Download This Paper
Preprints with The Lancet is a collaboration between The Lancet Group of journals and SSRN to facilitate the open sharing of preprints for early engagement, community comment, and collaboration. Preprints available here are not Lancet publications or necessarily under review with a Lancet journal. These preprints are early-stage research papers that have not been peer-reviewed. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. The findings should not be used for clinical or public health decision-making or presented without highlighting these facts. For more information, please see the FAQs.
Integration of Radiomics Models of Tumor and Deep Learning Based Volumetric Segmentation of Three-Dimensional Abdominal Muscles Improves the Predictive Performance on the Prognosis of Gastrointestinal Stromal Tumor
24 Pages Posted: 25 Apr 2024
More...Abstract
Background: GISTs are recognized as a type of tumor exhibiting varying degrees of malignant potential. Muscle status serves as a representative marker for nutritional health. Nonetheless, the broader implications of general muscle status on GIST patient prognosis remain to be fully elucidated.
Method: We conducted a retrospective study, enrolling 218 patients from The First Hospital of China Medical University as the development cohort and 36 patients from Liaoning Cancer Hospital & Institute for external validation. We constructed a novel abdominal 3D-CT volumetric muscle segmentation framework consisted of two main components: muscle segmentation and post-processing. Based on the results of ROI annotation, we developed radiomics models based on features extracted from tumor and muscles, respectively.
Results: After post-fusion of models, the integrated radiomics model consisted of tumor, muscle, delta-muscle performed the best on the external validation cohort with a mean C-index of 0.788, and demonstrated proficiency in stratifying patient prognosis, as evidenced by Kaplan-Meier analysis, with a p-value < 0.05 observed in both the development and test cohorts. We also noted that the prediction generated by radiomics models of muscles were related with hypoalbuminemia, low level of blood creatinine, high level of gamma glutamyl transferase which are significant markers of malnutrition and low lean mass and weakness.
Conclusion: In the present study, we developed a novel deep learning model for automatic segmentation of abdominal muscle of patients with GIST. We noted that the muscle status could serve as a significant synergetic factor together with tumor characteristics.
Funding: This study is supported by a grant from the Natural Science Foundation of Liaoning Province (No. 2023-MS-170)
Declaration of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interests.
Ethical Approval: This study was conducted in accordance with the Declaration of Helsinki and has received approval from the Institutional Ethics Review Board of The First Hospital of China Medical University and Liaoning Cancer Hospital & Institute. As this was a retrospective study, the requirement for written signed informed consent was waived.
Keywords: Gastrointestinal stromal tumor, nnU-Net, radiomics, muscle status, machine learning, survival mode
Suggested Citation: Suggested Citation