ESTRO 2020 Abstract book

S974 ESTRO 2020

Figure 1 shows an example of MAA, 90Y with dose‐maps and QF for one patient. Lesions had a good, acceptable and poor concordance in 56%, 25% and 19% of the cases respectively. Similarly only 23% of the TL resulted in poor agreement, next to 38% acceptable and 39% good ones. In contrast, inconsistent results were observed for NTL where 15%/27%/58% appeared as good/acceptable and poor conformity. Table 1 summarises the univariates association of the dichotomised explanatory variables with QF values for the different structures. Interestingly, the delay between simulation and treatment (>9days vs. <9days) was a significant predictor across all VOIs. The median QF was 0.53 vs. 0.39 ( p =0.04) for individual lesions, 0.62 vs. 0.42 ( p =0.02) for TL and 1.20 vs. 0.60 ( p =0.002) for NTL. The volume of NTL was also a predictor of D 90Y and D MAA differences with a median QF of 1.22 for large vs. 0.61 for small volumes (p=0.02).

At the primary endpoint of 4 months, 6/25 patients from which complete data was available showed clinically relevant deterioration. This corresponds with the 28% decline in all patients participating in the study. Median change in the HVLT‐R was ‐2 and – 3.5 points at 4 and 8 months, respectively. There was no consistent correlation between tumour stage, pack years, smoking status, age, sex and verbal memory at 4 months. With a median split, relatively high cytokine IL1a levels at baseline (38.5% vs 8,3%) and pro‐inflammatory biomarkers IL6 and IL8 (36,4% vs 14,3% and 38,5% vs 8,3% respectively) coincided with clinically relevant cognitive decline at 4 months. There was a trend towards significance between the Δ NfL (T2‐ T1) and the HVLT‐R at 4 months with 1/12 (8.3%) patients with a decrease in NfL demonstrating neurocognitive decline, compared with 5/11 (45%) patients with an increase in NfL. The mean serum NfL was 120.48, 134,70 and 214.72 pg/ml, respectively at T1, T2 and T3. This was markedly higher when compared to published data on patients with neurological disease such as MS or healthy controls Conclusion NfL has previously been established as a promising blood biomarker for neurodegeneration. In this study both the Δ NfL and baseline inflammatory biomarkers were identified as potential early biomarkers correlating with neurocognitive decline in patients treated with PCI for SCLC. This study supports further investigation into these potential biomarkers to detect early neuronal damage after cranial irradiation PO-1677 Quality Volume Histogram concept for personalized radioembolization based on MAA SPECT- CT simulation M. Burghelea 1 , H. Levillain 2 , B. Vanderlinden 1 , T. Guiot 1 , A. Gulyban 1 , N. Reynaert 1 , P. Flamen 1 1 Jules Bordet Institute, Medical Physics Department, Bruxelles, Belgium ; 2 Jules Bordet Institute, Nuclear Medicine Department, Brussels, Belgium Purpose or Objective It has been proven that the absorbed dose measured in 99m Tc‐MAA‐SPECT/CT (MAA) or 90Y PET/CT (90Y) can predict the radioembolization treatment response, however this precision remains better in 90Y. Therefore it is necessary to study the predictive power of pre‐ treatment MAA for 90Y treatment dosimetry. For this purpose, we propose the quality volume histogram (QVH) as voxel‐to‐voxel evaluation method to compare simulated and treatment absorbed doses distributions. Material and Methods Thirty liver‐only mCRC patients were included in this retrospective investigation, where all patients underwent treatment simulation based on MAA followed by 90Y. Individual lesions were delineated using a standardized threshold. PlanetOnco 3.0 workstation (Dosisoft, France) was used for initial co‐registration and 3D personalized voxel‐based internal dosimetry computation for simulation (D MAA ) and treatment (D 90Y ). The 90Y was considered as reference, where Liver, tumoral‐liver (TL) and non‐ tumoral‐liver (NTL) were delineated as volumes of interest (VOI). The QVH tool was implemented for dose maps comparison, using MICE Toolkit 1.20 (NONPI Medical AB, Sweden). QVHs were computed for TL, NTL and all lesions. The Quality factor (QF) (the spread of the QVH curve around 1, ideal = 0) were also determined per VOI. QF between 0 and 0.5 were considered good, between 0.5 and 1 acceptable, while QF > 1 were considered as poor correspondence. Univariate analysis based on gender, former surgery, previous chemotherapy, delay in treatment, administered activity, TL and NTL volumes were performed to identify possible predictive factors for the QF, where p<0.05 considered statistically significant. Results

Conclusion Quality Volume Histogram concept seems to be adequate to assess the predictive value of MAA simulation dosimetry for personalized radioembolization.