Thyroid carcinomas are often diagnosed by histopathology, which is widely regarded as the most reliable method. However, alternative imaging modalities may also provide meaningful information about thyroid tumors. Multiphoton Microscopy (MPM) images may be one of them. MPM images include Second Harmonic Generation (SHG) and Two-Photon Excitation Fluorescence (TPEF) images. Nevertheless, the field of automated analysis of MPM images for the diagnosis of cancer is in its infancy. We propose a strategy for the differential diagnosis of thyroid tumors through information fusion from different types of MPM images. We introduce a novel fusion autoencoder (FAE) for this task. The fused information from the FAE is subsequently used by a classifier module for the differential diagnosis of thyroid tumors. Our method is one of the first approaches to look into the possibility of using MPM images for the diagnosis of thyroid tumors. Extensive experiments demonstrate the superiority of the proposed method compared to a number of state-of-the-art classification techniques. The code for the paper can be found at https://github.com/HarshithK13/ICPR2024-Thyroid-Diagnosis.git .