This model is a combination of a detection, volumetric segmentation and classification of Lung Nodules for Chest CT studies. It aims at classifying lung nodules into malignant or benign based on learned image features. The malignancy score for each detected nodule allows radiologists to evaluate the likelihood of malignancy of a detected nodule and increase confidence in decision making.
Runs on screening Low dose Chest CT DICOM or regular dose Chest CT DICOM studies.
Not trained CAP CT.
Information on training data
Trained on NLST data. Nodules were annotated by experienced radiologists using biopsy confirmed malignant and radiology adjudicated benign as ground truth.
Model performance metrics
Sensitivity = 0.89
Specificity = 0.73
AUC = 0.93
Model available on Arterys.
Arterys is the World's First Online Medical Imaging Platform 100% Cloud native, powered by AI and FDA cleared.