Detection Instead of Inversion: T1-T2 Microstructure Quantification With Detection Transformers
Abstract
T1-T2 correlation MRI has strong potential to quantify microstructural compartments without imposing a biological model. Traditionally, full T1-T2 spectra per voxel are reconstructed with an ill-conditioned and time-consuming Inverse Laplace Transform. We propose to view this challenge as an object detection task: A Detection Transformer (DETR) takes 64 TI/TE signals and directly predicts a set of compartments specified by T1, T2, and signal fraction. We achieve a mean average precision mAP@10 of 76.8% on simulated data, outperforming both our previous model design (53.8%) and regularized NNLS (66.1%). We show that our DETR approach outperforms NNLS in almost every setting across compartment count, compartmental signal fraction, and SNR. Reconstruction is orders of magnitude faster than NNLS. The results are fully reproducible from our public code: github.com/SebastianEndtTHI/t1t2-detr
Related
- Fiber-Resolved Microstructure Quantification from Multi-Shell Diffusion MRI using Detection Transformers
- Transformers with Physics-Informed Encodings and Simulation-Based Inference for Robust Detection of Eccentric Binary Black Holes in Pulsar Timing Array Data
- Generalized Glauber theorem for dark-matter axion and graviton detection
- Granger causality detection with Kolmogorov-Arnold networks
- Optical detection and manipulation of pseudospin orders in Wigner crystals
BibTeX
@misc{endt2026detection,
title = {{Detection Instead of Inversion: T1-T2 Microstructure Quantification With Detection Transformers}},
author = {Sebastian Endt and Fatih Özkan and Marion I. Menzel},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.01790},
eprint = {2611.01790},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.01790},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Detection Instead of Inversion: T1-T2 Microstructure Quantification With Detection Transformers
AU - Sebastian Endt
AU - Fatih Özkan
AU - Marion I. Menzel
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.01790
UR - https://arxiv.org/abs/2611.01790
AB - T1-T2 correlation MRI has strong potential to quantify microstructural compartments without imposing a biological model. Traditionally, full T1-T2 spectra per voxel are reconstructed with an ill-conditioned and time-consuming Inverse Laplace Transform. We propose to view this challenge as an object detection task: A Detection Transformer (DETR) takes 64 TI/TE signals and directly predicts a set of compartments specified by T1, T2, and signal fraction. We achieve a mean average precision mAP@10 of 76.8% on simulated data, outperforming both our previous model design (53.8%) and regularized NNLS (66.1%). We show that our DETR approach outperforms NNLS in almost every setting across compartment count, compartmental signal fraction, and SNR. Reconstruction is orders of magnitude faster than NNLS. The results are fully reproducible from our public code: github.com/SebastianEndtTHI/t1t2-detr
ER -