MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. The Feasibility of MVCT-Based Radiomics for Delta-Radiomics in Head and Neck Cancer

The Feasibility of MVCT-Based Radiomics for Delta-Radiomics in Head and Neck Cancer

K Abe1,2*, N Kadoya2 , S Tanaka2 , Y Nakajima1,2 , S Hashimoto1 , T Kajikawa2 , K Karasawa1 , K Jingu2 , (1) Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Tokyo, Japan,(2) Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendal, Japan

Presentations

(Sunday, 7/14/2019) 3:00 PM - 3:30 PM

Room: Exhibit Hall | Forum 8

Purpose: Changes in radiomic feature during treatment (delta-radiomics) can be potentially used as an imaging biomarker of treatment response. Megavoltage computed tomography (MVCT) images acquired using TomoTherapy (Hi-Art, Accuray) may be useful for delta-radiomics because they are typically obtained daily to verify the positioning of patients with head and neck (HN) cancer. This study aimed to evaluate the prognostic power of MVCT-based radiomics in patients with HN cancer.

Methods: In 37 patients with pharyngeal cancer (stages I–IV) treated with radiotherapy alone or chemotherapy using TomoTherapy, MVCT images obtained from the first fraction were analyzed. A total of 1,026 radiomic features (including the intensity histogram, intensity direct, gray level co-occurrence matrix [GLCM] 25, GLCM3, shape, gray level run length [GLRL] 25, GLRL3, neighbor intensity difference) were extracted from GTV of all scans. To clarify the prognostic power, the relationship between radiomic features and the 3-year overall survival (OS) was evaluated using the multivariate analysis with random forest machine learning. First, all 37 patients were divided into training (27 patients) and validation (10 patients) cohort. The prediction accuracy and area under the ROC curve (AUC) were obtained to evaluate the performance of our machine-learning-based classification. Finally, the accuracy and AUC in the validation cohort were evaluated.

Results: The feature selection with boruta showed that 25 features (GLCM3: 14; GLCM25: 11) had importance significantly higher than other features for predicting the 3-year OS in patients with HN. The accuracy values in training and validation cohorts were 0.96 and 0.80, respectively, and their AUC values were 0.95 and 0.83, respectively.

Conclusion: Prediction performance using MVCT-based radiomic features had high accuracy and AUC values. Our results suggested that radiomic features extracted from MVCT images have a prognostic power and can potentially be used as an imaging biomarker for adaptive radiation therapy in delta-radiomics.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by Foundation for Promotion of Cancer Research in Japan

Keywords

Megavoltage Imaging, Image Analysis, Radiation Therapy

Taxonomy

IM- CT: Machine learning, computer vision

Contact Email

Additional Links