Abstract 1449: Label free virtual HE from 3D holotomography enables reliable nuclear and microanatomical readouts in thick cancer tissues.

Key Points

• The aim is to develop a label-free method for imaging thick cancer tissues using 3D holotomography and virtual H&E staining.
• Utilized holotomography (HT) for imaging FFPE breast cancer tissues.
• Developed a diffusion-based model for image-to-image translation of HT volumes into virtual H&E.
• Employed weakly aligned chemical H&E supervision for training.
• Evaluated fidelity through nuclear morphology agreement and quantification metrics.
• Achieved 97.8% concordance with chemical H&E in whole-field nuclear counts.
• Demonstrated continuous nuclear contours across axial planes.
• Maintained robust performance across varying tissue thicknesses (4–50 µm) and organ types.
• Expert reviewers confirmed accurate chromatin texture and essential nuclear patterns for cancer grading.