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

Image registration is essential for medical image applications where alignment of voxels across multiple images is needed for qualitative or quantitative analysis. With recent advances in deep neural networks and parallel computing, deep learning-based medical image registration methods become competitive with their flexible modeling and fast inference capabilities. However, compared to traditional optimization-based registration methods, the speed advantage may come at the cost of registration performance at inference time. Besides, deep neural networks ideally demand large training datasets while optimization-based methods are training-free. To improve registration accuracy and data efficiency, we propose a novel image registration method, termed Recurrent Inference Image Registration (RIIR) network. RIIR is formulated as a meta-learning solver for the registration problem in an iterative manner. RIIR addresses the accuracy and data efficiency issues, by learning the update rule of optimization, with implicit regularization combined with explicit gradient input. We extensively evaluated RIIR on brain MRI, lung CT, and quantitative cardiac MRI datasets, in terms of both registration accuracy and training data efficiency. Our experiments showed that RIIR outperformed a range of deep learning-based methods, even with only 5% of the training data, demonstrating high data efficiency. Key findings from our ablation studies highlighted the important added value of the hidden states introduced in the recurrent inference framework for meta-learning. Our proposed RIIR offers a highly data-efficient framework for deep learning-based medical image registration.

In this work, we aimed to develop and evaluate multi-band outer volume suppression pulses for increased acceleration rates in simultaneous multi-slice accelerated cardiac MRI. MB-OVS pulses were constructed from a multi-band combination of two slab-selective saturation pulses and tested for various pulse shapes using Bloch simulation and phantom experiment. The MB-OVS pulses were interleaved between imaging pulses to ensure homogeneous suppression throughout the cardiac cycle/imaging window in vivo. Simultaneous multi-slice (SMS) CINE and first-pass myocardial perfusion scans with and without the proposed MB-OVS pulses were compared in terms of residual artifacts at high acceleration rates. Among the tested pulses, both Bloch simulation and phantom experiments showed that amplitude-optimized sinc pulses provided the best trade-off in suppression efficiency, the required $B_1^{+}$, SAR, and slab profile. CINE imaging with 5-fold SMS-OVS acceleration significantly outperformed imaging without MB-OVS, maintaining leakage-free image quality, even when adding 2-fold in-plane acceleration. SMS-OVS also enabled perfusion imaging in 9 slices with 1.7 × 1.7 mm² resolution, achieving a 16-fold spatial-only acceleration while ensuring accurate contrast dynamics without leakage artifacts. Interleaved MB-OVS modules enabled thorough leakage artifact suppression in cardiac SMS-accelerated CINE and perfusion imaging, particularly at high acceleration rates. The proposed approach may be promising for unlocking further acceleration potential of SMS in cardiac imaging.

$T_{1\rho }$ mapping is emerging as a potential, contrast-free alternative to late gadolinium enhancement (LGE) for assessment of myocardial viability. However, strong signal contributions from the blood pool can impede quantitative evaluation at the (sub)endocardium. In this work, we study the effectiveness of dark-blood (DB) contrast in adiabatic $T_{1\rho }$ ($T_{1\rho ,adiab}$) mapping at 3T, using slice-selective and non-selective adiabatic spin-lock pulses. Adiabatic DB-$T_{1\rho ,adiab}$ preparations consisted of an odd number of slice-selective adiabatic full passage (AFP) pulses, followed by a final, non-selective AFP pulse. This preparation induces $T_{1\rho ,adiab}$ decay within the imaging slice while inverting the magnetization outside. A delay ($\delta$) between preparation and imaging allowed for relaxation and inflow of the inverted blood to achieve DB contrast. Bias and precision of DB- and bright-blood (BB)-$T_{1\rho ,adiab}$ were compared in phantom and in healthy subjects (n = 10). Blood suppression efficacy and apparent myocardial thickness in DB imaging were investigated in simulations, phantom, and in vivo. The clinical feasibility of DB-$T_{1\rho ,adiab}$ mapping was evaluated in a small cohort of patients (n = 7) with suspected cardiovascular diseases. DB-$T_{1\rho ,adiab}$ values were in agreement with reference BB-$T_{1\rho ,adiab}$ values in phantom (myocardium-like vial BB: 219.27 ± 4.80 ms, DB: 218.09 ± 8.22 ms) and in healthy subjects (BB: 182.32 ± 28.27 ms, DB: 183.49 ± 45.54 ms). A moderate increase in intra-($wC{V}_{i,r}$) and inter-scan variability ($\overline{wC{V}_i}$) was observed in the DB method, compared with conventional BB imaging, for phantom and healthy subjects (in vivo $wC{V}_{i,r}$ BB: 15.51 ± 2.65%, DB: 24.82 ± 4.18%; in vivo $\overline{wC{V}_i}$ BB: 3.38 ± 0.86%, DB: 7.24 ± 2.55%). Longer delay times improved blood suppression in vivo for DB-$T_{1\rho ,adiab}$, albeit at increased intra-scan variability in phantom and in vivo (DB $wC{V}_{i,r}$ for $\delta$ = 0 ms: 4.90 ± 0.83% in phantom, 13.44 ± 2.91% in vivo, for $\delta$ = 600 ms: 8.14 ± 1.88% in phantom, 26.25 ± 5.19% in vivo). Average apparent myocardial thickness was slightly higher when using DB-$T_{1\rho ,adiab}$ compared with BB-$T_{1\rho ,adiab}$ (BB: 7.33 ± 2.05 mm, DB: 7.99 ± 2.46 mm). DB-$T_{1\rho ,adiab}$ maps yielded comparable image quality to BB-$T_{1\rho ,adiab}$ maps in patients. DB-$T_{1\rho ,adiab}$ mapping represents an alternative to BB-$T_{1\rho ,adiab}$ for myocardial assessment with the potential for improved visualization of the (sub-)endocardium.