Rapid, label-free differentiation of benign and malignant gastric lesions using two-photon microscopy and quantitative collagen analysis

Previous investigations have explored two-photon microscopy (TPM) for differentiating normal and malignant lesions in the digestive tract. The aim of this study is to evaluate label-free TPM for rapid, ex vivo classification of fresh endoscopic human gastric tissues across the spectrum of neoplastic progression, identify quantitative imaging biomarkers, and establish a basis for near real-time diagnostics.

Sixty fresh gastric samples (8 normal, 34 chronic inflammation/metaplasia, 10 atypical hyperplasia, 8 malignant lesions) from 32 patients undergoing routine endoscopic biopsy were imaged using TPM without fixation/staining, generating 60 mosaics and 122 3D images. Key morphological features were established and validated by two blinded pathologists. Normal, chronic inflammation/metaplasia lesions were classified into the low-risk group, while atypical hyperplasia and malignant lesions were assigned to the high-risk group. Second-harmonic generation (SHG) quantified collagen area percentage (Area%) and orientation dispersion (Dispersion) to differentiate low-risk group from high-risk group.

TPM provided high-resolution visualization of cellular/stromal features essential for pathology, comparable to histology but significantly faster. Blinded review showed substantial inter-reader agreement (Cohen's κ = 0.793) and 80.3% diagnostic accuracy across five tissue types. Quantitative analysis revealed high-risk lesions had significantly greater collagen Dispersion (U = 684, P < 0.001, r = -0.498) and higher Area% (U = 376, P < 0.001, r = -0.644) than low-risk lesions.

TPM offers a robust, label-free, rapid method for ex vivo assessment of gastric biopsies, accurately differentiating lesions from normal to malignant lesions. Increased collagen Dispersion and Area% serve as objective biomarkers for distinguishing high-risk lesions. These findings support TPM's transformative potential for near real-time pathological evaluation, accelerating diagnostics, improving biopsy targeting, and enabling AI-assisted bedside/intraprocedural tools in gastric atypical hyperplasia and malignant lesions management.