Morphometric Analysis of Nuclei and Cytoplasm in Lung Cancer Brightfield Images for Drug Target Classification

Identifying drug target genes is essential for improving the efficacy of chemotherapeutic drugs and for personalized medicine. The drug treatments targeting cancer-causing genes result in structural and functional changes in nuclei and cytoplasm, which are manifested as morphological variations. Brightfield imaging is the most widely available technique for visualizing the morphological alterations in these organelles. In this work, brightfield images of lung cancer cells treated with drugs targeting Aurora Kinase (AURK) and Rapidly Accelerated Fibrosarcoma (RAF) genes are obtained from a public database. The morphometric features are extracted from the segmented nuclei and cytoplasm. After statistical analysis, the significant features are fed to the random forest, XGBoost, and multilayer perceptron (MLP) for the classification of drug targeting AURK and RAF genes. The obtained results show that combined morphometric features from nuclei and cytoplasm with the XGBoost algorithm achieved the maximum classification accuracy of 78.08%. It indicates the potential of morphometric features to capture the morphological changes in the cell organelles due to the drugs targeting specific cancer-causing genes.