ESTRO 2025 - Abstract Book

S2669

Physics - Detectors, dose measurement and phantoms

ESTRO 2025

Ingbert, Germany. 6 Information Engineering, University of Florence, Florence, Italy. 7 Electrical Engineering, KU Leuven, Gent, Belgium

Purpose/Objective: Superheated injectable nanodroplets (NDs) were proposed for in vivo proton range verification, as they vaporize into echogenic microbubbles upon exposure to ionizing radiation [1,2,3]. Ultrafast ultrasound imaging allows to super-localize sparse vaporizations using active [4] or passive [5] Ultrasound Localization Microscopy (ULM and P ULM, respectively). 2D-ULM can simultaneously image ND vaporizations and surrounding tissues, but is sensitive to flow as it cannot separate new vaporizations and microbubbles flowing in and out of the image plane. In comparison, P-ULM specifically detects acoustic signatures from individual vaporizations [5]. To verify the radiation beam position in 3D, we extend here both techniques to volumetric imaging and compare their range verification performances. Material/Methods: NDs (PVA shell, C 4 F 10 core [2]) were dispersed in water at 50°C and irradiated with a 62 MeV proton beam (Fig 1). Active and passive ultrasound acquisitions were performed during irradiation using a 2 MHz square matrix array [6]. The peak dose was 4 and 2 Gy delivered in 4 and 6 seconds, for passive and active acquisitions, respectively. The flow speed was constant (250 rpm) for passive acquisitions (n=3), while the magnetic stirrer speed varied for active acquisitions (3 samples at 0, 100 and 250 rpm). A previous P-ULM pipeline was extended to 3D to super-localize vaporizations from passive acquisitions [5]. Active acquisitions were filtered to retain sparse microbubbles which were detected and tracked with a standard ULM algorithm. 3D vaporization maps were built using the start positions of every track (i.e. the vaporization event), registered to room coordinates and compared to an absolute range measurement.

Results: Fig. 2 shows 3D vaporization maps overlaid on the proton stopping distribution for a passive and active acquisition. Most vaporizations appear at the Bragg peak as primary protons reach their peak LET, while high-LET secondaries cause vaporizations proximal to the Bragg peak [2,4,5]. P-ULM localized more vaporizations than ULM (19.3 ± 2.8 and 4.0 ± 0.2 events /10 8 protons/µM of NDs, respectively). ULM suffered from false localizations distal to the Bragg peak worsening with increasing flow. The proton range estimation error was 0.98 ± 0.04 mm for P-ULM and 0.70 ± 0.02 mm for ULM, i.e. below the combined uncertainties of the absolute range measurement and registration to room coordinates.