Spin-Echo (SE) MRSI can encode J-coupling information and is desirable for brain imaging applications. But it uses long TR, leading to long scan time and thus low resolution. Consequently, removal of subcutaneous lipid signals from low-resolution SE data is challenging. This paper presents a novel method to solve this problem. The proposed method uses a high-resolution FID reference and a neural network to transform it into SE signals, which are then used to construct a generalized-series model for lipid signal removal from the SE 1H-MRSI data. The proposed method has been tested using in vivo data, producing very encouraging results.