Machine learning approach for distinguishing malignant and benign lung nodules utilizing standardized perinodular parenchymal features from CT

Radiomic features extracted from the parenchyma surrounding lung nodules contain valid signals with spatial relevance for the task of lung cancer risk classification. Through standardization of feature extraction regions from the parenchyma, ML tool validation performance of 100% sensitivity and 96% specificity was achieved.