Abstract
Near-infrared diffuse optical tomography (DOT) is an emerging functional imaging modality for breast cancer assessment that measures optical contrast reflective of tumor vascularity and metabolism. However, reconstructing 3D DOT volumes from sparsely sampled source-detector measurements is an ill-posed inverse problem. Recent deep learning (DL) approaches improve upon classical finite-element method (FEM) reconstruction, but most rely on large fully connected (FC) sensor-to-image domain mapping layers whose parameter count (and thus model size and memory footprint) scales with the output resolution. This makes higher-resolution reconstruction impractical and restricts prior work to coarse voxel grids. We propose iDOT, a breast DOT reconstruction model that uses Implicit Neural Representations (INR) to enable clinically relevant, high-resolution reconstruction. Unlike competing methods that require per-sample optimization at test time, iDOT conditions on the optical measurements to reconstruct each volume in a single forward pass. On a test set of 400 simulated breast volumes, iDOT improves contrast recovery (0.17 increase in ICC) and lesion localization (19.2 mm decrease in centroid offset) for 8 mm lesions compared to the best baseline. iDOT is also lightweight, using 99.7% fewer trainable parameters than competing methods, supporting efficient deployment while enabling high-resolution reconstructions.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Deep_Brea3th_037.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=gWWpzRVyw0
BibTex
@InProceedings{PatJay_Image_MICCAISAT2026,
author = { Patel, Jay B. AND Mireles, Miguel AND Mindroc-Filimon, Diana AND Carp, Stefan A. AND Bridge, Christopher P. AND Deng, Bin},
title = { { Image Reconstruction of Breast Diffuse Optical Tomography via Implicit Neural Representations } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17256},
month = {pending},
page = {pending}
}