Yidong Zhao

Accurate segmentation of cardiac MRI is essential for assessment of cardiac function through biomarkers such as the left and right ventricular ejection fraction (LVEF, RVEF). Although AI methods have achieved high average segmentation accuracy, the precision of biomarkers for individual patients – quantified by estimation variance, remains critical for reliable diagnosis. Calibrated biomarkers, whose uncertainty accurately reflects the true variability, are highly desirable. However, existing evaluations predominantly focus on population-level segmentation accuracy, leaving biomarker-level uncertainty and calibration largely underexplored. Intrinsic anatomical ambiguity and annotation variability are major sources of biomarker variability and cannot be fully eliminated, even when training on a single annotation set. To address this, we propose a probabilistic segmentation framework that explicitly models aleatoric uncertainty with the goal of improving calibration in the biomarker space. The framework disentangles two key sources of uncertainty: (1) detection uncertainty, arising from ambiguous inclusion of basal or apical slices in 2D cardiac MRI, modeled via objectness probabilities; and (2) contour uncertainty, reflecting variability in ventricular boundary delineation, modeled through mean–variance regression of elliptic Fourier descriptors, a compact representation of closed contours. By propagating these uncertainties to derived biomarkers, the proposed method produces more informative and better-calibrated confidence estimates for ejection fraction. Compared to conventional pixel-wise approaches, our framework improves biomarker reliability, particularly in realistic settings dominated by annotation ambiguity and limited domain shift.

Deformable image registration is inherently a multi-objective optimization (MOO) problem, requiring a delicate balance between image similarity and deformation regularity. These conflicting objectives often lead to poor optimization outcomes, such as being trapped in unsatisfactory local minima or experiencing slow convergence. Deep learning methods have recently gained popularity in this domain due to their efficiency in processing large datasets and achieving high accuracy. However, they often underperform during test time compared to traditional optimization techniques, which further explore iterative, instance-specific gradient-based optimization. This performance gap is more pronounced when a distribution shift between training and test data exists. To address this issue, we focus on the instance optimization (IO) paradigm, which involves additional optimization for test-time instances based on a pre-trained model. IO effectively combines the generalization capabilities of deep learning with the fine-tuning advantages of instance-specific optimization. Within this framework, we emphasize the use of gradient projection to mitigate conflicting updates in MOO. This technique projects conflicting gradients into a common space, better aligning the dual objectives and enhancing optimization stability. We validate our method using a state-of-the-art foundation model on the 3D Brain inter-subject registration task (LUMIR) from the Learn2Reg 2024 Challenge. Our results show significant improvements over standard gradient descent, leading to more accurate and reliable registration results.

Deep learning (DL)-based methods have achieved state-of-the-art performance for many medical image segmentation tasks. Nevertheless, recent studies show that deep neural net- works (DNNs) can be miscalibrated and overconfident, leading to ”silent failures” that are risky for clinical applications. Bayesian DL provides an intuitive approach to DL failure de- tection, based on posterior probability estimation. However, the posterior is intractable for large medical image segmentation DNNs. To tackle this challenge, we propose a Bayesian learning framework using Hamiltonian Monte Carlo (HMC), tempered by cold posterior (CP) to accommodate medical data augmentation, named HMC-CP. For HMC compu- tation, we further propose a cyclical annealing strategy, capturing both local and global geometries of the posterior distribution, enabling highly efficient Bayesian DNN training with the same computational budget as training a single DNN. The resulting Bayesian DNN outputs an ensemble segmentation along with the segmentation uncertainty. We evaluate the proposed HMC-CP extensively on cardiac magnetic resonance image (MRI) segmentation, using in-domain steady-state free precession (SSFP) cine images as well as out-of-domain datasets of quantitative T1 and T2 mapping. Our results show that the proposed method improves both segmentation accuracy and uncertainty estimation for in- and out-of-domain data, compared with well-established baseline methods such as Monte Carlo Dropout and Deep Ensembles. Additionally, we establish a conceptual link between HMC and the commonly known stochastic gradient descent (SGD) and provide general insight into the uncertainty of DL. This uncertainty is implicitly encoded in the training dynamics but often overlooked. With reliable uncertainty estimation, our method provides a promising direction toward trustworthy DL in clinical applications. We release our code in https://gitlab.tudelft.nl/yidongzhao/hmc_uncertainty