We developed a deep learning (DL) model to enhance mesoscopic fluorescence lifetime imaging (FLI). By fusing low-resolution time-resolved images from a time-gated SPAD detector with high-resolution intensity images from a shortwave infrared (SWIR) InGaAs detector, we produced upsampled high-resolution FLI images. The DL model, featuring LSTM and 3D convolutional layers, was trained on simulated temporal point spread functions and validated with tissue-mimicking phantoms. Results showed highly accurate upsampled images, with future testing planned for tumor xenograft imaging to study tumor heterogeneity.