Multi-Task Learning Approach for Detecting Multiple Tumors Using CT and MRI Images

This paper presents a multi-task learning approach for detection of brain and lung tumors from CT and MRI images. The Inception v3 model, which is a CNN that has an efficient feature extraction process with low computational complexity, was used to classify and locate tumors. The model utilized its multi-scale feature extraction ability to achieve high accuracy and robustness in tumor region differentiation across different datasets. Preprocessing steps such as normalization and data augmentation were found to enhance model generalization, while segmentation techniques improved the accuracy of localizing tumors. Experimental results show that the model was effective, with over 85% accuracy in tumor classification tasks. The proposed system aims to streamline diagnostic workflows, assist radiologists, and enhance healthcare outcomes by providing reliable, automated tumor detection. Future developments include expanding the system to detect additional tumor types and integrating real-time processing capabilities for clinical deployment..