Authors - A Aruna kumari, Tamminana Visweswari Abstract - Lung cancer is a fatal illness that causes several deaths worldwide and detection of lung cancer remains a challenge for medical professionals. Detection of cancer in early stages is difficult as the size of the tumor is very small making it difficult for medical professionals to detect. Cancer detected in the early stages can be treated with proper techniques which can save the lives of the patients. Due to excessive information in the CT scans, MRIs, X-rays, and PET scans the manual detection of lung tumor becomes extremely difficult. The methodology helps in detecting the presence of cancerous tissues in the lungs and predicting which stage of lung cancer is present. The methodology mainly includes image preprocessing, training the model, extracting features using deep learning algorithms and classifying the stage of cancer present as Normal, Benign, Malignant Stage 1, Malignant Stage 2, and Malignant Stage 3. Proper image processing techniques like image augmentation, image normalization and image resizing are applied on the IQ-OTH/NCCD dataset for extracting the necessary features which will be used while training the model. A hybrid model is created by combining two deep learning models, the Xception and MobileNetV2 architectures which can accurately distinguish between the different lung cancer stages and predict the stage of cancer. The performance metrices which include accuracy, precision, recall, f1-score and confusion matrix were also calculated to determine the accuracy of the proposed hybrid model. The proposed model helps in accurate and reliable diagnosis of lung cancer at early stages.
Open Zoom