Oates, James T Chapman, DavidJamdade, Vaishnavi Avinash2023-04-052023-04-052022-01-0112639http://hdl.handle.net/11603/27343We present an AI-assisted approach for classification of malignancy of lung nodules in CT scans for explainable AI-assisted lung cancer screening. We evaluate this explainable classification to estimate lung nodule malignancy against the LIDC-IDRI dataset. The LIDC-IDRI dataset includes biomarkers from Radiologist's annotations thereby providing a training dataset for nodule malignancy suspicion and other findings. The algorithm employs a 3D Convolutional Neural Network (CNN) to predict both the malignancy suspicion level as well as the biomarker attributes. Some biomarkers such as malignancy and subtlety are ordinal in nature, but others such as internal structure and calcification are categorical. Our approach is uniquely able to predict a multitude of fields such as to not only estimate malignancy but many other correlated biomarker variables. We evaluate the malignancy classification algorithm in several ways including presentation of the accuracy of malignancy screening, as well as comparable metrics for biomarker fields.application:pdfText