Spots Global Cancer Trial Database for Influence of PET/CT Radiomic Features on the Outcome of Lung Cancer Patients

The following info and data is provided "as is" to help patients around the globe.
We do not endorse or review these studies in any way.

Trial Identification

Brief Title: Influence of PET/CT Radiomic Features on the Outcome of Lung Cancer Patients

Official Title: Influence of PET/CT Radiomic Features on the Outcome of Lung Cancer Patients

Study ID: NCT03648151

Conditions

Lung Cancer

Image, Body

Interventions

Study Description

Brief Summary: Radiomics is an attractive field in objectively quantifying image features, and may overcome the subjectivity of visually interpreting computed tomography (CT), or positron emission tomography (PET). It is reported that the features related to treatment response, outcomes, tumor staging, tissue identification, and cancer genetics. Therefore, the investigators try to explore the key features for the outcome of lung cancer patients.

Detailed Description: Radiomic Features: PET/CT images, including other kinds of CT serials, were transported into a personal computer. Using the open source software of 3D-Slicer, volumes of interest (VOIs) for primary tumor, or even lymph nodes, was semi-automatically or manually segmented. And then, radiomic features were extracted. PET Parameters: Using combined CT VOIs, corresponding PET standard uptake value (SUV, no unit) were measured. For a foci (either tumor, or lymph node), mean, sum and maximum SUV were documented, and were used for training and validating models alongside radiomic features. Feature Selection: Data were analyzed by deep learning or random forests method, and top 20 variables were scored by their contribution to the regression (variable importance, VIMP). The generalized features were identified as the same ones between two kinds of image serials (for example, ordinary and thin-section CT, or PET and CT). Additionally, when three or more features met the criterion, a lower value of Akaike information criterion (AIC) which measures the relative quality of statistical models was used to find appropriate features with lower overfitting possibility. Model Validation: The developed model was validated internally and externally. The internal indices for independent continuous variable were accuracy (bias and absolute bias) and precision (correlation coefficient and R square), and that for independent classified or survival variable was c-index. The patients enrolled from another medical center were used for external validation.