<p>Positron emission tomography (PET) serves as an essential tool for diagnosis of encephalopathy and brain science research. However, its utility is limited by the availability of tracers. Recently, with the growing use of PET imaging in neuropsychiatric treatment, 6-¹⁸F-fluoro-3,4-dihydroxy-L-phenylalanine (DOPA) has shown to be more effective than ¹⁸F-labeled fluorodeoxyglucose (FDG). Despite this, DOPA is less commonly used than FDG due to its complex preparation and other limitations. To address this issue, we proposed a tracer conversion invertible neural network (TC-INN) for image projection, aiming to map FDG images to DOPA images through deep learning. This approach allows for the generation of PET images from FDG to DOPA, thus providing more diagnostic information. Specifically, the proposed TC-INN involves two separate phases: one for training traceable data and the other for reconstructing new data. During the tracer conversion training process, the reference DOPA PET image serves as the learning target for the corresponding network. Meanwhile, the invertible network iteratively estimates the resultant DOPA PET data and compares it to the reference DOPA PET data. Notably, the reversible model employs a variable enhancement technique to achieve better power generation. Moreover, image registration needs to be performed before training due to the angular deviation of the acquired FDG and DOPA data information. Experimental results exhibited excellent generation capability in mapping between FDG and DOPA, suggesting that PET tracer conversion has great potential in the case of limited tracer applications. The absolute mean error (MAE) and root mean squared error (RMSE) of brain synthesis DOPA images are about 2.72% and 5.34%, respectively, under the 3-channel condition. The strength similarity between synthetic DOPA and reference DOPA is satisfactory. In the context of limited tracer availability, the TC-INN method presents considerable promise for PET projection imaging.</p>