An evaluation of the Clarity 3D ultrasound system for prostate localization

Abstract

The purpose of this study is to evaluate the accuracy and precision of the Clarity 3D ultrasound system to track prostate gland positional variations due to setup error and organ motion. Seventeen patients (n = 17) undergoing radical external beam radiation therapy for localized prostate cancer were studied. Subsequent to initial reference ultrasound and planning CT scans, each patient underwent seven repeat weekly tracking CT and ultrasound (US) scans during the course of treatment. Variations in the location of the prostate between reference and tracking scans were measured. Differences reported by CT and ultrasound scans are compared. Ultrasound tracking was initially performed clinically by a group of trained general users. Retrospective prostate localization was then performed by a trained dedicated user upon the original raw data set and also a reduced data set derived from the original by an expert user from Resonant Medical. Correlation accuracy between ultrasound and CT shifts acquired and delineated by a pool of trained general users was deemed unacceptable for radiotherapy purposes. A mean discrepancy between CT and US localizations of greater than 10 mm, with a 5 mm or greater discrepancy rate of nearly 90%, was observed. Retrospective analysis by a dedicated user of both the original and Resonant Medical reduced data sets yielded mean CT-Us discrepancies of 8.7 mm and 7.4 mm, respectively. Unfortunately, the 5 mm or greater CT-US discord rate for these retrospective analyses failed to drop below 80%. The greatest disparity between CT and ultrasound was consistently observed in the superior-inferior direction, while greatest agreement was achieved in the lateral dimension. Despite an expert reanalysis of the original data, the Clarity ultrasound system failed to deliver an acceptable level of geometric accuracy required for modern radiotherapy purposes.

Figure 1

Scatter plots of ultrasound vs. CT shifts (mm) [O CT‐US1, • CT‐US2] in the AP, LAT, and SI directions for the initial dataset (1(a), 1(b), and 1(c), respectively), for the dedicated user dataset (1(d), 1(e), and 1(f), respectively), and for the RMr dataset (1(g), 1(h), and 1(i), respectively). Dotted lines represent $\text{CT}-\text{US}=\pm 5\hspace{0.17em}\text{mm}$.

Figure 2

Histograms of the Euclidean distance (mm) between CT and ultrasound shifts for the initial (a), dedicated user (b), and RMr (c) datasets.

Figure 3

Histograms of the distances between consecutive ultrasound localization determinations (US1–US2) in the AP, LAT, and SI directions for the initial dataset (3(a), 3(d), and 3(g), respectively), for the dedicated user dataset (3(b), 3(e), and 3(h), respectively), and for the RMr dataset (3(c), 3(f), and 3(i), respectively).

Figure 4

Histograms of the Euclidean distance US1–US2 (mm) between two consecutive ultrasound localization determinations (US1–US2) for the initial (a), dedicated user (b), and RMr (c) datasets.

Figure 5

Representative images drawn from the RMr dataset showing the prostate localization provided by a tracking CT scan (a) and its two associated ultrasound tracking scans: US1 (b) and US2 (c).