A model based on radiomic features extracted from CT scans can help predict which ground glass nodule (GGN) cases require surgery and may reduce overtreatment of the condition, according to researchers at the Affiliated Suzhou Hospital of Nanjing Medical University in Suzhou, China.
Ground glass nodule detection has risen alongside CT’s increased use in lung cancer screening, but pulmonary GGNs can represent various abnormalities, wrote Chen-Lu Liu, MD, with Nanjing Medical University’s Department of Radiology, and colleagues. Therefore, an improved predictive model is sorely needed.
“As the prevalence of lung adenocarcinoma and the detection rates of GGNs in nonsmokers increase, we need not only a more accurate prediction model for GGNs but also a predictive model to judge whether the GGNs require surgical resection,” the authors wrote in the study published Nov. 19 in the Journal of the American College of Radiology.
The study included CT images of 239 patients with GGNs. A total of 160 cases were used to train the predictive model and 79 to validate and verify it.
Overall, a predictive model which combined clinical information and the radiomics nomogram achieved the best predictive ability and calibration in both the training set and validation set, with an area under the curve (AUC) of 0.831 and 0.816, respectively. A model created with only clinical information achieved an AUC of 0.711.
The researchers did note a few limitations of their study, including potential bias involved in the manual contouring of regions of interest. Overall, the authors believe their model could help individualize GGN care.
“This predictive model can intuitively help the clinicians to identify the GGNs that require surgical resection and avoid the occurrence of overtreatment and the waste of medical resources, which is in line with the trend of precision medicine,” Liu et al. concluded.