a) build and improve computed tomography (CT) image processing applications that use shape and geometry information to highlight lung nodules for further radiological assessments,
b) extract image-based nodule features such as size, type, location etc,
c) generate dense nodule-specific feature vectors (NFVs) that combine image-based nodule features with clinical factors such as age, sex, history of smoking and family history of lung cancer,
d) evaluate the performance of previously validated lung cancer risk tools to differentiate benign versus malignant nodules using components of the derived NFVs as needed and finally,
e) utilize these dense NFVs in conjunction with the known nodule–specific lung cancer or benign statuses to develop and validate a classifier and risk model that can assign a probability of “benign” versus “malignant” to each detected nodule with improved accuracy.

Extended Aim:
Statistical analysis of the data and correlation with known nodule-specific lung cancer and benign statuses, to evaluate the manufacturer specific performance and efficacy of using low dose CT as a screening modality.

Dann, Robert (GE Healthcare);
Alam, Shamsul (GE Healthcare);
Talla Souop, Alain (GE Healthcare);
Doan, Huy (GE Healthcare);
Jaeckle, John (GE Healthcare);
Sirohey, Saad (GE Healthcare);
Duliskovich, Tibor (GE Healthcare);
Richard Petrisko (GE Healthcare);
Keyur Desai (GE-GRC);
Shubao Liu (GE-GRC);
Xiaojie Huang (GE-GRC)