2025 - Automated Clinical Target Volume to Improve Efficiency of Contouring for Spine Metastasis Patients Treated with Stereotactic Body Radiation Therapy

03:00pm - 04:00pm ET

Poster Hall - Exhibit Hall A

Screen: 14

POSTER

Presenter(s)

Gregory Buti, PhD - Massachusetts General Hospital, Winchester, MA

G. Buti¹, A. Ajdari², C. Bridge¹, F. Lofman³, G. Sharp¹, A. E. Marciscano⁴, and T. R. Bortfeld¹; ¹Massachusetts General Hospital, Boston, MA, ²Department of Radiation Oncology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, ³RaySearch Laboratories AB, Stockholm, Sweden, ⁴Department of Radiation Oncology, Mass General Brigham / Massachusetts General Hospital, Boston, MA

Purpose/Objective(s):

Spine stereotactic body radiation therapy (SBRT) is an emerging standard of care for patients with symptomatic spinal metastatic disease. The accuracy of target volume delineation is of critical importance for safe and effective spine SBRT. The adjacent normal tissues, including the spinal cord, are often closely approximated to the target and inadequate target coverage can result in local failure, whereas excessive target coverage can result in unnecessary and serious toxicity. Algorithmic tools can assist in improving the accuracy and efficiency of target delineation practices. We developed an automated method for CTV expansion based on internationally published guidelines from the International Spine Radiosurgery Consortium (ISRC).

Materials/Methods:

Following ISCR guidelines, the bony expansion of the automated CTV includes the elective vertebral segments located near the segment(s) containing the lesion. Elective vertebral segments are included in CTV delineation on a patient-specific basis using a three-step process: (1) segmentation of the vertebrae on the patient's CT scan using a deep learning model from RayStation (research version 2025); (2) estimation of the patient's spinal curvature to accurately define the vertebral segments; and (3) inclusion of potentially involved vertebral segments in a lesion-to-CTV expansion algorithm. The automated method was tested on eleven patients with various primary cancers, including breast, colorectal, pancreatic, lung and renal cancers. Eight of the patients had metastases in the thoracic spine and three had metastases in the lumbar spine.

Results:

The accuracies of the lesion-containing vertebrae and the automated CTVs were evaluated by retrospectively comparing the contours to manual physician contours. The accuracy metrics were Dice Similarity Coefficient (DSC) for the volume overlap and Surface Dice Coefficient (SDSC) with a tolerance of 2 mm for surface similarity. On average (± standard deviation), the DSC for the vertebrae was 87.0% ± 7.6%, and SDSC was 91.9% ± 7.0%, and the DSC and SDSC for the automated CTV were 83.0% ± 9.9% and 79.8% ± 11.4%, respectively. Generating an autodelineated CTV took less than one minute on average, compared to an estimated 20 minutes for manual CTV delineation.

Conclusion:

We propose an automated approach to delineating the CTV for bony spine expansion that uses a sub-1-minute computational pipeline to calculate elective spine segments. This approach has the potential to support CTV delineation practices within the emerging treatment paradigm of spine SBRT.