3465 - Feasibility of a New Paradigm: Patient-Driven Radiation Therapy Using Smart-Phone Photogrammetry

L. Zhao¹, S. Klawikowski¹, X. Chen², M. J. Awan², and E. S. Paulson²; ¹Medical College of Wisconsin, Milwaukee, WI, ²Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI

Purpose/Objective(s): Accurate representation of patient anatomy is essential for radiation therapy. Surface changes arising from weight gain/loss or tissue edema occurring during a treatment course can currently only be detected by on-treatment imaging modalities during a clinical visit. However, delayed detection of surface changes may result in postponed treatments due to the need to generate a new treatment plan. We investigate here the feasibility of a new paradigm: patient-driven radiation therapy, in which patients can self-perform 3D surface scans using affordable smartphone photogrammetry. The surface-scanned models can then be compared to reference surfaces obtained from planning images to detect surface changes, triggering the need for an adaptive plan. Materials/

Methods: An initial feasibility study was performed using an anthropomorphic head phantom and two healthy volunteers. The combination of an iPhone 15 Pro Max (Apple Inc., USA) and smartphone application Polycam were used to create 3D surface models using photogrammetry. Ground truth surface contours of the phantom and volunteers were obtained with CT and MRI, respectively. Image processing and analysis was performed in 3D Slicer. The CT and MRI defined surface contours were saved as standard tessellation language (STL) files and registered with smartphone-generated 3D surface models. The accuracy of the smartphone generated models relative to reference models was evaluated using Hausdorff distances. To evaluate the ability of photogrammetry to detect surface changes, smartphone surface models were also measured after placing a 1cm thick 3D printed bolus on the phantom and volunteers.

Results: Overall, good agreement was obtained between surface models generated from smartphone photogrammetry and ground truth CT or MRI scans. As shown in Table 1, mean and standard deviation Hausdorff distances were 1.25 ± 0.70mm, 1.04 ± 0.74mm, and 0.64 ± 0.66 mm for the phantom and two human studies, respectively. Smartphone photogrammetry was able to detect the changes in surface anatomy, induced by the 3D printed bolus. The maximum Hausdorff distance in the surface region where 3D printed bolus was placed was 10.9 mm.

Conclusion: Accurate 3D surface model generation using smartphone photogrammetry is feasible. The portable and affordable surface scanning with smartphone presents a promising solution for early detection of patient surface changes using home-based monitoring. Early detection can prompt the need for plan adaptation outside the clinic, preventing further treatment delays.

Abstract 3465 – Table 1

Hausdorff distance (mm)	Head phantom	Human subject 1		Human subject 2
Maximum	4.92		4.30	5.30
Mean	1.25		1.04	0.64
Standard deviation	0.70		0.74	0.66