Most source estimation techniques based on Magneto- and Electroencephalography (M/EEG) derive their estimates sample-wise and independently over time. However, neural arrangements are closely interconnected, restricting the temporal evolution of the neural activity captured by MEG and EEG. The observed neural currents must, therefore, be highly context-dependent. Here, we use a network of Long Short-Term Memory (LSTM) cells, with the input being a sequence of past source estimates and the output being a prediction of the following estimate. This prediction is then used to correct the estimate. In this study, we applied this technique to noise-normalized Minimum Norm Estimation (MNE). Since the correction is determined based on past activities (context), we refer to this implementation as Contextual MNE (CMNE), although this technique can be used in conjunction with any source estimation method. We test CMNE on simulated epileptiform activity and recorded Auditory Steady State Response (ASSR) data, demonstrating that CMNE estimates exhibit a higher degree of spatial fidelity in the tested cases compared to unfiltered estimates.