ESTRO 2025 - Abstract Book

S2963

Physics - Image acquisition and processing

ESTRO 2025

Purpose/Objective: 4DCT imaging is essential for target definition and safety margin setup in 4D SBRT; however, breathing-induced artifacts can impede treatment planning and affect clinical outcomes, such as local control [1]. Online breathing signal-guided scanning protocols, such as the intelligent 4DCT (i4D) protocol, reduce breathing artifacts, as shown by phantom and case studies. This study determines artifact frequency and local metastasis control for comparison between clinical i4DCTs, acquired using the i4D protocol, and conventional spiral scanning (c4DCTs). Material/Methods: This retrospective study includes 103 in-house stereotactic treatments of thoracic metastasis (>96 Gy BED surrounding the ITV). For each treatment, a 10-phase 4DCT scan for treatment planning was acquired. The 4DCT scans were divided into two cohorts based on scanning type: (1) i4DCT, including 27 scans from a SOMATOM go.Open Pro scanner (Siemens Healthineers AG), and (2) c4DCT, with 76 scans from a Siemens Definition AS+ scanner (Siemens Healthineers AG). A deep learning-based convolutional neural network [2] was employed to classify each axial slice within a phase image as either “artifact-affected” or not. Neighboring “artifact-affected” slices were grouped as individual artifacts. Each artifact was further categorized by the network as either double structure (DS) or interpolation (INT). For each 4DCT, the total number of detected artifacts was calculated by summing the artifacts identified across all 10 phase images. Local control (LC) was assessed for each metastasis 12 months post-treatment, with locally controlled defined as complete remission or stable tumor size based on CT, MRI, or PET imaging.

Results: Figure 1 shows examples of detected INT and DS artifacts. Table 1 lists the frequency of detected artifacts and LC for both cohorts.

While INT artifacts were only observed in c4DCTs, DS artifacts were present in both cohorts. In the i4DCT cohort, fewer DS artifacts per 4DCT were detected compared to the c4DCT cohort. Both cohorts demonstrated high variability in the number of detected DS artifacts per 4DCT, indicating that their presence is specific to individual scans and may be primarily attributable to breathing irregularities.