medical-informatics-2016_posters-accepted-abstracts

2 / 14

Page 46
Notes:
conferenceseries.com
Volume 7, Issue 5 (Suppl) J Health Med Inform
ISSN: 2157-7420 JHMI, an open access journal Medical Informatics 2016
October 6-7, 2016
October 6-7, 2016 | London, UK
4th International Conference on
Medical Informatics & Telehealth
RADIOMIC FEATURESANALYSIS IN PET IMAGES FOR HEADAND NECK CANCER
Kuei-Ting Choua, Chiou-Yu Roua, Geoffrey Zhangb, Shih-Neng Yang a and Tzung-Chi Huang a
aChina Medical University, Taiwan
bMoffitt Cancer Center, USA
Positron emission tomography (PET) image has been used routinely in oncology for tumor diagnosis, staging and assessment
of treatment response. However, the information extracted from image-based features for diagnosis is still under development
during the past decade. In recent years, radiomics texture analysis has been used in medical imaging to obtain quantitative data
through automated and reproducible analysis, reflecting the characteristics of the tumor, providing additional clinical diagnostic
information. In this study, we analyzed 80 head and neck cancer and extracted image features from four metabolic volumes (MTV2.5,
MTV3.0, MTV40% and MTV50%) of PET images. The features include shapes, intensity-based, grey level co-occorence, size zone,
length, neighborhood grey-tone difference etc.. ANOVA and Kruskal-Wallis test was used to assess the differences of image texture
features in different groups of patients. Receiver-operating characteristic analysis was used to find out the optimal cutoff point of
overall survival (OS) and primary, relapse free survival (PRFS) with different image features. The results showed that 16 image texture
features had significant differences in early tumor stage (T1, T2) and lately tumor stage (T3, T4). We found 5 and 2 image textural
features had ability to predict the tumor response and recurrence, respectively. The histogram entropy is the one predictor of OS and
PRFS of head and neck cancer patients. We found that image textural features provide predictive and prognostic information on
tumor staging, tumor response, recurrence, and can be a prognosticator for OS and PRFS in head and neck cancer in PET images.
Kuei-Ting Chou et al., J Health Med Informat 2016, 7:5 (Suppl)