ESTRO 37 Abstract book

S1175

ESTRO 37

2 Shandong Normal University, School of Physics and Electronics, Ji'nan, China

patient were calculated. The ratio of the MEDIC mean volume to the CT mean volume for each patient and the equivalent ratio of the SPACE to the CT was determined. Results The prostate delineations using the MEDIC sequence had the highest similarity measures for four out of five patients (see figure 1). Patient 5 appears to be anomalous for the MEDIC sequence, which may be due to significant motion artefact present on the MEDIC images for that patient. There was a wide range of variability in the definition of seminal vesicles, with the MEDIC and CT performing similarly (see figure 2). Both MR prostate volumes were consistently smaller than the CT. The SPACE-CT volume ratio was the smallest, 0.74 ± 0.04 (mean ± s.e.m.) compared to 0.82 ± 0.05 for the MEDIC- CT ratio. Similarly for the seminal vesicles, the SPACE-CT ratio was 0.79 ± 0.05 and the MEDIC-CT ratio 0.88 ± 0.06.The MEDIC sequence was preferred by all delineators and was the quickest to delineate, taking 10 ± 1 minutes (mean ± s.e.m.). The CT took 13 ± 1 minutes and the SPACE 14 ± 1 minutes.

Purpose or Objective To evaluate the impact of b-values on radiomic features of diffusion-weighted imaging (DWI) and to seek stable radiomic features which could be potentially used in quantitative clinical analysis of hepatic cirrhosis. Material and Methods The DWI data set of 35 patients who have been diagnosed with hepatic cirrhosis were randomly enrolled. DWI scanning was performed with ten b-values (0, 20, 50, 100, 200, 400, 800, 1000, 1200 and 1500 s/mm 2 , respectively) using a GE 3.0T Discovery scanner. Rigid registration was used to align ten b-value DWIs to T1WI image in MIM software. Using the anatomical information of T1WI as reference, we defined three VOIs (volume of interest, VOI) in parenchyma of liver segment II/III, V/VI and VII of b=0 s/mm 2 DWI images. Then, the VOIs in b=0 s/mm 2 were mapped to the other nine b-value DWI images. After that, 78 radiomic features were extracted by 3D Slicer Radiomics Extension from these VOIs, including 19 first order intensity histogram (IH) based features, 27 gray-level co-occurrence matrix (GLCM) based features, 16 gray-level run-length matrix (GLRLM) based features and 16 gray-level size zone matrix (GLSZM) based features. The percentage coefficient of variation (%COV) was used to determine the robust radiomic features against different b-values and the feature with %COV < 30 was considered as a stable one. The concordance correlation coefficient (CCC) was used to evaluate the reproducibility of different b-value DWI images and the image with CCC > 0.8 was considered to have higher reproducibility. Results 30 radiomic features were stable across ten b-value DWI images is presented in Table 1, including 4 IH-based features, 15 GLCM-based features, 7 GLRLM-based features and 4 GLSZM-based features. For different b- value DWI images, the higher reproducible radiomic features can be extracted from DWIs with nearby b- values. Especially, the most higher reproducible radiomic features were extracted from DWI image group b=0, 20, 50, 100, 200s/mm 2 and group b=800, 1000, 1200 s/mm 2 . Table 1 The stable radiomic features across ten b-value DWI images (%COV < 30)

IH

GLCM GLCM- IDMN

GLRLM GLSZM

GLCM- InverseVar

GLRLM- SRE GLRLM- R.Perce nt GLRLM- RLNUN GLRLM- R.Entro py GLRLM- RLNU GLRLM- LRE GLRLM- GLNU

GLSZM- SAE GLSZM- Z.Entro py GLSZM- GLNU

IH-Kurtosis

GLCM- IMC2 GLCM- Homo2 GLCM- IDMN

IH-Entropy

GLCM-IDN

IH- RobustAbsDe v IH- InterQuarRan ge

GLCM- SumEntro py

Conclusion The MEDIC sequence appears the most efficient for prostate delineation and is preferred by all oncologists in this study. It also appears to have the least variability in prostate delineations and is no worse than the SPACE sequence and CT in delineating the seminal vesicles. Volumes delineated on MR are consistently smaller than those delineated on CT, with the SPACE volumes being smaller than the MEDIC. The MEDIC sequence appears a more efficient option for routine prostate delineation. EP-2129 The impact of b-values on radiomic features of diffusion-weighted imaging in hepatic cirrhosis Q. Qiu 1,2 , J. Zhang 1,2 , J. Duan 1 , Y. Yin 1 1 Shandong Cancer Hospital Affiliated to Shandong University, Department of Radiation Oncology, Ji'nan, China

GLCM- Entropy

GLSZM- SZNUN

GLCM- Homo1

GLCM- SumAver

GLCM-ID GLCM-

AverInten

GLCM- DiffEntrop y

GLCM- DiffAver

GLCM- Dissimilari ty