Simplified in vivo corneal confocal volumetric imaging using adaptive axial scanning and computational registration

Corneal confocal microscopy (CCM) provides optical sections with cellular-level detail, but large-depth volumetric visualization is limited by acquisition time and inter-frame motion. Here, we present a simplified optical-digital framework that combines a motorized objective module with a clinically familiar TomoCap contact interface, layer-aware MobileNetV3 speed control, expanded-canvas coarse registration, and dynamic-window VoxelMorph refinement. In 11 matched same-eye sequence pairs, adaptive-speed acquisition produced lower image-estimated stack drift across the tested depth range. Across four test sequences, coarse plus VXM produced the highest adjacent-frame NCC and the lowest masked MAE among the tested processing variants. The registered rabbit and human stacks supported large-depth volumetric visualization and oblique viewing with cellular-level detail. Deformation-field, mask-based, and expert-review analyses found no widespread geometric folding or visually suspicious failure in the reviewed cases. This framework provides a practical route for converting in vivo CCM sequences into interpretable depth-continuous volume representations while retaining cellular-level confocal detail.