Radiogenomic and Deep Learning Network Approaches to Predict KRAS Mutation from Radiotherapy Plan CT
Bum-Sup Jang1, Changhoon Song1, Sung-Bum Kang2
1Department of Radiation Oncology, Seoul National University Bundang Hospital, Seongnam, Republic of Korea.
Anticancer Research
|July 20, 2021
Summary
Radiogenomic scoring effectively predicts KRAS mutation status in rectal cancer patients using CT scans. This method proved more accurate than deep learning for identifying KRAS mutations.
Area of Science:
- Oncology
- Radiology
- Medical Imaging
Background:
- KRAS mutations are significant in locally advanced rectal cancer (LARC).
- Accurate prediction of KRAS mutation status is crucial for treatment planning.
- Radiotherapy planning computed tomography (CT) images offer a potential data source.
Purpose of the Study:
- To evaluate radiogenomic and deep learning models for predicting KRAS mutation status.
- To utilize radiotherapy planning CT images for non-invasive prediction.
- To compare the efficacy of radiogenomic versus deep learning approaches.
Main Methods:
- A cohort of 110 LARC patients was analyzed post-surgery.
- KRAS mutation status was determined, with 30 (27.3%) patients positive.
- Radiogenomic analysis involved extracting 378 texture features from the boost clinical target volume (CTV).
- A deep learning model was developed using 3D input from the CTV.
Main Results:
- The radiogenomic score model achieved an Area Under the Curve (AUC) of 0.73 for predicting KRAS mutation.
- The deep learning model showed a lower predictive performance with an AUC of 0.63.
- Radiogenomic analysis demonstrated superior predictive ability compared to deep learning.
Conclusions:
- The radiogenomic score model is a more feasible and effective approach for predicting KRAS mutation status in LARC.
- Radiotherapy planning CT imaging holds potential for non-invasive biomarker prediction.
- Further research may refine radiogenomic models for personalized cancer therapy.
