Characterizing the pathological features of hepatocellular carcinoma by label-free multimodal nonlinear optical microscopy

Multimodal nonlinear optical imaging (MNOI) provided detailed morphological and biochemical information for pathological analysis without extra labeling. In this study, an integrated MNOI was developed for revealing the pathological characteristics of hepatocellular carcinoma (HCC) and its alterations during disease progression. A computational image enhancement algorithm, Zero-shot Deconvolution Network (ZS-DeconvNet), was employed to improve the spatial resolution of the captured images. After multichannel nonlinear optical imaging, second harmonic generation (SHG) signals from collagen showed marked spatial heterogeneity. Peritumoral stroma, particularly at the invasive front, showed increased collagen signals consistent with fibrosis, whereas the tumor core showed intratumoral collagen loss compared with normal tissue. These complementary patterns may be associated with tumor invasion and poor prognosis. Normal tissue surrounding neoplastic regions could be distinguished from early-stage HCC lesions based on third harmonic generation (THG) signal variations associated with changes in refractive index. Three-photon-excited fluorescence (3PEF) identified alterations in endogenous fluorophores associated with different cellular metabolic states. After image processing, image quality was improved by reducing background noise and allowing the extraction of fine structural features. This enabled quantification of collagen levels in HCC tissues with different degrees of differentiation. By characterizing the pathological features of HCC, the proposed approach provides new insights into liver cancer development and demonstrates promising prognostic potential for future clinical applications and biomedical studies.