Due to the close resemblance between overlapping and cancerous nuclei, the\nmisinterpretation of overlapping nuclei can affect the final decision of cancer cell detection. Thus,\nit is essential to detect overlapping nuclei and distinguish them from single ones for subsequent\nquantitative analyses. This paper presents a method for the automated detection and classification\nof overlapping nuclei from single nuclei appearing in cytology pleural effusion (CPE) images. The\nproposed system is comprised of three steps: nuclei candidate extraction, dominant feature extraction,\nand classification of single and overlapping nuclei. A maximum entropy thresholding method\ncomplemented by image enhancement and post-processing was employed for nuclei candidate\nextraction. For feature extraction, a new combination of 16 geometrical and 10 textural features was\nextracted from each nucleus region. A double-strategy random forest was performed as an ensemble\nfeature selector to select the most relevant features, and an ensemble classifier to differentiate between\noverlapping nuclei and single ones using selected features. The proposed method was evaluated on\n4000 nuclei from CPE images using various performance metrics. The results were 96.6% sensitivity,\n98.7% specificity, 92.7% precision, 94.6% F1 score, 98.4% accuracy, 97.6% G-mean, and 99% area under\ncurve. The computation time required to run the entire algorithm was just 5.17 s. The experiment\nresults demonstrate that the proposed algorithm yields a superior performance to previous studies\nand other classifiers. The proposed algorithm can serve as a new supportive tool in the automated\ndiagnosis of cancer cells from cytology images.
Loading....