Abstract
Synthetic optical coherence tomography (OCT) is an inverse physics problem: recover a digital phantom (a cloud of optical scatterers) that a fixed Virtual Scanner renders into a given real skin B-scan. Scoring uses structural and physics-derived maps (attenuation, speckle contrast). The scanner is a compiled, forward-only binary, so prior work relies on approximate differentiable surrogates, and the dominant residual is the raw speckle, whose exact realization is ill-posed to reproduce. We re cover the scanner’s exact coherent-summation model and re-implement it as a bit-faithful, autograd-differentiable GPU twin (correlation 0.9998, 35ms/B-scan), enabling backpropagation of the exact objective. On it we build a generator: analytic attenuation initialization, a structured per pixel scatterer grid, amortized inference, and a speckle-phase refinement that moves only sub-pixel positions to align the coherent phase without touching the physics maps. On a never-tuned test split (exact scanner) it reaches MS-SSIM 0.984 / LPIPS 0.018 with near-perfect physics maps, near the speckle ceiling: independent realizations of the same structure differ at MS-SSIM ≈ 0.58. The method runs as a portable single-GPU generator within a 12GB / 600s budget
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SynthOCT_001.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/SynthOCT_001_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=eT8xnrF58D
BibTex
@InProceedings{SukDan_SpecklePhase_MICCAISAT2026,
author = { Sukhorukov, Daniil AND Antipushina, Ekaterina AND Koush, Yury AND Makarov, Ilya},
title = { { Speckle-Phase Refinement on an Exact Differentiable Scanner Twin for Inverse OCT Phantom Generation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17258},
month = {pending},
page = {pending}
}