3D MRI Research Articles

A balance between native footprint coverage and overlap of the anterolateral meniscal root in tibial tunnel positioning during anterior cruciate ligament reconstruction: A 3D MRI study

BackgroundTibial footprint of anterior cruciate ligament (ACL) is situated close to the anterior lateral meniscal root (ALMR) attachment. PurposeTo investigate the impact of the size and location of the tibial tunnel for ACL reconstruction on the ACL footprint coverage and overlap to the ALMR. Study designControlled laboratory study. MethodsTwenty knee MRI scans from twenty healthy subjects were recruited, and three-dimensional (3D) tibia models were created to show the tibial attachment sites of ACL and ALMR. Surgical simulation of the tibial tunnel drilling was performed on each 3D model, entering the joint at an angle set at 60 degrees from the tibial plateau plane and 55 degrees from the posterior tibial condylar axis, with analysis for six different drill sizes; 7.5, 8, 8.5, 9, 9.5 and 10 mm; and nine locations; the center of the ACL attachment and eight locations 2% of the tibial width apart surrounding it. The width of the tibial plateau, the distance between ACL and ALMR attachment centers, and the size and location of the potential tibial tunnel were evaluated to determine association with the area of the ACL footprint coverage and ALMR overlap using a linear mixed effects model. ResultsA large tunnel (p <.001), a central and anterior location (p <.029), and small tibial width (p =.015) were all associated with larger coverage of the ACL footprint. A large tunnel (p <.001), posteriorly and laterally located (p ≤ 0.001), and a small distance between the ACL and ALMR centers (p =.001) were significantly associated with a larger ALMR overlap. The association of the tunnel size to ALMR overlap reduced with a medial tunnel location. ConclusionsThe short distance between the centers of the ALMR attachment and native ACL footprint suggests that the ALMR will always be susceptible to overlap when the tibial tunnel is drilled in ACL reconstruction. Small alterations in tunnel location can lead to a statistically significant alteration with the amount of ALMR overlap. To minimize this overlap, whilst maintaining acceptable coverage of the ACL footprint, a tibial tunnel positioned in a medial or anteromedial location from the center of the ACL footprint is recommended.

A Deep Learning Method for Quantification of Femoral Head Necrosis Based on Routine Hip MRI for Improved Surgical Decision Making.

(1) Background: To evaluate the performance of a deep learning model to automatically segment femoral head necrosis (FHN) based on a standard 2D MRI sequence compared to manual segmentations for 3D quantification of FHN. (2) Methods: Twenty-six patients (thirty hips) with avascular necrosis underwent preoperative MR arthrography including a coronal 2D PD-w sequence and a 3D T1 VIBE sequence. Manual ground truth segmentations of the necrotic and unaffected bone were then performed by an expert reader to train a self-configuring nnU-Net model. Testing of the network performance was performed using a 5-fold cross-validation and Dice coefficients were calculated. In addition, performance across the three segmentations were compared using six parameters: volume of necrosis, volume of unaffected bone, percent of necrotic bone volume, surface of necrotic bone, unaffected femoral head surface, and percent of necrotic femoral head surface area. (3) Results: Comparison between the manual 3D and manual 2D segmentations as well as 2D with the automatic model yielded significant, strong correlations (Rp > 0.9) across all six parameters of necrosis. Dice coefficients between manual- and automated 2D segmentations of necrotic- and unaffected bone were 75 ± 15% and 91 ± 5%, respectively. None of the six parameters of FHN differed between the manual and automated 2D segmentations and showed strong correlations (Rp > 0.9). Necrotic volume and surface area showed significant differences (all p < 0.05) between early and advanced ARCO grading as opposed to the modified Kerboul angle, which was comparable between both groups (p > 0.05). (4) Conclusions: Our deep learning model to automatically segment femoral necrosis based on a routine hip MRI was highly accurate. Coupled with improved quantification for volume and surface area, as opposed to 2D angles, staging and course of treatment can become better tailored to patients with varying degrees of AVN.

3D bi-directional transformer U-Net for medical image segmentation.

As one of the popular deep learning methods, deep convolutional neural networks (DCNNs) have been widely adopted in segmentation tasks and have received positive feedback. However, in segmentation tasks, DCNN-based frameworks are known for their incompetence in dealing with global relations within imaging features. Although several techniques have been proposed to enhance the global reasoning of DCNN, these models are either not able to gain satisfying performances compared with traditional fully-convolutional structures or not capable of utilizing the basic advantages of CNN-based networks (namely the ability of local reasoning). In this study, compared with current attempts to combine FCNs and global reasoning methods, we fully extracted the ability of self-attention by designing a novel attention mechanism for 3D computation and proposed a new segmentation framework (named 3DTU) for three-dimensional medical image segmentation tasks. This new framework processes images in an end-to-end manner and executes 3D computation on both the encoder side (which contains a 3D transformer) and the decoder side (which is based on a 3D DCNN). We tested our framework on two independent datasets that consist of 3D MRI and CT images. Experimental results clearly demonstrate that our method outperforms several state-of-the-art segmentation methods in various metrics.

Rapid lumbar MRI protocol using 3D imaging and deep learning reconstruction.

Three-dimensional (3D) imaging of the spine, augmented with AI-enabled image enhancement and denoising, has the potential to reduce imaging times without compromising image quality or diagnostic performance. This work evaluates the time savings afforded by a novel, rapid lumbar spine MRI protocol as well as image quality and diagnostic differences stemming from the use of an AI-enhanced 3D T2 sequence combined with a single Dixon acquisition. Thirty-five subjects underwent MRI using standard 2D lumbar imaging in addition to a "rapid protocol" consisting of 3D imaging, enhanced and denoised using a prototype DL reconstruction algorithm as well as a two-point Dixon sequence. Images were graded by subspecialized radiologists and imaging times were collected. Comparison was made between 2D sagittal T1 and Dixon fat images for neural foraminal stenosis, intraosseous lesions, and fracture detection. This study demonstrated a 54% reduction in total acquisition time of a 3D AI-enhanced imaging lumbar spine MRI rapid protocol combined with a sagittal 2D Dixon sequence, compared to a 2D standard-of-care protocol. The rapid protocol also demonstrated strong agreement with the standard-of-care protocol with respect to osseous lesions (κ = 0.88), fracture detection (κ = 0.96), and neural foraminal stenosis (ICC > 0.9 at all levels). 3D imaging of the lumbar spine with AI-enhanced DL reconstruction and Dixon imaging demonstrated a significant reduction in imaging time with similar performance for common diagnostic metrics. Although previously limited by long postprocessing times, this technique has the potential to enhance patient throughput in busy radiology practices while providing similar or improved image quality.

Mild behavioral impairment linked to progression to Alzheimer's disease and cortical thinning in amnestic mild cognitive impairment.

Mild behavioral impairment (MBI) is a neurobehavioral syndrome characterized by later life emergence of sustained neuropsychiatric symptoms, as an at-risk state for dementia. However, the associations between MBI and a risk of progression to Alzheimer's disease (AD) and its neuroanatomical correlates in mild cognitive impairment (MCI) are still unclear. A total 1,184 older adults with amnestic MCI was followed for a mean of 3.1 ± 2.0 years. MBI was approximated using a transformation algorithm for the Neuropsychiatric Inventory at baseline. A two-step cluster analysis was used to identify subgroups of individuals with amnestic MCI based on profiles of 5 MBI domain symptoms (decreased motivation, affective dysregulation, impulse dyscontrol, social inappropriateness, abnormal perception/thought content). A Cox regression analysis was applied to investigate differences in the risk of progression to AD between subgroups. A subset of participants (n = 202) underwent 3D T1-weighted MRI scans at baseline and cortical thickness was compared between the subgroups of amnestic MCI patients. The cluster analysis classified the patients into 3 groups: (1) patients without any MBI domain symptoms (47.4%, asymptomatic group); (2) those with only affective dysregulation (29.4%, affective dysregulation group); (3) those with multiple MBI domain symptoms, particularly affective dysregulation, decreased motivation and impulse dyscontrol (23.2%, complex group). Compared to the asymptomatic group, the complex group was associated with a higher risk of progression to AD (hazard ratio = 2.541 [1.904-3.392], p < 0.001), but the affective dysregulation group was not (1.214 [0.883-1.670], p = 0.232). In cortical thickness analysis, the complex group revealed cortical thinning bilaterally in the inferior parietal, lateral occipital, lateral superior temporal, and frontopolar regions compared with the affective dysregulation group. The multiple co-occuring MBI domains in individuals with amnestic MCI are associated with a higher risk of progression to AD and cortical thinning in temporal, parietal and frontal areas. These results suggest that evaluation of MBI could be useful for risk stratification for AD and appropriate intervention in MCI individuals.

Non-Invasive In Vivo Estimation of HbA1c Using Monte Carlo Photon Propagation Simulation: Application of Tissue-Segmented 3D MRI Stacks of the Fingertip and Wrist for Wearable Systems

The early diagnosis of diabetes mellitus in normal people or maintaining stable blood sugar concentrations in diabetic patients requires frequent monitoring of the blood sugar levels. However, regular monitoring of the sugar levels is problematic owing to the pain and inconvenience associated with pricking the fingertip or using minimally invasive patches. In this study, we devise a noninvasive method to estimate the percentage of the in vivo glycated hemoglobin (HbA1c) values from Monte Carlo photon propagation simulations, based on models of the wrist using 3D magnetic resonance (MR) image data. The MR image slices are first segmented for several different tissue types, and the proposed Monte Carlo photon propagation system with complex composite tissue support is then used to derive several models for the fingertip and wrist sections with different wavelengths of light sources and photodetector arrangements. The Pearson r values for the estimated percent HbA1c values are 0.94 and 0.96 for the fingertip transmission- and reflection-type measurements, respectively. This is found to be the best among the related studies. Furthermore, a single-detector multiple-source arrangement resulted in a Pearson r value of 0.97 for the wrist. The Bland-Altman bias values were found to be -0.003 ± 0.36, 0.01 ± 0.25, and 0.01 ± 0.21, for the two fingertip and wrist models, respectively, which conform to the standards of the current state-of-the-art invasive point-of-care devices. The implementation of these algorithms will be a suitable alternative to the invasive state-of-the-art methods.

Generating Realistic Brain MRIs via a Conditional Diffusion Probabilistic Model.

As acquiring MRIs is expensive, neuroscience studies struggle to attain a sufficient number of them for properly training deep learning models. This challenge could be reduced by MRI synthesis, for which Generative Adversarial Networks (GANs) are popular. GANs, however, are commonly unstable and struggle with creating diverse and high-quality data. A more stable alternative is Diffusion Probabilistic Models (DPMs) with a fine-grained training strategy. To overcome their need for extensive computational resources, we propose a conditional DPM (cDPM) with a memory-efficient process that generates realistic-looking brain MRIs. To this end, we train a 2D cDPM to generate an MRI subvolume conditioned on another subset of slices from the same MRI. By generating slices using arbitrary combinations between condition and target slices, the model only requires limited computational resources to learn interdependencies between slices even if they are spatially far apart. After having learned these dependencies via an attention network, a new anatomy-consistent 3D brain MRI is generated by repeatedly applying the cDPM. Our experiments demonstrate that our method can generate high-quality 3D MRIs that share a similar distribution to real MRIs while still diversifying the training set. The code is available at https://github.com/xiaoiker/mask3DMRI_diffusion and also will be released as part of MONAI, at https://github.com/Project-MONAI/GenerativeModels.

Performance analysis of 3D brain MRI modalities for brain tumour sub-region segmentation using U-Net architecture

Performance analysis of 3D brain MRI modalities for brain tumour sub-region segmentation using U-Net architecture

Aging and Sex Differences in Brain Volume and Cerebral Blood Flow.

How do regional brain volume ratios and cerebral blood flow (CBF, mL/min) change with aging, and are there sex differences? This study aimed to comprehensively evaluate the relationships between regional brain volume ratios and CBF in healthy brains. The study participants were healthy volunteers who underwent three-dimensional T1-weighted MRI, time-of-flight MR angiography, and four-dimensional (4D) flow MRI between 2020 and 2022. The brain was automatically segmented into 21 brain subregions from 3D T1-weighted MRI, and CBF in 16 major intracranial arteries were measured by 4D flow MRI. The relationships between segmented brain volume ratios and CBFs around the circle of Willis were comprehensively investigated in each decade and sex. This study included 129 healthy volunteers (mean age ± SD, 48.2 ± 16.8; range, 22-92 years; 43 males and 86 females). The association was strongest between the cortical gray matter volume ratio and total outflow of the intracranial major arteries distal to the circle of Willis (Pearson's correlation coefficient, r: 0.425). In addition, the mean flow of the total inflow and outflow around the circle of Willis were significantly greater in women than men, and significant left-right differences were observed in CBFs even on the peripheral side of the circle of Willis. Moreover, the correlation was strongest between the left cortical gray matter volume ratio and the combined flows of the left anterior and posterior cerebral arteries distal to the circle of Willis (r: 0.486). There was a trend toward greater total intracranial CBF, especially among women in their 40s and younger, who had a larger cortical gray matter volume. This finding may be one of the reasons for the approximately twofold higher incidence of cerebral aneurysms and subarachnoid hemorrhage, and a threefold higher incidence of migraine headaches.

Objectively-captured Changes in Trigeminal Fibers before and after Microvascular Decompression Using 3D T2-SPACE MRI Might Relate to Eventual Residual Symptoms.

Although microvascular decompression (MVD) is a reliable treatment for trigeminal neuralgia (TN), neurosurgeons sometimes encounter patients whose symptoms do not improve postoperatively or who experience good treatment efficacy but develop other sensory disturbances. This study aims to objectively identify changes in nerve fibers before and after surgery by MRI and to clarify the relationship between the changes and residual postoperative symptoms. We retrospectively analyzed data from 36 consecutive patients who underwent MVD for classical TN at our hospital between November 2019 and November 2020. Cases that fulfilled the diagnostic criteria for multiple sclerosis were excluded. We confirmed the changes on the brainstem side of the trigeminal nerve preoperatively and at seven days postoperatively using 3D T2-SPACE MRI, in which the patients were divided into three groups: preoperative T2 high intensity positive (A), postoperative T2 high intensity positive (B), and no T2 high-intensity region (C). The primary outcome measures were therapeutic efficacy and frequency of postoperative numbness. The results of MVD surgery were evaluated one year postoperatively. The percentage of cases in which treatment outcomes were rated as excellent or good at one year: group A: 0 (0%), group B: 6 (100%), and group C: 25 (96.2%) (p < 0.05); the frequency of numbness: 2 (50%) in group A, 3 (50%) in group B, and 1 (3.8%) in group C, indicating significant differences between the three groups (p < 0.05). 3D T2-SPACE MRI sequences can be used to identify changes in trigeminal nerve fibers before and after MVD, which might correlate with eventual residual symptoms.

Semi-supervised dual-task cross-consistency constraint network for 3D cardiac MRI segmentation

目的 针对人体组织器官及病灶区域的3维图像分割是计算机辅助医疗诊断的重要前提，是医学影像3维可视化的重要技术基础。深度学习方法在医学图像分割任务中的成功通常取决于大量有标注数据。半监督学习利用未标注数据容易获取的优点，在模型训练过程中使用少量标注数据和大量未标注数据进行学习，缓解了数据标注昂贵耗时的问题，在医学图像分割中受到了广泛关注。为更好地利用无标注数据，提升医学图像分割效果，提出一种新的一致性正则方法用于半监督3维医学图像分割。方法 模型以V-Net为基础架构，通过扩展网络结构，在均带有分割任务及回归任务属性的双任务主副解码器之间添加了用于正则化约束的交叉损失，构建了具有形状感知的基于双任务的交叉一致性正则网络SACC-Net （shape-aware cross-consistency regular network based on dual tasks），实现将数据层面和模型层面的扰动融合进多任务机制的一致性正则方法，使模型能够更好地利用未标注数据的有效先验信息，并且具有更好的泛化性能。结果 在MICCAI 2018（Medical Image Computing and Computer Assisted Intervention Society）心房分割挑战赛公布的数据集中的3维左心房核磁共振成像上验证本文算法，在仅使用训练集中10%的有标注数据的实验组中，Dice系数、Jaccard指数、HD（Hausdorff distance）距离和平均对称表面距离分别为88.01%、78.89%、8.19和2.09。在另一组仅使用20%的有标注数据的实验中Dice系数、Jaccard指数、HD距离和平均对称表面距离分别达到90.11%、82.11%、6.57和1.78。结论 本文提出的半监督分割模型性能显著，综合了数据、模型和任务层面的一致性正则约束，与其他半监督方法相比分割效果更好且具有更佳的泛化性能。;Objective Three-dimensional（3D）image segmentation of human tissues，organs and lesion areas is projected for computer-aided diagnosis and medical images-related 3D visualization. Thanks to the emerging deep learning technique，fully-supervised network models have been developing intensively in relevant to medical image segmentation tasks. However，it is challenged for a large amount of annotated data and 3D image segmentation data-labeled is costly and inefficient. Semi-supervised learning is focused on a small size of data-labeled and sufficient data-unlabeled in terms of easy acquisition of unlabeled data，which can alleviate the cost and time-consuming problem of data labeling. Our research is focused on new consistency regular method for semi-supervised 3D medical image segmentation model. To improve the medical image segmentation effect，our model can use unlabeled data through the fusion of different consistency methods. Method The network model is demonstrated on the V-Net，which can remove the residual structure of encoding and decoding. To get efficient features of unlabeled data，the proposed shape-aware cross-consistency regular network is introduced via the V-Net network structure extension on the basis of an encoder and two independent decoders-involved dual tasks （shape-aware cross-consistency regular network based on dual tasks（SACC-Net）），which are divided into a main decoder a and an auxiliary decoder b. The output of encoder-shared is transmitted to the two decoders after noise disturbance. At the same time，the two decoders can output the prediction results after each iteration. To increase the generalization and anti-noise ability of the model，it can minimize the difference between the two parts of the results during the training process. Additionally，the proportion of labeled samples in the training samples is extremely small because the feature distributions between the pre-processed medical image samples are relatively similar. To improve the learning ability of the model to segmented samples further，geometric prior information constraints are melted into the segmentation target. A shape-aware regression layer is added at the end of each decoder as well. During the training phase，each decoder can output two parts of the prediction results at the same time. That is，the total output of each iteration-after network consists of four parts. It can be used to decode the segmentation map SA and the signed distance map output by the decoder A，and the segmentation map SB and the signed distance map output by the decoder B through the dual-task consistency of each decoding part. To enhance the model’s ability and learn the effective features of segmentation targets to a greater extent， constraints and cross constraints can be used to realize a consistent regular method that combines data-level and modellevel disturbances with multi-task mechanisms，and make better use of unlabeled data. Result Our algorithm is validated on the MRI data set published in Atrium Segmentation Challenge held by MICCAI（Medical Image Computing and Computer Assisted Intervention Society）in 2018. The experiment is divided into two test groups based on the amounts proportion of labeled data. In the training set，10% annotated data is used only in the experimental group，the Dice coefficient， Jaccard index，HD （Hausdorff distance） distance，and average symmetric surface distance is reached to 88. 01%， 78. 89%，8. 19，and 2. 09 of each. In the other group，20% annotated data of the experiments are used only. The median Dice coefficient，Jaccard index，HD distance and average symmetric surface distance can be reached to 90. 14%， 82. 11%，6. 57，and 1. 78 each as well. Furthermore，In respect of the shape perception method using the level set function for regression tasks，the Dice evaluation index can be improved by 0. 69 and 0. 60 in comparison with shape-aware semi-supervised net（SASS Net）in 10% and 20% of the marked training results. Each improvement is reached to 1. 44% and 0. 72% in terms of comparative results of dual-task consistency（DTC）trained with 10% and 20% labeled data. Conclusion The semi-supervised segmentation model（SACC-Net）is illustrated for the criteria optimization for both of the region and boundary-based segmentation，which can incorporate the level-consistency of its data，model and task. The constrained method has its potential segmentation effect and generalization performance for semi-supervised methods.

Lesion location across diagnostic regions in multiple sclerosis

BackgroundLesions in the periventricular, (juxta)cortical, and infratentorial region, as visible on brain MRI, are part of the diagnostic criteria for Multiple sclerosis (MS) whereas lesions in the subcortical region are currently only a marker of disease activity. It is unknown whether MS lesions follow individual spatial patterns or whether they occur in a random manner across diagnostic regions. AimFirst, to describe cross-sectionally the spatial lesion patterns in patients with MS. Second, to investigate the spatial association of new lesions and lesions at baseline across diagnostic regions. MethodsExperienced neuroradiologists analyzed brain MRI (3D, 3T) in a cohort of 330 early MS patients. Lesions at baseline and new solitary lesions after two years were segmented (manually and by consensus) and classified as periventricular, (juxta)cortical, or infratentorial (diagnostic regions) or subcortical—with or without Gadolinium-enhancement. Gadolinium enhancement of lesions in the different regions was compared by Chi square test. New lesions in the four regions served as dependent variable in four zero-inflated Poisson models each with the six independent variables of lesions in the four regions at baseline, age and gender. ResultsAt baseline, lesions were most often observed in the subcortical region (mean 13.0 lesions/patient), while lesion volume was highest in the periventricular region (mean 2287 µl/patient). Subcortical lesions were less likely to show gadolinium enhancement (3.1 %) than juxtacortical (4.3 %), periventricular (5.3 %) or infratentorial lesions (7.2 %). Age was inversely correlated with new periventricular, juxtacortical and subcortical lesions. New lesions in the periventricular, juxtacortical and infratentorial region showed a significant autocorrelative behavior being positively related to the number of lesions in the respective regions at baseline. New lesions in the subcortical region showed a different behavior with a positive association with baseline periventricular lesions and a negative association with baseline infratentorial lesions. ConclusionAcross regions, new lesions do not occur randomly; instead, new lesions in the periventricular, juxtacortical and infratentorial diagnostic region are associated with that at baseline. Lesions in the subcortical regions are more closely related to periventricular lesions. Moreover, subcortical lesions substantially contribute to lesion burden in MS but are less likely to show gadolinium enhancement (than lesions in the diagnostic regions).

An 3D MRI Unsupervised Graph-based Skull Stripping Algorithm

Brain tumours and strokes are increasingly common, even from a young age. For this reason preliminary screenings, such as MRI scans, are crucial steps in detecting the abnormalities and finding a treatment. Given the increasing number of people developing different brain problems, there is a need in developing automated segmentation systems. Regardless the various methods presented in the literature, the lack of ground truth images determines the researchers to move towards unsupervised techniques.This paper presents an modified graph-based unsupervised brain segmentation method that uses a minimal spanning tree to segment brain from non-brain tissue. The original method uses human intervention for node selection. The aim is to eliminate human intervention and reduce the computational time required for brain segmentation.The presented method was compared with the original one in two ways: by resizing and scaling the images, and using images without any pre-processing step. Experimental results were obtained on the NFBS dataset and in both experiments the proposed approach delivers better results, both visually and numerically. Performance measures have improved, leading to better overall segmentation. The new method increased the precision and dice coefficient by 20%, resulting in a more accurate segmentation.

Data Augmentation-Driven Constantly Corrected Mean Teacher for Semi-Supervised 3D MRI Segmentation

Data Augmentation-Driven Constantly Corrected Mean Teacher for Semi-Supervised 3D MRI Segmentation

Three dimensional reconstruction of brain tumours from 2D MRI scans: optimised curve fitting process

Three dimensional reconstruction of brain tumours from 2D MRI scans: optimised curve fitting process

Deep learning-based Automated Localisation of Anterior Commissure and Posterior Commissure Landmarks in 3D space from three-plane 2D MRI localiser slices of the brain

The operator first performs the three-plane MRI localiser slices acquisition protocol during brain MRI scan acquisition. Based on various anatomical landmarks such as anterior commissure AC), posterior commissure (PC), and mid-sagittal plane (MSP), the operator plans MRI position and orientation before a full brain MRI scan which is essential for good quality MRI. Automatic localisation of these landmarks is vital to automatise the process and minimise operator error. Prior approaches focused on automated AC and PC detection on 2D mid-sagittal MRI slices. However, improper head positioning leads to improper 2D mid-sagittal MRI slice; therefore, it may impact the localisation error, and the localisation is not in 3D space with respect to brain volume. In the present work, the AC and PC landmarks' locations were predicted in 3D space from three-plane 2D MRI localiser slices using a convolutional neural network-based approach. Six publically available brain MRI datasets were used. The mean AC and PC localisation error obtained was less than 2mm in a within-dataset evaluation and less than 3mm in a cross-dataset evaluation.

Three dimensional reconstruction of brain tumours from 2D MRI scans: optimised curve fitting process

Three dimensional reconstruction of brain tumours from 2D MRI scans: optimised curve fitting process

Real-time 3D MRI reconstruction from cine-MRI using unsupervised network in MRI-guided radiotherapy for liver cancer.

Respiration has a major impact on the accuracy of radiation treatment for thorax and abdominal tumours. Instantaneous volumetric imaging could provide precise knowledge of tumour and normal organs' three-dimensional (3D) movement, which is the key to reducing the negative effect of breathing motion. Therefore, this study proposed a real-time 3D MRI reconstruction method from cine-MRI using an unsupervised network. Cine-MRI and setup 3D-MRI from eight patients with liver cancer were utilized to establish and validate the deep learning network for 3D-MRI reconstruction. Unlike previous methods that required 4D-MRI for network training, the proposed method utilized a reference 3D-MRI and cine-MRI to generate the training data. Then, a network was trained in an unsupervised manner to estimate the relationship between the cine-MRI acquired on coronal plane and deformation vector field (DVF) that describes the patient's breathing motion. After the training process, the coronal cine-MRI were inputted into the network, and the corresponding DVF was obtained. By wrapping the reference 3D-MRI with the generated DVF, the 3D-MRI could be reconstructed. The reconstructed 3D-MRI slices were compared with the corresponding phase-sorted cine-MRI using dice similarity coefficients (DSCs) of liver contours and blood vessel localization error. In all patients, the liver DSC had mean value >96.1% and standard deviation<1.3%; the blood vessel localization error had mean value <2.6mm, and standard deviation was <2.0mm. Moreover, the time for 3D-MRI reconstruction was approximately 100ms. These results indicated that the proposed method could accurately reconstruct the 3D-MRI in real time. The proposed method could accurately reconstruct the 3D-MRI from cine-MRI in real time. This method has great potential in improving the accuracy of radiotherapy for moving tumours.

Specific features of designing a database for neuro-oncological 3D MRI images to be used in training artificial intelligence

The research was aimed at analyzing current approaches to the organization and design methodology of visualization database built on the basis of computer vision . Such approaches are necessary for effective development of diagnostic systems using artificial intelligence (AI). A training data set of high quality is a mandatory prerequisite for that . Material and methods . The paper presents the technology for designing an annotated database (SBT Dataset ) that contains about 1000 clinical cases based on the archived data acquired by the Federal Neurosurgical Center, Novosibirsk , Russia including data on patients with astrocytoma , glioblastoma , meningioma , neurinoma , and patients with metastases of somatic tumors . Each case is represented by a preoperative MRI. The Results and discussion . The dataset was built (SBT Dataset ) containing segmented 3D MRI images of 5 types of brain tumors with 991 verified observations . Each case is represented by four MRI sequences T1-WI, T1C ( with Gd-contrast ), T2-WI and T2-FLAIR with histological and histochemical postoperative confirmation . Tumors segmentation with verification of the tumor core elements boundaries and perifocal edema was approved by two certified experienced neuroradiologists . Conclusion . The database built during the research is comparable in its volume and quality ( verification level ) with the state-of-the-art databases . The methodological approaches proposed in this paper were focused on designing the high-quality medical computer vision systems . The database was used to create artificial intelligence systems with the “ physician assistant ” functions for preoperative MRI diagnostics in neurosurgery.

Quantifying intrafractional gastric motion using auto-segmentation on MRI: Deformation and respiratory-induced displacement compared.

For accurate pre-operative gastric radiotherapy, intrafractional changes must be taken into account. The aim of this study is to quantify local gastric deformations and compare these deformations with respiratory-induced displacement. Coronal 2D MRI scans (15-16min; 120 repetitions of 25-27 interleaved slices) were obtained for 18 healthy volunteers. A deep-learning network was used to auto-segment the stomach. To separate out respiratory-induced displacements, auto-segmentations were rigidly shifted in superior-inferior (SI) direction to align the centre of mass (CoM) within every slice. From these shifted auto-segmentations, 3D iso-probability surfaces (isosurfaces) were established: a reference surface for POcc =0.50 and 50 other isosurfaces (from POcc =0.01 to 0.99), with POcc indicating the probability of occupation by the stomach. For each point on the reference surface, distances to all isosurfaces were determined and a cumulative Gaussian was fitted to this probability-distance dataset to obtain a standard deviation (SDdeform ) expressing local deformation. For each volunteer, we determined median and 98th percentile of SDdeform over the reference surface and compared these with the respiratory-induced displacement SDresp , that is, the SD of all CoM shifts (paired Wilcoxon signed-rank, α=0.05). Larger deformations were mostly seen in the antrum and pyloric region. Median SDdeform (range, 2.0-2.9mm) was smaller than SDresp (2.7-8.8mm) for each volunteer (p<0.00001); 98th percentile of SDdeform (3.2-7.3mm) did not significantly differ from SDresp (p=0.13). Locally, gastric deformations can be large. Overall, however, these deformations are limited compared to respiratory-induced displacement. Therefore, unless respiratory motion is considerably reduced, the need to separately include these deformation uncertainties in the treatment margins may be limited.