Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Towards Prospective Medical Image Reconstruction via Knowledge-Informed Dynamic Optimal Transport
Authors: Taoran Zheng, Yan Yang, Xing Li, Xiang Gu, Jian Sun, Zongben Xu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on MRI and CT reconstruction demonstrate KIDOT s superior performance. Code is available at https://github.com/Taoran Zheng717/KIDOT. ... 4 Experiments ... Quantitative results on the simulated MRI dataset are presented in Table 1. |
| Researcher Affiliation | Academia | Taoran Zheng1, , Yan Yang1, , , Xing Li1, Xiang Gu1, , Jian Sun1,2, Zongben Xu1 1School of Mathematics and Statistics, Xi an Jiaotong University, Xi an, China 2State Industry-Education Integration Center for Medical Innovations at Xi an Jiaotong University EMAIL; EMAIL |
| Pseudocode | Yes | A.1 KIDOT Training Algorithm Input: Real prospective degraded dataset Y (samples y Y); high-quality dataset X (samples x X); unfolding transport network Tϕ and potential network φθ; learning rates αϕ, αθ; number of critic updates per generator update Nc. |
| Open Source Code | Yes | Code is available at https://github.com/Taoran Zheng717/KIDOT. |
| Open Datasets | Yes | Our experiments on simulated MRI data utilized the publicly accessible fast MRI multi-coil knee dataset [42]. |
| Dataset Splits | Yes | We selected 2500 fully sampled MRI slices, partitioning them into 1000 for training, 500 for validation, and 1000 for testing. ... The dataset, comprising images of size 256 256 176, was divided into 3077 training, 1860 validation, and 3077 test slices. |
| Hardware Specification | Yes | Experiments were performed using Py Torch on an NVIDIA 4090 GPU. |
| Software Dependencies | No | Experiments were performed using Py Torch on an NVIDIA 4090 GPU. |
| Experiment Setup | Yes | Optimization utilized the RMSProp algorithm with differential learning rates: 1 10 4 for the transport network (Tϕ) and 2 10 4 for the potential network (φθ). The inner iteration parameter Nc was consistently set to 1. A learning rate decay schedule applied a factor of 10 reduction after each block of 30 training epochs. |