner (Somatom Definition AS 40; Siemens Medical Systems, Erlangen, Germany). Tube voltage was 120 kV with a pitch of 0.6, and tube current was modulated according to Siemens CareDose4D. Images were reconstructed using filtered back projection. The anonymized CT-exams were extracted from the PACS, and the bilateral PMMs were manually segmented by a board-certified radiologist (6 years of experience) using ITK-SNAP (www.itksnap.org), an open-source software tool used to segment structures in 3D medical images [15 ]. Therefore, a total of 68 single psoas segmentations were available. Per sample, the number of slices varied between 76 and 139 with a scan matrix of 512 × 512 pixels.

Helical computed tomography (CT) scans (SOMATOM Definition AS40; Siemens) were performed under the following condition: a scan and reconstruction slice thickness of 5 mm. The patients were scanned in upper and lower segment, respectively. The upper segments were scanned from the skull to 10 cm below the boundary, while the lower segments scanned from the toes to 10 cm above the boundary.

The data consisted of 19 stage III/IV esophageal cancer patients who were treated from December 2018 to July 2019 in our department. The inclusion criteria of patients were proven and diagnosed histologically as esophageal cancer according to the guideline of the TNM staging system. The detailed demographics of the included patients are shown in Table 1. All patients were set up with the supine position on a commercial “bellyboard” and immobilized using a thermoplastic mask. The data of planning computed tomography (CT) images were acquired from the Somatom Definition AS 40 (Siemens Healthcare, Forchheim, Germany) or the Brilliance CT Big Bore (Philips Healthcare, Best, the Netherlands) systems on helical scan mode. CT images were reconstructed using a matrix size of 512 × 512 and a slice thickness of 5 mm. The delineation of OARs was delineated on CT images according to RTOG 0617 and RTOG 1106 standard contouring atlas (8 , 9 ). Meanwhile, the delineation of OARs was delineated and approved by senior clinicians for this study.