A Master’s thesis entitled “Bone Cancer Detection Using Deep Learning and Meta-Heuristic Algorithms” was discussed at the College of Computer Science and Information Technology, Al-Qadisiyah University.

The thesis was presented by student Yaqeen Ali Mohsen, under the supervision of Professor Dr. Osama Majeed Hilal.

The thesis aimed to develop an intelligent model for detecting bone cancer based on X-ray images. This was achieved by employing deep learning techniques with a meta-exploratory optimization algorithm, thereby enhancing the model’s ability to differentiate between cases of bone cancer and normal cases.

The thesis utilized the BTXRD-2024 dataset of X-ray images of bone tumors. The data was prepared and pre-processed through a series of steps, including image cropping, resizing, and formatting to meet the requirements of the deep learning model. The study employed the DenseNet121 model, pre-trained using the ImageNet dataset, leveraging its ability to extract distinctive features from X-ray images. Additional classification layers were then constructed to achieve the final image classification.

The study also utilized the Walrus Optimization Algorithm (WaOA) to optimize several of the model’s hyperparameters, including the learning rate, number of neurons, and dropout rate. This optimized the model’s settings and improved its detection and classification performance. The model was trained in two phases. The first phase involved extracting features while maintaining the model’s basic layers. The second phase involved fine-tuning several layers to enhance the model’s ability to learn from the specific features of bone X-ray images.

The performance of the proposed model was evaluated using a set of established metrics, including accuracy, precision, recall, F1-score, and AUC. Confusion matrix analysis was also performed to measure the model’s ability to differentiate between categories.

The model achieved an accuracy of 99.18%.

The thesis concluded that artificial intelligence and deep learning techniques can be employed in analyzing X-ray images to aid in the detection of bone cancer. It also utilized the WaOA algorithm to optimize the model’s hyperparameters and achieve settings that contribute to enhanced classification accuracy.