Prostate Cancer Characterization Based on Rapid SHG Imaging of Collagen Fiber Combined With Denoising Algorithm

Second harmonic generation (SHG) imaging technique can specifically image components such as collagen fibers, which provides diagnostic information for prostate cancer (PCa). Gleason grading serves as the primary criterion for assessing PCa. However, obtaining high signal-to-noise ratio (SNR) SHG images requires prolonged acquisition, limiting clinical feasibility. In this study, rapid SHG imaging was combined with a deep-learning-based denoising network, Selective Residual M-Net (SRMNet), to reconstruct high-fidelity images from low-SNR inputs. The denoising procedure markedly enhanced image quality while preserving collagen alignment features essential for quantitative assessment. The enhanced image fidelity enabled more robust extraction of collagen orientation metrics under low-SNR conditions, facilitating accurate assessment of stromal organization with reduced acquisition time. These results highlight that deep-learning-assisted SHG imaging provides an effective strategy for reliable collagen orientation analysis in PCa, supporting the development of rapid and objective optical histopathology.