Abstract:

Lung carcinoma grading is a critical task in the accurate diagnosis and treatment planning of lung cancer. In this study, we present “LungCarcinoGrade-EffNetSVM”, a novel approach that combines the powerful feature extraction capabilities of EfficientNetB0 with the classification prowess of Support Vector Machine (SVM) for lung carcinoma grading. The dataset utilized for this study was sourced from the Kaggle repository and includes images representing three types of lung carcinoma—Adenocarcinoma (ACA), Large Cell Carcinoma (LCC), Squamous Cell Carcinoma (SCC)— along with normal cell samples. Our proposed method achieved an Acc. of 86.88%, Sens. of 86.88%, and Spec. of 95.63%. The Prec. and F1 score were 87.06% and 86.64%, respectively, with a false positive rate (FPR) of 4.37%. The model also demonstrated robust performance with a Matthews correlation coefficient (MCC) of 0.8257 and a Kappa statistic of 0.65. The computational time for grading was recorded at 9.3082 seconds. These results indicate that the integration of EfficientNetB0 and SVM provides a reliable and efficient method for lung carcinoma grading, potentially aiding in more accurate and timely diagnosis of lung cancer.