Search

• Pelvis auto-segmentation tools were compared, from atlas-techniques to deep learning. • Deep learning showed better and faster segmentation results. • Differences found in cervical cancer tapered off in prostate cancer contouring. • Expert radiation oncologists qualitatively preferred deep learning. • Automated tools allow for faster radiotherapy workflow and high-quality treatments. Artificial intelligence can standardize and automatize highly demanding procedures, such as manual segmentation, especially in an anatomical site as common as the pelvis. This study investigated four automated segmentation tools on computed tomography (CT) images in female and male pelvic radiotherapy (RT) starting from simpler and well-known atlas-based methods to the most recent neural networks-based algorithms. The evaluation included quantitative, qualitative and time efficiency assessments. A mono-institutional consecutive series of 40 cervical cancer and 40 prostate cancer structure sets were retrospectively selected. After a preparatory phase, the remaining 20 testing sets per each site were auto-segmented by the atlas-based model STAPLE, a Random Forest-based model, and two Deep Learning-based tools (DL), MVision and LimbusAI. Setting manual segmentation as the Ground Truth, 200 structure sets were compared in terms of Dice Similarity Coefficient (DSC), Hausdorff Distance (HD), and Distance-to-Agreement Portion (DAP). Automated segmentation and manual correction durations were recorded. Expert clinicians performed a qualitative evaluation. In cervical cancer CTs, DL outperformed the other tools with higher quantitative metrics, qualitative scores, and shorter correction times. On the other hand, in prostate cancer CTs, the performance across all the analyzed tools was comparable in terms of both quantitative and qualitative metrics. Such discrepancy in performance outcome could be explained by the wide range of anatomical variability in cervical cancer with respect to the strict bladder and rectum filling preparation in prostate Stereotactic Body Radiation Therapy (SBRT). Decreasing segmentation times can reduce the burden of pelvic radiation therapy routine in an automated workflow. [ABSTRACT FROM AUTHOR]

Purpose In this study three digital breast tomosynthesis (DBT) devices equipped with different flat panel detectors (FPD) and a prototype equipped with a scan-spectral photon counting detector (PCD) are compared. The aim is to evaluate if a PCD device can guarantee a comparable or even better image quality delivering lower radiation doses than FPD devices. Methods Image quality performances were evaluated exposing two image quality DBT phantoms, Tomophan and Agatha and a homemade phantom for the evaluation of low contrast detectability (LCD). Exposures were performed in automatic exposure control (AEC) mode, both with standard (ST) and high resolution (HR) modality when available. The images with the lowest achievable post-processing were analyzed. Geometric distortion, signal-difference-to-noise ratio (SDNR), LCD with the statistical method, in-plane resolution in terms of modulation transfer function (MTF) in scan and sub-scan directions, slice-sensitivity profile (SSP) and artifact spread function (ASF) were evaluated. Results Geometric distortion resulted comparable between PCD and FPD devices. SDNR for PCD was comparable to two of the FDP devices. PCD device showed the best LCD. In-plane resolution was higher for PCD device: FPD devices exhibit a reduction in MTF 0.1 between 38% and 62%, both in scan and sub-scan direction. SSP evaluation showed comparable results for PCD and one of the FPD devices, while it was up to 60% worse for the others. PCD device showed a slightly worse ASF compared to FPD devices. Delivered dose in AEC with PCD device was 50% lower than doses delivered with FPD devices. Conclusions PCD device ensures a comparable or even better image quality than FPD devices, especially for spatial resolution and detectability of low contrast objects, delivering much lower radiation dose. Photon counting seems to be really an impressive approach to DBT. [ABSTRACT FROM AUTHOR]

Abstract Purpose To evaluate the effect of plan complexity parameters in patient-specific quality assurance (QA). Methods Evaluation of 902 clinical VMAT plans, optimized using a 3° control point separation, was achieved by local gamma index analysis (3%, 3 mm) between delivered and calculated dose. Confidence limit (CL) for 93.5% interval (p = 0.065) was evaluated too [1]. All plans were delivered by a 6MV Linac to a biplanar diode array for patient-specific QA. Plan complexity was investigated throughout different parameters: average leaf travel (LT), average in-field leaf travel (LTi) [2] , modulation complexity score applied to VMAT (MCSv) and a multiplicative combination of them (LTMCS) [3]. Pearson's correlation analysis was performed between gamma passing rates, CL and each parameter. Results 98.45% of plans was characterized by a gamma passing rates above 90%, while only 14 plans yielded values under the acceptance threshold. CL resulted to be more restrictive: 95,34% of plans satisfied the passing criteria, while 42 plans presented CL > 1. Nevertheless, gamma passing rates and CL showed a strong correlation (Pearson absolute value above 0.93). Plan complexity parameters underlined a moderate correlation with gamma passing rates, with an absolute value between 0.39 and 0.47. More deeply, a negative correlation was observed with LT, while MCS and LTMCS presented a positive correlation. CL showed similar results characterized by an absolute Pearson factor between 0.47 and 0.51. Results of Pearson analyses are summarized in Table 1. Pearson Coefficient Gamma passing rate CL LT −0.47 (−0.42–−0.52) 0.51 (0.46–0.55) LTi −0.44 (−0.38–−0.48) 0.47 (0.42–0.52) MCSv 0.39 (0.34–0.45) −0.47 (−0.42–−0.52) LTMCS 0.44 (0.39–0.49) −0.51 (−0.46–−0.55) Conclusions Local gamma analysis results underlined a significant correlation with plan complexity parameters, with larger correlations obtained with CL. [ABSTRACT FROM AUTHOR]

IMRT is recognized as a powerful technique to improve quality of dose distribution in head and neck (HN) radiotherapy. To compare Constant and Variable Dose Rate VMAT (CDR,VDR respectively) to Step-and-Shoot IMRT (S&S-IMRT) for treatment of HN cancer. S&S-IMRT, CDR and VDR plans were created for 10 patients. 3 PTVs were defined (prescribed doses 66 Gy, 60 Gy and 54 Gy, 30 fr). OARs included mandible (MA), spinal cord (SC), brain stem (BS), parotids (PA), salivary glands (SG), esophagus (ES), larynx (LA) and thyroid (TH). S&S-IMRT plans consisted of seven 6 MV coplanar beams; CDR and VDR plans consisted of two 6 MV arcs (182° to 178°; clockwise and counterclockwise direction). All plans were performed with Pinnacle3 TPS (v 9.8). Dose distributions were compared by evaluating PTVs' D mean , D 2% , D 50% , D 98% and Homogeneity Index (HI); for MA, SC, BS D 2% ; for PA, SG, ES, LA and TH D mean and a number of different dose-volume data in V 20 Gy – V 50 Gy range. To compare efficiency, MUs and treatment delivery time were evaluated. Concerning PTVs coverage, all techniques were equivalent although VMAT ones significantly improved HI and decreased D 2%. For OARs significantly better results were found for VMAT techniques with reduction in the volume of LA, TH and ES receiving medium and high doses. Compared with S&S-IMRT, VMAT techniques reduced delivery times although MUs were higher. VMAT techniques offer better results and CDR is an additional option of rotational arc radiotherapy for linacs without variable dose rate with a lower cost. [ABSTRACT FROM AUTHOR]