Leukemia Cancer Detection Using Image Analytics : (Comparative Study)

Leukemia is a cancer of white blood cells (WBC). It can be fatal if not detected early. Microscopic images are studied by hematologists for detecting cancer. This manual detection becomes very tedious and time-consuming process. Leukemia if detected in earlier stages, can be cured. But traditional process causes late detection of cancerous cells. Hence in order to minimize the death caused due to late detection, an automated system can be used. This paper proposes an automated system which uses image analytics. Based on image analytics and classification algorithms performed on cell image samples of patients, the proposed system will give correct output. The dataset for experimentation is obtained from TCIA (The Cancer Imaging Archive) repository. The dataset is already pre-processed. An open source tool, "Orange-Data Mining" is used for predictions. In this comparative study, it was found that K-means clustering performs well for segmentation phase and also Neural Networks gives better results for classification phase. We have achieved AUC (area under curve) 0.865, Calculation accuracy (0.838), precision (0.835) and F1(0.836) for neural networks.