Verification of field match lines in whole breast radiation therapy using Cherenkov imaging

Abstract

Purpose: In mono-isocentric radiation therapy treatment plans designed to treat the whole breast and supraclavicular lymph nodes, the fields meet at isocenter, forming the match line. Insufficient coverage at the match line can lead to recurrence, and overlap over weeks of treatment can lead to increased risk of healthy tissue toxicity. Cherenkov imaging was used to assess the accuracy of delivery at the match line and identify potential incidents during patient treatments.

Methods and materials: A controlled calibration was constructed from the deconvolved Cherenkov images from the delivery of a modified patient treatment plan to an anthropomorphic phantom with introduced separation and overlap. The trend from this calibration was then used to evaluate the field match line for accuracy and inter-fraction consistency for two patients.

Results: The intersection point between matching field profiles was directly correlated to the distance (gap/overlap) between the fields (anthropomorphic phantom R² = 0.994 "breath hold" and R² = 0.990 "free breathing"). The profile intersection points from two patients' imaging sessions yielded an average of +1.40 mm offset (overlap) and -1.32 mm offset (gap), thereby introducing roughly a 25.0% over-dose and a -23.6% under-dose (R² = 0.994).

Conclusions: This study shows that field match regions can be detected and quantified by taking deconvolved Cherenkov images and using their product image to create steep intensity gradients, causing match lines to stand out. These regions can then be quantitatively translated into a dose consequence. This approach offers a high sensitivity detection method which can quantify match line variability and errors in vivo.

Keywords: Breast; Cherenkov; Field Match; Imaging; In vivo; Radiation Therapy.

Fig. 1.

Deconvolution and Profile Extraction. In (a), an ROI is taken from an input image and vertically summed to create the edge spread function (ESF) in (b) which is differentiated to view the line spread function (LSF) in (c). A 2D gaussian was fit using the LSF as a cross-section (d) to yield the point spread function, (PSF). The PSF is then used to deconvolve the left field (a) and right field (not shown), to get left field outputs (e) and right field outputs (f). The product and square root of both output images forms the match or product image (g). An intensity threshold was applied to create the yellow mask in (h) where automated techniques isolated the rectangular ROI (red) circumscribing the threshold mask. This region becomes the mask for the output images in (i), resulting in their respective profiles (j). Match amplitude is indicated by central crosshair. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.

Anthropomorphic “Annie” Phantom Simulated Treatment Setup. Subfigure (a) provides a coronal-view screenshot of the delivered treatment plan, consisting of two supraclavicular beams (AP/PA), shown superior to the tangent fields (RPO/LAO), meeting at isocenter (red). In (b), the sagittal view (TPS) is shown. The plan information was exported from the TPS and rendered in the view of the Cherenkov camera in (c) for supraclavicular fields and (d) for tangent fields. Their respective, cumulative Cherenkov images are shown in (e) and (f), isocenter circled in yellow. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3.

Analysis of anthropomorphic phantom match lines. The Cherenkov match images and field sums were transformed to an en face view for fields with added (a) separation, (b) true match, and (c) overlap. These images have been rotated 90 degrees, column summed, and normalized such that the tangent field (blue) summed with the supraclavicular field (orange) yields the profile sum (black), with match/intersection point denoted by crosshair. In (d) a fit was applied between overlap distance and match point height for breath hold treatment (blue), and free breathing treatment (orange). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4.

Patient 1 Treatment and Analytics. In (a), both RPO and LAO supraclavicular (SCV) fields from the treatment plan are shown. In (b), the same is shown for all four tangent fields RPO/LAO 6X/10X summed. The match (product) image is shown in (c), followed by the treatment plan profile (d). In (e) the supraclavicular Cherenkov image and (f) the tangent fields Cherenkov image are shown from day one of the patient’s treatment. Likewise, the Cherenkov match image (g) and the respective profile are shown (h). In (i), the Cherenkov images and profiles are shown for the next four fractions of treatment. In (j) the ratio of the Cherenkov image match amplitude over the TPS match amplitude is plotted with respect to recorded treatment day.

Fig. 5.

Patient 2 Treatment and Analytics (Towel Bolus). In (a) the projected treatment plan is shown for the SCV fields AP/PA, and likewise for the summation of all four tangent beams in (b) RPO/LAO 6X/10X. Respective Cherenkov images are shown in (e) and (f). The treatment plan match image is shown in (c), and the Cherenkov match image is shown in (g). The tangent profile (blue), SCV profile (orange) and their sum (black) are shown for the treatment plan in (d), and for the Cherenkov images from the first fraction in (h). The following four fractions are shown in (i) and, their match amplitudes are normalized to the treatment plan match amplitude found in (d), plotted with respect to recorded fraction in (j). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6.

Surface Dose Consequence due to Gap/Overlap. In (a), surface dose maps of introduced gap show true match (0 mm), followed by the dose deficiency (underdose) from to 1 mm, 2 mm, 3 mm, 5 mm, and 10 mm of separation. Summed profiles are shown under each respective surface dose map. In (b), the same format is shown for overlap, which illustrates overdosing. In (c), the consequential percent overdose or underdose is mapped with respect to distance in mm, fit using the error function with error bars representative of the RMSE.