In look-ahead logging-while-drilling acoustic detection, the echoes from the formation ahead of the drill bit captured by the receiver array are severely overwhelmed by complex borehole mode waves, which significantly hinder subsequent imaging analysis. Existing wavefield separation methods are limited by frequency-domain overlaps, making it difficult to effectively extract weak formation echoes. In particular,echo signals from horizontal geological boundaries perpendicular to the wellbore axis are even more difficult to separate. To this end, a method for enhancing weak geological boundary echoes based on iterative-decoupled singular spectrum analysis (ISSA) is proposed to improve the SNR and detectability of weak echo signals. First, delay embedding coding is applied to the common-source gather signals to convert the one-dimensional time series into an invertible geometric matrix structure that satisfies the requirements for matrix decomposition. Low-rank singular value decomposition technology is then used to decompose the full signals, initially separating the borehole direct waves with low-rank characteristics from the geological boundary echoes with non-low-rank characteristics. Furthermore, the direct wave suppression is reformulated as a low-rank approximation reconstruction inversion problem, and the bi-factored gradient descent algorithm is introduced to iteratively decouple the direct wave structure and amplitude characteristics. Meanwhile, a residual weighting mechanism is employed to apply dynamic non-zero weights to the direct wave reconstruction residuals to guide the gradient direction, while applying a zero-gradient protection mechanism to weak reflection waves, thereby achieving independent decoupling of both. Finally, an exponential decay step strategy is integrated to stabilize the decoupling process and accelerate convergence, achieving optimal separation results. Validation using various synthetic arrays and simulated well acoustic data demonstrates that for weak echoes from geological boundaries with different formation properties and inclination angles, the SNR improves by 5~20 dB, and the structural similarity reaches over 0.90. The method exhibits good adaptability even under strong random noise interference. The results demonstrate that the proposed method can significantly suppress in-band direct waves from the borehole while effectively preserving the waveform structure and amplitude characteristics of weak reflected waves, giving it important theoretical significance and application value.