Jung Youn Kim1, Min Jae Yoon2, Ji Eun Park3, Eun Jung Choi4, Jongho Lee4, Ho Sung Kim5. 1. Department of Radiology, Sungkyunkwan University School of Medicine, Kangbuk Samsung Hospital, 29 Saemunan-ro, Jongro-gu, Seoul, 03181, South Korea. 2. Department of Radiology, University of Konyang College of Medicine, Konyang University Hospital, Daejeon, 35365, South Korea. 3. Department of Radiology and Research Institute of Radiology, University of Ulsan College of Medicine, Asan Medical Center, 88 Olympic-ro 43-gil, Songpa-Gu, Seoul, 05505, South Korea. jieunp@gmail.com. 4. Laboratory for Imaging Science and Technology, Department of Electrical and Computer Engineering, Seoul National University, Seoul, South Korea. 5. Department of Radiology and Research Institute of Radiology, University of Ulsan College of Medicine, Asan Medical Center, 88 Olympic-ro 43-gil, Songpa-Gu, Seoul, 05505, South Korea.
Abstract
PURPOSE: The peritumoral non-enhancing region (NER) is frequently not removed during the surgical resection of glioblastoma, with most recurrences occurring within the original treatment field. This study determined whether radiomics analysis of the NER can predict local recurrence and overall survival in patients with glioblastoma. METHODS: Preoperative magnetic resonance imaging (MRI) scans from 83 consecutive patients with glioblastoma were retrospectively reviewed and grouped into training (n = 59) and test sets (n = 24). A total of 6472 radiomic features were extracted from contrast-enhanced T1-weighted and fluid-attenuated inversion recovery images and from fractional anisotropy (FA) and normalized cerebral blood volume (CBV) maps. A diagnostic model to predict 6-month progression was tested using the area under the receiver operating characteristics curve (AUC) and compared with the single parameters of FA and CBV. A survival model was tested using Harrell's C-index and compared with clinical models that included age, sex, Karnofsky performance score, and extent of surgical resection. RESULTS: Four FA features and six CBV features were selected for the diagnostic model; no features were extracted from conventional MRI. Combined FA and CBV radiomics showed better predictive value for local progression (AUC, 0.79; 95% CI, 0.67-0.90) than single imaging radiomics (AUC, 0.70-0.76) or single imaging parameters (AUC, 0.51-0.54). The combined model (C-index, 0.87) improved prognostication when added to clinical models (C-index, 0.72). CONCLUSION: Radiomics features using FA and CBV in the NER have the potential to improve prediction of local progression and overall survival in patients with glioblastoma.
PURPOSE: The peritumoral non-enhancing region (NER) is frequently not removed during the surgical resection of glioblastoma, with most recurrences occurring within the original treatment field. This study determined whether radiomics analysis of the NER can predict local recurrence and overall survival in patients with glioblastoma. METHODS: Preoperative magnetic resonance imaging (MRI) scans from 83 consecutive patients with glioblastoma were retrospectively reviewed and grouped into training (n = 59) and test sets (n = 24). A total of 6472 radiomic features were extracted from contrast-enhanced T1-weighted and fluid-attenuated inversion recovery images and from fractional anisotropy (FA) and normalized cerebral blood volume (CBV) maps. A diagnostic model to predict 6-month progression was tested using the area under the receiver operating characteristics curve (AUC) and compared with the single parameters of FA and CBV. A survival model was tested using Harrell's C-index and compared with clinical models that included age, sex, Karnofsky performance score, and extent of surgical resection. RESULTS: Four FA features and six CBV features were selected for the diagnostic model; no features were extracted from conventional MRI. Combined FA and CBV radiomics showed better predictive value for local progression (AUC, 0.79; 95% CI, 0.67-0.90) than single imaging radiomics (AUC, 0.70-0.76) or single imaging parameters (AUC, 0.51-0.54). The combined model (C-index, 0.87) improved prognostication when added to clinical models (C-index, 0.72). CONCLUSION: Radiomics features using FA and CBV in the NER have the potential to improve prediction of local progression and overall survival in patients with glioblastoma.
Authors: Andrew T Parsa; Scott Wachhorst; Kathleen R Lamborn; Michael D Prados; Michael W McDermott; Mitchel S Berger; Susan M Chang Journal: J Neurosurg Date: 2005-04 Impact factor: 5.115
Authors: Russell T Shinohara; Elizabeth M Sweeney; Jeff Goldsmith; Navid Shiee; Farrah J Mateen; Peter A Calabresi; Samson Jarso; Dzung L Pham; Daniel S Reich; Ciprian M Crainiceanu Journal: Neuroimage Clin Date: 2014-08-15 Impact factor: 4.881
Authors: Yan Li; Yiqi Ma; Zijun Wu; Ruoxi Xie; Fanxin Zeng; Huawei Cai; Su Lui; Bin Song; Lei Chen; Min Wu Journal: Front Immunol Date: 2021-11-25 Impact factor: 7.561
Authors: Thomas C Booth; Mariusz Grzeda; Alysha Chelliah; Andrei Roman; Ayisha Al Busaidi; Carmen Dragos; Haris Shuaib; Aysha Luis; Ayesha Mirchandani; Burcu Alparslan; Nina Mansoor; Jose Lavrador; Francesco Vergani; Keyoumars Ashkan; Marc Modat; Sebastien Ourselin Journal: Front Oncol Date: 2022-01-31 Impact factor: 6.244
Authors: Jiun-Lin Yan; Chao Li; Anouk van der Hoorn; Natalie R Boonzaier; Tomasz Matys; Stephen J Price Journal: Sci Rep Date: 2020-06-16 Impact factor: 4.379