Abstract: To address the challenges of high computational complexity and manual parameter tuning in traditional compressed sensing based through-the-wall radar sparse imaging methods, a novel sparse imaging approach based on the adaptive threshold analytic learned iterative shrinkage thresholding network (AT-ALISTA-Net) for TWR system is proposed in this paper. The weight matrices for each layer of AT-ALISTA-Net are determined offline by solving the convex optimization problem. Only two scalar parameters need to be learned from the training data, which significantly simplifies the training process. Moreover, the AT-ALISTA-Net utilizes long short-term memory (LSTM) module to dynamically estimate parameters affecting the threshold and incorporates the adaptive thresholding shrinkage function to reconstruct the reflection coefficient vector. This approach enhances the imaging performance for weakly scattering human targets. Imaging results based on both simulated and measured data validate the effectiveness and accuracy of the proposed imaging method.