An overlap invariant entropy measure of 3D medical image alignment
Abstract
This paper is concerned with the development of entropy-based registration criteria for automated 3D multi-modality medical image alignment. In this application where misalignment can be large with respect to the imaged field of view, invariance to overlap statistics is an important consideration. Current entropy measures are reviewed and a normalised measure is proposed which is simply the ratio of the sum of the marginal entropies and the joint entropy. The effect of changing overlap on current entropy measures and this normalised measure are compared using a simple image model and experiments on clinical image data. Results indicate that the normalised entropy measure provides significantly improved behaviour over a range of imaged fields of view.
References (0)
Cited by (2017)
Anatomical registration of intracranial electrodes. Robust model-based localization and deformable smooth brain-shift compensation methods
2024, Journal of Neuroscience MethodsIntracranial electrodes are typically localized from post-implantation CT artifacts. Automatic algorithms localizing low signal-to-noise ratio artifacts and high-density electrode arrays are missing. Additionally, implantation of grids/strips introduces brain deformations, resulting in registration errors when fusing post-implantation CT and pre-implantation MR images. Brain-shift compensation methods project electrode coordinates to cortex, but either fail to produce smooth solutions or do not account for brain deformations.
We first introduce GridFit, a model-based fitting approach that simultaneously localizes all electrodes’ CT artifacts in grids, strips, or depth arrays. Second, we present CEPA, a brain-shift compensation algorithm combining orthogonal-based projections, spring-mesh models, and spatial regularization constraints.
We tested GridFit on ∼6000 simulated scenarios. The localization of CT artifacts showed robust performance under difficult scenarios, such as noise, overlaps, and high-density implants (<1 mm errors). Validation with data from 20 challenging patients showed 99% accurate localization of the electrodes (3160/3192). We tested CEPA brain-shift compensation with data from 15 patients. Projections accounted for simple mechanical deformation principles with < 0.4 mm errors. The inter-electrode distances smoothly changed across neighbor electrodes, while changes in inter-electrode distances linearly increased with projection distance.
GridFit succeeded in difficult scenarios that challenged available methods and outperformed visual localization by preserving the inter-electrode distance. CEPA registration errors were smaller than those obtained for well-established alternatives. Additionally, modeling resting-state high-frequency activity in five patients further supported CEPA.
GridFit and CEPA are versatile tools for registering intracranial electrode coordinates, providing highly accurate results even in the most challenging implantation scenarios. The methods are implemented in the iElectrodes open-source toolbox.
Characterization of white matter lesions in multiple sclerosis using proton density and T1-relaxation measures
2024, Magnetic Resonance ImagingAlthough lesion dissemination in time is a defining characteristic of multiple sclerosis (MS), there is a limited understanding of lesion heterogeneity. Currently, conventional sequences such as fluid attenuated inversion recovery (FLAIR) and T1-weighted (T1W) data are used to assess MS lesions qualitatively. Estimating water content could provide a measure of local tissue rarefaction, or reduced tissue density, resulting from chronic inflammation. Our goal was to utilize the proton spin density (PD), derived from a rapid, multi-contrast STAGE (strategically acquired gradient echo) protocol to characterize white matter (WM) lesions seen on T2W, FLAIR and T1W data.
Twenty (20) subjects with relapsing-remitting MS were scanned at 3 T using T1W, T2-weighted, FLAIR and strategically acquired gradient echo (STAGE) sequences. PD and T1 maps were derived from the STAGE data. Disease severity scores, including Extended Disability Status Scale (EDSS) and Multiple Sclerosis Functional Composite (MSFC), were correlated with total, high PD and high T1 lesion volumes. A probability map of high PD regions and all lesions across all subjects was generated. Five perilesional normal appearing WM (NAWM) bands surrounding the lesions were generated to compare the median PD and T1 values in each band with the lesional values and the global WM.
T1W intensity was negatively correlated with PD as expected (R = −0.87, p < 0.01, R2 = 0.756) and the FLAIR signal was suppressed for high PD volumes within the lesions, roughly for PD ≥ 0.85. The threshold for high PD and T1 regions was set to 0.909 and 1953.6 ms, respectively. High PD regions showed a high probability of occurrence near the boundary of the lateral ventricles. EDSS score and nine-hole peg test (dominant and non-dominant hand) were significantly correlated with the total lesion volume and the volumes of high PD and T1 regions (p < 0.05). There was a significant difference in PD/T1 values between the high PD/T1 regions within the lesions and the remaining lesional tissue (p < 0.001). In addition, the PD values of the first NAWM perilesional band directly adjacent to the lesional boundary displayed a significant difference (p < 0.05) compared to the global WM.
Lesions with high PD and T1s had the highest probability of occurrence at the boundary of the lateral ventricles and likely represent chronic lesions with significant local tissue rarefaction. Moreover, the perilesional NAWM exhibited subtly increasing PD and T1 values from the NAWM up to the lesion boundary. Unlike on the T1 maps, the perilesional band adjacent to the lesion boundary possessed a significantly higher PD value than the global WM PD values. This shows that PD maps were sensitive to the subtle changes in NAWM surrounding the lesions.
Symmetrical SyncMap for imbalanced general chunking problems
2023, Physica D: Nonlinear PhenomenaRecently, SyncMap pioneered an approach to learn complex structures from sequences as well as adapt to any changes in underlying structures. This is achieved by using only nonlinear dynamical equations inspired by neuron group behaviors, i.e., without loss functions. Here we propose Symmetrical SyncMap that goes beyond the original work to show how to create dynamical equations and attractor–repeller points which are stable over the long run, even dealing with imbalanced continual general chunking problems (CGCPs). The main idea is to apply equal updates from negative and positive feedback loops by symmetrical activation. We then introduce the concept of memory window to allow for more positive updates. Our algorithm surpasses or ties other unsupervised state-of-the-art baselines in all 12 imbalanced CGCPs with various difficulties, including dynamically changing ones. To verify its performance in real-world scenarios, we conduct experiments on several well-studied structure learning problems. The proposed method surpasses substantially other methods in 3 out of 4 scenarios, suggesting that symmetrical activation plays a critical role in uncovering topological structures and even hierarchies encoded in temporal data.
LiverHccSeg: A publicly available multiphasic MRI dataset with liver and HCC tumor segmentations and inter-rater agreement analysis
2023, Data in BriefAccurate segmentation of liver and tumor regions in medical imaging is crucial for the diagnosis, treatment, and monitoring of hepatocellular carcinoma (HCC) patients. However, manual segmentation is time-consuming and subject to inter- and intra-rater variability. Therefore, automated methods are necessary but require rigorous validation of high-quality segmentations based on a consensus of raters. To address the need for reliable and comprehensive data in this domain, we present LiverHccSeg, a dataset that provides liver and tumor segmentations on multiphasic contrast-enhanced magnetic resonance imaging from two board-approved abdominal radiologists, along with an analysis of inter-rater agreement.
LiverHccSeg provides a curated resource for liver and HCC tumor segmentation tasks. The dataset includes a scientific reading and co-registered contrast-enhanced multiphasic magnetic resonance imaging (MRI) scans with corresponding manual segmentations by two board-approved abdominal radiologists and relevant metadata and offers researchers a comprehensive foundation for external validation, and benchmarking of liver and tumor segmentation algorithms. The dataset also provides an analysis of the agreement between the two sets of liver and tumor segmentations. Through the calculation of appropriate segmentation metrics, we provide insights into the consistency and variability in liver and tumor segmentations among the radiologists. A total of 17 cases were included for liver segmentation and 14 cases for HCC tumor segmentation. Liver segmentations demonstrates high segmentation agreement (mean Dice, 0.95 ± 0.01 [standard deviation]) and HCC tumor segmentations showed higher variation (mean Dice, 0.85 ± 0.16 [standard deviation]).
The applications of LiverHccSeg can be manifold, ranging from testing machine learning algorithms on public external data to radiomic feature analyses. Leveraging the inter-rater agreement analysis within the dataset, researchers can investigate the impact of variability on segmentation performance and explore methods to enhance the accuracy and robustness of liver and tumor segmentation algorithms in HCC patients. By making this dataset publicly available, LiverHccSeg aims to foster collaborations, facilitate innovative solutions, and ultimately improve patient outcomes in the diagnosis and treatment of HCC.
Deformation equivariant cross-modality image synthesis with paired non-aligned training data
2023, Medical Image AnalysisCross-modality image synthesis is an active research topic with multiple medical clinically relevant applications. Recently, methods allowing training with paired but misaligned data have started to emerge. However, no robust and well-performing methods applicable to a wide range of real world data sets exist. In this work, we propose a generic solution to the problem of cross-modality image synthesis with paired but non-aligned data by introducing new deformation equivariance encouraging loss functions. The method consists of joint training of an image synthesis network together with separate registration networks and allows adversarial training conditioned on the input even with misaligned data. The work lowers the bar for new clinical applications by allowing effortless training of cross-modality image synthesis networks for more difficult data sets.
Impact of the confluence of cardiac motion and high spatial resolution on performance of ECG-gated imaging with an investigational photon-counting CT system: A phantom study
2023, Physica MedicaPhoton-counting CT (PCCT) has higher spatial resolution that conventional EID CT which improves imaging of stationary coronary plaques and stents.. In this work, we evaluated the relationship between higher spatial resolution and motion acquisition on an investigational PCCT system.
An investigational photon-counting CT scanner (Siemens CounT) with ECG gating was used to image a coronary tree phantom with models of healthy, stenotic, and stented arteries using a motion simulator. Images were acquired with matched clinical parameters at rest and 60 beats per minute. An additional set of high dose stationary images were averaged to generate a motion-free, reduced noise reference. Scans were completed at standard (0.5 mm2) and high-resolution (0.25 mm2). Motion images were reconstructed at multiple phases. Regions of interest were drawn around vessels and segmented. Percentage difference from the reference standard was evaluated for vessel diameter and circularity. Mutual information between the reference and stationary and motion datasets was used as a measure of volumetric similarity.
The stenotic vessel showed the most variation from the reference when compared to healthy or stented vessels. Compared to standard resolution, high-resolution images had lower bias for diameter (-0.012 ± 0.19% vs −0.052 ± 0.14%) and lower variability for circularity (-0.13 ± 0.138% vs −0.12 ± 0.144%). Both differences were found to be statistically significant. High-resolution images had a slightly lower mutual information (1.28) than standard resolution (1.31).
The higher spatial resolution enabled by photon-counting CT can be harnessed for cardiac imaging as the benefits of high spatial resolution acquisitions remain relevant in the presence of motion.