ESTRO 2025 - Abstract Book

S3063

Physics - Image acquisition and processing

ESTRO 2025

3784

Digital Poster MR-CT multimodal fusion in stereotactic radiosurgery for brain metastases: a comparison between manual and automatic outputs Valeria Faccenda 1 , Denis Panizza 1 , Riccardo Ray Colciago 2 , Stefano Mascellani 2 , Bianca Bordigoni 3 , Stefano Carminati 4 , Chiara Ingraito 4 , Elena De Ponti 1 , Stefano Arcangeli 2 1 Medical Physics, Fondazione IRCCS San Gerardo, Monza, Italy. 2 School of Medicine and Surgery, University of Milan Bicocca, Milan, Italy. 3 Physics and Astronomy, University of Padua, Padua, Italy. 4 Physics, University of Milan, Milan, Italy Purpose/Objective: To investigate the accuracy and quantify uncertainties of manual vs. automatic MR-CT registration in stereotactic radiosurgery (SRS) for brain metastases (BM). Material/Methods: Twenty MR-CT rigid registrations were independently performed twice by three expert operators, yielding a total of 120 analyzed fusions. For each patient, a mean reference fusion was derived by averaging the translational and rotational degrees of freedom across all operator fusions, with deviations from this reference used for further analysis. The range of the values were computed to assess inter- and intra-operator variability in registration accuracy. Thirty-nine BM volumes contoured on the MR images were mapped onto the corresponding CT scans based on each individual fusion, and centroid shift measured the impact of registration accuracy on GTV localization. The outputs of automatic fusions performed by MIM software using 3 different algorithms (standard, box-based, and contour-based) were compared to the manual reference registrations. Finally, four non-experts completed the same registrations, and the resulting uncertainty range and impact on GTV volumes were also analyzed. Results: Inter- and intra-operator variability was notable. The pitch demonstrated the highest variability, with a mean deviation of 0.56° and a range extending up to 2.9°. Translational mean deviations were <0.4 mm and a maximum range of 2.4 mm was observed in the longitudinal direction. Automated registration values consistently fell within the range of manual measurements. The contour-based algorithm exhibited the smallest mean variations, with translations deviating by less than 0.3 mm and rotations by less than 0.3°. Non-expert registrations exhibited a broader range of variability, with pitch deviations up to 4.9° and longitudinal shifts up to 5.2mm. The analysis of GTV localization revealed that different expert registrations resulted in a mean barycenter displacement of 0.8 mm (range: 0.1–2.8), with approximately 18% of GTVs displaced by more than 1 mm. The contour-based algorithm, by comparison, produced a mean barycenter displacement of 0.7 mm (range: 0.1–1.8) vs. 1.0 mm (0.2 – 3.1) of the other automated methods. Non-expert registrations yielded a higher mean barycenter displacement of 1.6 mm (range: 0.1–6.2). Conclusion: Inter- and intra-operator variability, along with operator experience, must be accounted for when planning high precision SRS treatments. Even experts showed GTV barycenter deviations exceeding 1 mm, prompting reconsideration of a “fusion gold-standard” and the use of sub-millimeter margins. Automated algorithms, particularly contour-based methods, showed the potential for reducing variability. Further investigations will assess variability and uncertainties with operator-revised automated outputs.

Keywords: Multimodal fusion, SRS, brain metastases