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 These preprints are early-stage research papers that have not been peer-reviewed.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.
Radiomic Features, the Bridge Between CT and FDG PET Images: Estimating the PET Parameters of Lymph Node from CT with Deep Learning Method
28 Pages Posted: 16 Jun 2019
More...Abstract
Background: Radiomic features might bridge CT to FDG PET parameters (SUVmax, SUVmean and SUVsum). Therefore, the estimation models were trained and validated for the lymph nodes of cancer patients.
Methods: The lymph nodes (LNs) of the lung cancer (LC) patients from SMU (n= 1239) were used as the training cohort, and both PET/CT and thin-section CT were collected. Those of the esophagus cancer (EC) from SMU (n=301), and the LC from AMU (n=204) and TCIA (n=47 and 82) were used as the validation cohorts. The radiomic features of LNs on CT serials (short axis diameter >3mm) were extracted. The feature selection and model training strategy were investigated, and the influential factors of proposed deep learning models were explored.
Findings: From each LN of the patients, 141 features (1683 variables) were extracted. In the training cohort, the top 20 features on the combined and thin- section CT serials were sorted by the deep learning method, and the identical 2, 2 and 3 features between the serials were trained against SUVmax, SUVmean and SUVsum by the logarithmically transferred model, respectively. In the validations, the logarithms of SUMsum model performed well in the SMU EC cohort (r=0.847, bias= 0.024, P=0.843) with particular coefficients for different scanners. The absolute biases of all the models were slightly affected by LN volume, injected dose, acquisition time, pathology classification and LN metastasis.
Interpretation: Several radiomic features could be used to estimate PET parameters of lymph nodes.
Trial Registration: ClinicalTrials.gov (NCT03648151).
Funding Statement: Collaborative Innovation Center for Molecular Imaging and Precise D&T Center of Shanxi Medical University (Grant No. MP201604).
Declaration of Interests: The authors declare that they have no conflict of interest.
Ethics Approval Statement: The protocol was approved by the ethics committees at the First Affiliated Hospital of Shanxi Medical University (SMU) and the First Affiliated Hospital of Anhui Medical University (AMU). For a retrospective study, the review boards approved to waive the requirement for informed consent.
Keywords: Radiomics; lung cancer; Standard uptake value; Positron emission tomography; Computed tomography
Suggested Citation: Suggested Citation