ESTRO 2023 - Abstract Book

S599

Monday 15 May 2023

ESTRO 2023

The primary objective of this study was to pilot an in-house VLR. Prior to validating as a departmental training method for prostate SABR IGRT a comparison of inter-observer image registration variability between trainees either educated remotely using a preliminary VLR version or via established F2F training was conducted. Materials and Methods 8 RTTs participated and were divided into two cohorts. Cohort A: 4 participants underwent conventional F2F training delivered by an experienced trainer Cohort B: 4 participants underwent remote learning using a VLR training package Observers completed offline image registration for 12 prostate SABR CBCT images in XVI v5.0.4 (Elekta AB, Sweden) before training. After completion of training observers repeated the same 12 prostate SABR images (<21 days after 'before' registrations to minimise the effects of repetition). Resulting translations, time taken for registration and confidence scores using Likert scale of 1 (not confident) to 5 (extremely confident) were recorded and compared between cohorts using descriptive statistics. Inter-observer registration variability ( σ registration) was calculated for each plane (R/L, S/I, A/P) with respect to a gold standard registration (a consensus match agreed between three experienced IGRT specialist RTTs) according to Equation 1 (Image 1). Vectors were calculated for each registration and combined to determine the interobserver variability ( σ magnitude) according to Equation 2 (Image 1). Larger σ magnitude values indicated greater interobserver variability.

Results Results in Table 1 indicate that application of both F2F and VLR learning results in improved group inter-observer agreement. The improvement in mean agreement was greater for F2F (36%) that VLR (12%). F2F and VLR had similar improvements in standard deviation (36% and 38% respectively).

Pre training F2F registration confidence levels were 2.8, and for VLR 2.4. Post training F2F improved to 3.2 (14.3%) whilst VLR demonstrated lesser improvement in confidence (2.6, 8.3%). Combined group mean registration times for F2F and VLR were 4:35 mins and 4:13 mins respectively. Post training F2F recorded times were 4:17 mins (7%), whereas VTL improved by 31.8% (3:12 mins). Conclusion Whilst preliminary data indicates that both training methods improve inter-observer agreement, confidence and time to register it additionally suggests established F2F training should remain the preferred method within the department for the time being. Using participant feedback iterative work is underway to revise the pilot VLR package. Once complete it is anticipated that a larger substantive study will compare a modified version with F2F training in order to validate this novel training method. PD-0733 RADAR Project: pediatric patients AI-profiling, focus on clinical and psychological outcomes E. Marconi 1,2 , E. Meldolesi 1 , F. Beghella Bartoli 1 , F.P. Kraja 3 , S. Mariani 4 , G. Panza 1 , A. Mastronuzzi 5 , A. Ruggiero 6 , M.A. Gambacorta 1 , V. Valentini 1 , M. Balducci 1 , D.P.R. Chieffo 2 , S. Chiesa 1 1 Fondazione Policlinico Universitario Agostino Gemelli IRCCS, UOC di Radioterapia Oncologica, Dipartimento Diagnostica per Immagini, Radioterapia Oncologica ed Ematologia, , Rome, Italy; 2 Fondazione Policlinico Universitario Agostino Gemelli IRCCS, UOS di Psicologia Clinica, Rome, Italy; 3 University Hospital Center Mother Teresa, Genitourinary Oncology Department , Tirana, Albania; 4 Università̀ Cattolica del Sacro Cuore, UOC di Radioterapia Oncologica, Dipartimento Diagnostica per Immagini, Radioterapia Oncologica ed Ematologia, , Rome, Italy; 5 Ospedale Pediatrico