DR. M. REKA,DR. M. RAJASEKARAN,MALINI. M,DR. P. SATHIYAMURTHI
DOI: https://doi.org/An all-in-one framework of integrated image processing for precise lung cancer detection in CT images, comprising sophisticated pre-processing, segmentation and classification and extensive performance evaluation, is proposed in this work. Pre-processing with Anisotropic diffusion filtering at first is performed, which improves or enhances the quality of the image quality by getting rid of noise while maintaining the necessary edges and fine physiological structures of the image. Then, segmentation in Morpho-Geometric Region Segmentation (MGRS) is applied to correctly extract candidate tumour regions, which can further separate the nodules from the surrounding tissues by characteristics such as shape, size, and compactness. For classification, we use a fusion of Stacked Neural Network (SNN) and Optimized Deep Neural Network (ODNN) to achieve better diagnostic accuracy by studying complex features and model parameters in better tuning. The proposed method is compared with the other two methods by accuracy 96.84%,Recall 97.10%, specificity 96.45 %, precision 96.92% and F1-score 96.99% to confirm the performance of reliable identification between malignant and benign nodules. Experimental results indicate that the proposed integrated method is effective in achieving higher classification accuracy and strong generalizability, for early diagnosis of lung cancer and improving clinical decisions in lung cancer detection.
