Time-frequency mixed-norm estimates: Sparse M/EEG imaging with non-stationary source activations

► Our method solves the M/EEG inverse problem without assuming source stationarity. ► We localize sources in space, time and frequency in one step. ► We provide results on simulations and two publicly available MEG datasets. ► We use short time Fourier transforms (STFT) and modern optimization techniques. ► The solver is tractable and fast on real data.