CANCER DIAGNOSIS DEEP LEARNING-BASED FUSION OF CT, MRI, AND PET IMAGING FOR IMPROVED DIAGNOSIS AND STAGING OF COMPLEX CANCERS

Abstract

Cancer can be quite complex and have a variety of anatomical, metabolic, and functional features that may not be sufficiently represented by a single imaging modality. In this paper, a deep learning model for multimodal fusion computed tomography (CT), magnetic resonance imaging (MRI) and positron emission tomography (PET) is proposed to enhance the accuracy of cancer diagnosis and staging. CT allows for detailed structural and anatomical information, MRI gives better soft tissue contrast and tumor boundary characterization and PET gives information on patterns of functional and metabolic activity, associated with malignancy progression. The proposed method not only extracts the features of each modality but also applies deep fusion layers to learn complementary representation across different modalities. The model combines spatial, textual, and metabolic characteristics to assist in the precise localization, classification, and prediction of cancer malignancy in complex tumors. It can help guide clinicians in minimizing diagnostic uncertainty, enhancing staging uniformity, and guiding individualized treatment planning. Overall, the study highlights the potential of deep learning–driven multimodal imaging fusion as a clinically useful decision-support tool for precision oncology.