DETECTION OF BLOOD CANCER

The Deep Learning classifier outperforms existing Red Blood Cell cancer CT image classifiers with a good accuracy. The findings of the experiment suggest that the current computer-aided diagnostic method offers considerable promise and security for automated diagnosis of benign and malignant cancers. This classifier dramatically reduces processing complexity. Additionally, the suggested approach might be expanded to identify many types of blood cell abnormalities. The results of the segmentation approach indicate that the online region-based segmentation method outperforms other techniques in terms of segmentation quality. After the initial extraction phase, map segmentation is applied in combination with online region-based segmentation for refined tumour area segmentation. The nodules identified through this segmentation are then used to extract texture features, which include energy, contrast, homogeneity, and correlation. Additionally, histogram-based features such as entropy, kurtosis, mean, skewness, and standard deviation are also extracted. These features are then compared against images of benign and malignant tumours. The DL classifier significantly outperforms other methods, achieving an accuracy of 97.1%. This marks an improvement of 5.4% over BRT, 4.4% over ANN, 2.4% over SVM, and 6.1% over KNN. These findings highlight the advantages of the proposed DL based approach for tumour classification.