ECLARE: Efficient cross-planar learning for anisotropic resolution enhancement

In clinical imaging, magnetic resonance (MR) image volumes are often acquired as stacks of 2D slices, permitting decreased scan times, improved signal-to-noise ratio, and image contrasts unique to 2D MR pulse sequences. While this is sufficient for clinical evaluation, automated algorithms designed for 3D analysis perform sub-optimally on 2D-acquired scans, especially those with thick slices and gaps between slices. Super-resolution (SR) methods aim to address this problem, but previous methods do not address all of the following: slice profile shape estimation, slice gap, domain shift, and non-integer / arbitrary upsampling factors. In this paper, we propose ECLARE (Efficient Cross-planar Learning for Anisotropic Resolution Enhancement), a self-SR method that addresses each of these factors. ECLARE estimates the slice profile from the 2D-acquired multi-slice MR volume, trains a network to learn the mapping from low-resolution to high-resolution in-plane patches from the same volume, and performs SR with anti-aliasing. We compared ECLARE to cubic B-spline interpolation, SMORE, and other contemporary SR methods. We used realistic and representative simulations so that quantitative performance against a ground truth could be computed, and ECLARE outperformed all other methods in both signal recovery and downstream tasks. On real data for which there is no ground truth, ECLARE demonstrated qualitative superiority over other methods as well. Importantly, as ECLARE does not use external training data it cannot suffer from domain shift between training and testing. Our code is open-source and available at https://www.github.com/sremedios/eclare.

Is Registering Raw Tagged-MR Enough for Strain Estimation in the Era of Deep Learning?

Magnetic Resonance Imaging with tagging (tMRI) has long been utilized for quantifying tissue motion and strain during deformation. However, a phenomenon known as tag fading, a gradual decrease in tag visibility over time, often complicates post-processing. The first contribution of this study is to model tag fading by considering the interplay between $T_1$ relaxation and the repeated application of radio frequency (RF) pulses during serial imaging sequences. This is a factor that has been overlooked in prior research on tMRI post-processing. Further, we have observed an emerging trend of utilizing raw tagged MRI within a deep learning-based (DL) registration framework for motion estimation. In this work, we evaluate and analyze the impact of commonly used image similarity objectives in training DL registrations on raw tMRI. This is then compared with the Harmonic Phase-based approach, a traditional approach which is claimed to be robust to tag fading. Our findings, derived from both simulated images and an actual phantom scan, reveal the limitations of various similarity losses in raw tMRI and emphasize caution in registration tasks where image intensity changes over time.

Towards an accurate and generalizable multiple sclerosis lesion segmentation model using self-ensembled lesion fusion

Automatic multiple sclerosis (MS) lesion segmentation using multi-contrast magnetic resonance (MR) images provides improved efficiency and reproducibility compared to manual delineation. Current state-of-the-art automatic MS lesion segmentation methods utilize modified U-Net-like architectures. However, in the literature, dedicated architecture modifications were always required to maximize their performance. In addition, the best-performing methods have not proven to be generalizable to diverse test datasets with contrast variations and image artifacts. In this work, we developed an accurate and generalizable MS lesion segmentation model using the well-known U-Net architecture without further modification. A novel test-time self-ensembled lesion fusion strategy is proposed that not only achieved the best performance using the ISBI 2015 MS segmentation challenge data but also demonstrated robustness across various self-ensemble parameter choices. Moreover, equipped with instance normalization rather than batch normalization widely used in literature, the model trained on the ISBI challenge data generalized well on clinical test datasets from different scanners.

Harmonization-enriched domain adaptation with light fine-tuning for multiple sclerosis lesion segmentation

Deep learning algorithms utilizing magnetic resonance (MR) images have demonstrated cutting-edge proficiency in autonomously segmenting multiple sclerosis (MS) lesions. Despite their achievements, these algorithms may struggle to extend their performance across various sites or scanners, leading to domain generalization errors. While few-shot or one-shot domain adaptation emerges as a potential solution to mitigate generalization errors, its efficacy might be hindered by the scarcity of labeled data in the target domain. This paper seeks to tackle this challenge by integrating one-shot adaptation data with harmonized training data that incorporates labels. Our approach involves synthesizing new training data with a contrast akin to that of the test domain, a process we refer to as "contrast harmonization" in MRI. Our experiments illustrate that the amalgamation of one-shot adaptation data with harmonized training data surpasses the performance of utilizing either data source in isolation. Notably, domain adaptation using exclusively harmonized training data achieved comparable or even superior performance compared to one-shot adaptation. Moreover, all adaptations required only minimal fine-tuning, ranging from 2 to 5 epochs for convergence.

Deep filter bank regression for super-resolution of anisotropic MR brain images

In 2D multi-slice magnetic resonance (MR) acquisition, the through-plane signals are typically of lower resolution than the in-plane signals. While contemporary super-resolution (SR) methods aim to recover the underlying high-resolution volume, the estimated high-frequency information is implicit via end-to-end data-driven training rather than being explicitly stated and sought. To address this, we reframe the SR problem statement in terms of perfect reconstruction filter banks, enabling us to identify and directly estimate the missing information. In this work, we propose a two-stage approach to approximate the completion of a perfect reconstruction filter bank corresponding to the anisotropic acquisition of a particular scan. In stage 1, we estimate the missing filters using gradient descent and in stage 2, we use deep networks to learn the mapping from coarse coefficients to detail coefficients. In addition, the proposed formulation does not rely on external training data, circumventing the need for domain shift correction. Under our approach, SR performance is improved particularly in "slice gap" scenarios, likely due to the constrained solution space imposed by the framework.

Contrast Adaptive Tissue Classification by Alternating Segmentation and Synthesis

Deep learning approaches to the segmentation of magnetic resonance images have shown significant promise in automating the quantitative analysis of brain images. However, a continuing challenge has been its sensitivity to the variability of acquisition protocols. Attempting to segment images that have different contrast properties from those within the training data generally leads to significantly reduced performance. Furthermore, heterogeneous data sets cannot be easily evaluated because the quantitative variation due to acquisition differences often dwarfs the variation due to the biological differences that one seeks to measure. In this work, we describe an approach using alternating segmentation and synthesis steps that adapts the contrast properties of the training data to the input image. This allows input images that do not resemble the training data to be more consistently segmented. A notable advantage of this approach is that only a single example of the acquisition protocol is required to adapt to its contrast properties. We demonstrate the efficacy of our approaching using brain images from a set of human subjects scanned with two different T1-weighted volumetric protocols.

Extracting 2D weak labels from volume labels using multiple instance learning in CT hemorrhage detection

Multiple instance learning (MIL) is a supervised learning methodology that aims to allow models to learn instance class labels from bag class labels, where a bag is defined to contain multiple instances. MIL is gaining traction for learning from weak labels but has not been widely applied to 3D medical imaging. MIL is well-suited to clinical CT acquisitions since (1) the highly anisotropic voxels hinder application of traditional 3D networks and (2) patch-based networks have limited ability to learn whole volume labels. In this work, we apply MIL with a deep convolutional neural network to identify whether clinical CT head image volumes possess one or more large hemorrhages (> 20cm$^3$), resulting in a learned 2D model without the need for 2D slice annotations. Individual image volumes are considered separate bags, and the slices in each volume are instances. Such a framework sets the stage for incorporating information obtained in clinical reports to help train a 2D segmentation approach. Within this context, we evaluate the data requirements to enable generalization of MIL by varying the amount of training data. Our results show that a training size of at least 400 patient image volumes was needed to achieve accurate per-slice hemorrhage detection. Over a five-fold cross-validation, the leading model, which made use of the maximum number of training volumes, had an average true positive rate of 98.10%, an average true negative rate of 99.36%, and an average precision of 0.9698. The models have been made available along with source code to enabled continued exploration and adaption of MIL in CT neuroimaging.

Distributed deep learning for robust multi-site segmentation of CT imaging after traumatic brain injury

Machine learning models are becoming commonplace in the domain of medical imaging, and with these methods comes an ever-increasing need for more data. However, to preserve patient anonymity it is frequently impractical or prohibited to transfer protected health information (PHI) between institutions. Additionally, due to the nature of some studies, there may not be a large public dataset available on which to train models. To address this conundrum, we analyze the efficacy of transferring the model itself in lieu of data between different sites. By doing so we accomplish two goals: 1) the model gains access to training on a larger dataset that it could not normally obtain and 2) the model better generalizes, having trained on data from separate locations. In this paper, we implement multi-site learning with disparate datasets from the National Institutes of Health (NIH) and Vanderbilt University Medical Center (VUMC) without compromising PHI. Three neural networks are trained to convergence on a computed tomography (CT) brain hematoma segmentation task: one only with NIH data,one only with VUMC data, and one multi-site model alternating between NIH and VUMC data. Resultant lesion masks with the multi-site model attain an average Dice similarity coefficient of 0.64 and the automatically segmented hematoma volumes correlate to those done manually with a Pearson correlation coefficient of 0.87,corresponding to an 8% and 5% improvement, respectively, over the single-site model counterparts.

Alternating Segmentation and Simulation for Contrast Adaptive Tissue Classification

A key feature of magnetic resonance (MR) imaging is its ability to manipulate how the intrinsic tissue parameters of the anatomy ultimately contribute to the contrast properties of the final, acquired image. This flexibility, however, can lead to substantial challenges for segmentation algorithms, particularly supervised methods. These methods require atlases or training data, which are composed of MR image and labeled image pairs. In most cases, the training data are obtained with a fixed acquisition protocol, leading to suboptimal performance when an input data set that requires segmentation has differing contrast properties. This drawback is increasingly significant with the recent movement towards multi-center research studies involving multiple scanners and acquisition protocols. In this work, we propose a new framework for supervised segmentation approaches that is robust to contrast differences between the training MR image and the input image. Our approach uses a generative simulation model within the segmentation process to compensate for the contrast differences. We allow the contrast of the MR image in the training data to vary by simulating a new contrast from the corresponding label image. The model parameters are optimized by a cost function measuring the consistency between the input MR image and its simulation based on a current estimate of the segmentation labels. We provide a proof of concept of this approach by combining a supervised classifier with a simple simulation model, and apply the resulting algorithm to synthetic images and actual MR images.

Synthesizing CT from Ultrashort Echo-Time MR Images via Convolutional Neural Networks

With the increasing popularity of PET-MR scanners in clinical applications, synthesis of CT images from MR has been an important research topic. Accurate PET image reconstruction requires attenuation correction, which is based on the electron density of tissues and can be obtained from CT images. While CT measures electron density information for x-ray photons, MR images convey information about the magnetic properties of tissues. Therefore, with the advent of PET-MR systems, the attenuation coefficients need to be indirectly estimated from MR images. In this paper, we propose a fully convolutional neural network (CNN) based method to synthesize head CT from ultra-short echo-time (UTE) dual-echo MR images. Unlike traditional $T_1$-w images which do not have any bone signal, UTE images show some signal for bone, which makes it a good candidate for MR to CT synthesis. A notable advantage of our approach is that accurate results were achieved with a small training data set. Using an atlas of a single CT and dual-echo UTE pair, we train a deep neural network model to learn the transform of MR intensities to CT using patches. We compared our CNN based model with a state-of-the-art registration based as well as a Bayesian model based CT synthesis method, and showed that the proposed CNN model outperforms both of them. We also compared the proposed model when only $T_1$-w images are available instead of UTE, and show that UTE images produce better synthesis than using just $T_1$-w images.