Three-dimensional transrectal ultrasound guided high-dose-rate prostate brachytherapy: A comparison of needle segmentation accuracy with two-dimensional image guidance

Toward a 3D transrectal ultrasound system for verification of needle placement during high‐dose‐rate interstitial gynecologic brachytherapy

“…For the patient study, the mean Euclidean distance AE SD between corresponding needle tips in 3D TRUS and CT was 3.82 AE 1.86 mm with 32.8% of identifiable needles having differences less than 3 mm and 70.7% of identifiable needles having differences less than 5 mm for the total of 58 identifiable needles used in five patients. This is comparable to values reported by Hrinivich et al, 38 when comparing 2D and 3D TRUS for visualizing needles during HDR prostate brachytherapy, reporting tip differences of less than 3 mm for 33% of needles and less than 5 mm for 64% of needles in a study of 12 patients undergoing a similar imaging procedure and evaluating 183 needles. In the needle template coordinates, the direction with the lowest mean absolute difference was the needle insertion direction with a maximum over all patients of 1.92 AE 2.03 mm (Patient D), but had the greatest variability during treatment since the other two directions are physically constrained by the template.…”

Simultaneous automatic segmentation of multiple needles using 3D ultrasound for high-dose-rate prostate brachytherapy

“…Figure shows example segmentations produced for two patients, indicating the 3D needle tip distances between manual and automatic segmentations. Treatment planning studies recommend needle tip localization accuracy within 3 mm 2 and previous HDR‐BT imaging studies described tip errors in terms of 3 mm and 5 mm thresholds . The geometric performance of the algorithm is summarized in terms of these thresholds in Table .…”

“…SR3D image spatial resolution varies with distance from the probe, with the highest spatial resolution component in the direction parallel to the transducer ( z ), and the lowest spatial resolution component in the reconstructed direction tangential to the direction of probe rotation ( t ). The device calibration and HDR‐BT workflow have been described previously . Following the procedure, a medical physicist manually segmented and labeled all needles present in the SR3D images using Brachyvision treatment planning software (Varian Medical Systems, Palo Alto CA, USA).…”

“…Transrectal ultrasound (TRUS)‐guided HDR‐BT techniques have been developed, enabling the segmentation of needles using multiple live 2D sagittal images and volumetric segmentation of organs using contiguous sets of axial images, thereby eliminating the need for patient repositioning between imaging and treatment . We recently completed a study investigating 3D TRUS‐guided HDR‐BT based on robotic sagittally reconstructed 3D (SR3D) images, which are acquired by rotating the probe using a motor while simultaneously tracking probe position and acquiring sagittal images in a fan geometry . This method of image reconstruction mitigates tip localization uncertainty by eliminating the need to switch between sagittal and axial transducers or move the probe in the superior/inferior direction .…”

“…We recently completed a study investigating 3D TRUS‐guided HDR‐BT based on robotic sagittally reconstructed 3D (SR3D) images, which are acquired by rotating the probe using a motor while simultaneously tracking probe position and acquiring sagittal images in a fan geometry . This method of image reconstruction mitigates tip localization uncertainty by eliminating the need to switch between sagittal and axial transducers or move the probe in the superior/inferior direction . Increasing needle and organ segmentation accuracy remains a primary objective in HDR‐BT imaging, but all imaging and segmentation must be performed intraoperatively, typically while the patient is under general anesthesia, adding a time constraint to the procedure.…”