In the field of biomedicine, high-throughput and multi-dimensional characterization of large-scale specimens is of critical importance for pathological diagnosis, tissue engineering, and drug screening. However, when dealing with thick or multi-layer stacked samples, traditional optical imaging techniques based on scalar models struggle to capture the rich internal anisotropic polarization characteristics due to the complex spatial evolution of polarization states and the non-linear scrambling coupling between anisotropic layers. Furthermore, because the evolution of vector light fields usually follows non-commutative Jones matrix multiplications, the forward propagation exhibits strong path dependency, making it highly ill-posed to unambiguously decouple independent multi-dimensional vector information from highly confounded diffraction patterns.
To address these challenges, we propose a high-throughput polarization-encoded multi-slice ptychographic tomography (PM-PT) method. The strategic introduction of ultra-thin film polarizers (with a thickness of approximately 0.1 mm) between adjacent target slices as deterministic physical constraints successfully decouples the complex, path-dependent vector-cascading model, reducing it to a series of independently observable scalar projection components. In the data acquisition stage, under multi-angle plane-wave illumination modulated by a programmable 2D translation stage, all polarizers within the system are collaboratively rotated to four characteristic selection angles to record a sequence of diverse sub-polarized diffraction intensity patterns on a lensless sensor covered with a random polystyrene microsphere coding layer. For reconstruction, a multi-slice beam-propagation forward model combined with an alternating projection ptychographic iterative engine is implemented to independently retrieve the high-resolution complex amplitude of each polarization channel. Subsequently, the eigenvalues and eigenvectors of the reconstructed Jones matrices are computed to quantitatively extract the slow-axis retardance and optic axis orientation distributions.
Experimental validations were thoroughly conducted on a dual-layer heterogeneous stacked sample composed of anisotropic aortic valve tissue sections and potato starch granule slices. Quantitative comparisons with a commercial polarized light microscope (PLM) confirm that the proposed PM-PT method achieves highly accurate and artifact-free depth-resolved reconstructions. Specifically, the reconstructed optic axis orientation profiles show a high agreement with the PLM ground truth, yielding a root-mean-square error (RMSE) of 3.55° and a mean absolute error (MAE) of 0.22° for the first-layer aortic valve tissue, and an RMSE of 6.57° and an MAE of 0.88° for the second-layer starch granules. Benefiting from the compact lensless computational architecture, the system simultaneously achieves a macro-scale effective field-of-view (FOV) of 59 mm2 and a depth-of-field (DOF) of approximately 3.18 mm without requiring any mechanical axial scanning or chemical staining. This work successfully bridges the gap between high-throughput scalar ptychography and multi-dimensional vector tomographic imaging, providing a precise, efficient, and label-free computational digital pathology platform for large-scale clinical diagnostics and heterogeneous biological tissue analysis.