Identifying LDCT imaging features associated with second primary lung cancer risk among lung cancer patients
Over the past decades, low-dose computed tomography (LDCT) has transformed the field of lung cancer screening and surveillance. Along with the popularity of LDCT screening, an increasing number of lung ground-glass opacity (GGO) lesions are detected. Although GGO, different from a solid nodule or carcinoma in situ, is a non-specific radiologic finding, it has the potential to evolve into a malignant lesion, i.e., SPLC. However, what characteristics of specific GGO among IPLC patients are associated with the one that eventually develop and grow as SPLC is unknown. Given that LDCT imaging features are highly predictive of malignancy of a GGO and subsolid nodule in the general population (Baldwin DR, et al. [Thorax. 2020]; Massion PP, et al. [AJRCM. 2020]), Mikhael et al. (JCO, 2023) developed and validated a LDCT image-based deep learning model (called Sybil) to predict IPLC risk in the general population. So, we would like to examine whether the performance of the LDCT imaging-based deep learning tool (Sybil) can identify lung cancer survivors who have a high risk of SPLC. As the next step, we aim to identify the LDCT imaging features of pulmonary nodules associated with SPLC risk among lung cancer survivors, and to evaluate the predictive potential of the imaging features using radiomics and deep learning methods.
(1) To evaluate the predictive accuracy of SYBIL lung cancer risk prediction model for SPLC among LC survivors using NLST data
(2) To identify the LDCT imaging features of pulmonary nodules (especially focusing on GGOs and subsolid nodules) associated with SPLC risk among lung cancer survivors
(3) To evaluate the predictive potential of the imaging features using radiomics and deep learning methods
Summer Han; Eunji Choi