Xin Zhao
PhD candidate, Department of Neurophysics
Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
I am interested in adapting pretrained generative models to new tasks and conditions, through fine-tuning and test-time adaptation. My current work develops these ideas for diffusion models, with simulation-based inference and inverse problems as testbeds.
My PhD project is on deep learning reconstruction for fast, ultra-high-resolution MRI. Before that, I worked on medical image analysis, including cardiac MRI segmentation and CT-based assessment of pulmonary hypertension.
My PhD is supervised by Nikolaus Weiskopf and Nico Scherf, with Martin Weigert (TU Dresden) as external advisor.
News
- Spectra, on test-time prior adaptation for simulation-based inference, is on arXiv. [arXiv]
- Presented a poster at the Machine Learning Summer School (MLSS) 2026, Tübingen.
- Presented a poster on deep learning MRI reconstruction at OHBM 2026, Bordeaux.
- Gave a talk at the 1st SFB MR-DYNAMO Workshop, TU Graz.
Publications
Google ScholarSpectra: Exact Component Transport for Test-Time Prior Adaptation in Simulation-Based Inference
Machine Learning–Based Hemodynamics Quantitative Assessment of Pulmonary Circulation Using Computed Tomographic Pulmonary Angiography
GLRP: Global and Local Contrastive Learning Based on Relative Position for Medical Image Segmentation on Cardiac MRI
A CMR Short-Axis Images Segmentation Method Based on Multi-Attention Mechanism and Boundary Distance Map
Cardiac MRI Image Enhancement Based on GAN Network
Deep Multi-Scale Dilated Convolution Network for Coronary Artery Segmentation
Machine Learning Based on Computed Tomography Pulmonary Angiography in Evaluating Pulmonary Artery Pressure in Patients with Pulmonary Hypertension
Talks and posters
- Evidence Response: What a Probability Law Must Do When the Evidence Moves
Poster · Machine Learning Summer School (MLSS) 2026, Tübingen, Germany - Deep Learning Reconstruction of 8× Accelerated Ultra-High-Resolution Ex Vivo Chimpanzee Brain MRI
Poster · OHBM Annual Meeting, Bordeaux, France - Physical Correctness of Data Consistency in State-Space-Model-Based MRI Reconstruction
Talk · 1st SFB MR-DYNAMO Workshop (Mathematical Methods in MRI), TU Graz, Austria