Magnetic Resonance Imaging Slices Research Articles

MRI super-resolution using similarity distance and multi-scale receptive field based feature fusion GAN and pre-trained slice interpolation network

Challenges arise in achieving high-resolution Magnetic Resonance Imaging (MRI) to improve disease diagnosis accuracy due to limitations in hardware, patient discomfort, long acquisition times, and high costs. While Convolutional Neural Networks (CNNs) have shown promising results in MRI super-resolution, they often don't look into the structural similarity and prior information available in consecutive MRI slices. By leveraging information from sequential slices, more robust features can be obtained, potentially leading to higher-quality MRI slices. We propose a multi-slice two-dimensional (2D) MRI super-resolution network that combines a Generative Adversarial Network (GAN) with feature fusion and a pre-trained slice interpolation network to achieve three-dimensional (3D) super-resolution. The proposed model requires consecutively acquired three low-resolution (LR) MRI slices along a specific axis, and achieves the reconstruction of the MRI slices in the remaining two axes. The network effectively enhances both in-plane and out-of-plane resolution along the sagittal axis while addressing computational and memory constraints in 3D super-resolution. The proposed generator has a in-plane and out-of-plane Attention (IOA) network that fuses both in-plane and out-plane features of MRI dynamically. In terms of out-of-plane attention, the network merges features by considering the similarity distance between features and for in-plane attention, the network employs a two-level pyramid structure with varying receptive fields to extract features at different scales, ensuring the inclusion of both global and local features. Subsequently, to achieve 3D MRI super-resolution, a pre-trained slice interpolation network is used that takes two consecutive super-resolved MRI slices to generate a new intermediate slice. To further enhance the network performance and perceptual quality, we introduce a feature up-sampling layer and a feature extraction block with Scaled Exponential Linear Unit (SeLU). Moreover, our super-resolution network incorporates VGG loss from a fine-tuned VGG-19 network to provide additional enhancement. Through experimental evaluations on the IXI dataset and BRATS dataset, using the peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM) and the number of training parameters, we demonstrate the superior performance of our method compared to the existing techniques. Also, the proposed model can be adapted or modified to achieve super-resolution for both 2D and 3D MRI data.

Content-aware Nakagami morphing for incremental brain MRI

Within the carcinogenesis mechanism, from the initiation of the very first tumor cell to the preneoplastic and neoplastic cancer cell groups, cancer cells omnidirectionally and unpredictably proliferate in three-dimensional (3D) space during the promotion and progression steps. Whole tumors areas, consisting of edema, necrosis, enhancing and non-enhancing tumor subareas, could easily be identified by the typical sequences of magnetic resonance imaging (MRI), mostly by FLAIR, the fluid attenuated inversion recovery sequence. Depending on the increment value of the scanner, not every scan could be contiguous though, which might create some gaps between the slices and cause some crucial loss of information. Shape-based morphing by erosion and dilation operations might be inadequate when the source or target lesions are not found on the MRI slices, which might emerge due to unpredictable shape of tumors. Furthermore, the real trajectories and genuine shapes of the tumors in spatial view might not be seen in the planar MRI slices due to approximations during the voxel-to-pixel transformations executed in MRI devices. Therefore, we propose a novel content-aware morphing procedure to generate imaginary slices by m-parametric Nakagami imaging. The mathematical and parametric design in our framework not only leads to precise segmentation of the whole tumors according to the ground truth images; but also, to realistic morphing, based on the content of tumor footprints in MRI slices. For testing purposes, a total number of 5133 morphs are created using 1831 genuine images containing at least one lesion taken from the Brats 2012 dataset. Given the average dice score coefficients (DSC), we obtained an average of 88.44% DSC with our morphing framework outperforming the conventional dilation-based morphing result which is 83.25%. Besides being a novel morphing methodology, the outcomes of this research could also be very beneficial in smoother 3D reconstruction and more realistic volumetric calculations of lesions to help the experts working on assessment of lesions.

SSA‐UNet: Whole brain segmentation by U‐Net with squeeze‐and‐excitation block and self‐attention block from the 2.5D slice image

AbstractWhole brain segmentation from magnetic resonance images (MRI) is crucial in diagnosing brain diseases and analyzing neuroimaging data. Despite advances through deep learning, challenges such as uneven gray distribution and the presence of artifacts still present hurdles in medical image processing. These limitations are often a result of insufficient spatial contextual information and lack of attention to important regions within existing models. To address these issues, this paper presents SSA‐UNet (Squeeze‐and‐Excitation and Self‐Attention UNet), a uniquely designed deep convolutional neural network that integrates spatial constraints by converting three consecutive 2D MRI slices into a single 2.5D image. This facilitates capturing inter‐slice dependencies effectively. Additionally, the newly formulated SSA block, which sequentially incorporates channel attention and Self‐Attention mechanisms, is placed before the decoders in the conventional U‐Net architecture. This enables the network to automatically weight different feature maps and focus more effectively on regions requiring precise segmentation. Rigorous evaluations on LPBA40 and IBSR18 datasets substantiate the remarkable improvements in accuracy and stability achieved by SSA‐UNet. Results indicate Dice coefficients of 98.38% and 97.47%, specificity of 99.69% and 99.57%, and sensitivity of 98.5% and 97.98% for the respective datasets. Compared to other existing models, SSA‐UNet shows significant improvements on both the LPBA40 and IBSR18 datasets. On the LPBA40 dataset, SSA‐UNet's Dice coefficient improved by 0.33% compared to the sub‐optimal model, while on the IBSR18 dataset, the improvement reached 1.78%. These empirical findings demonstrate SSA‐UNet's heightened capability in addressing the long‐standing challenges in MRI‐based whole‐brain segmentation.

Knee landmarks detection via deep learning for automatic imaging evaluation of trochlear dysplasia and patellar height

Abstract Objectives To develop and validate a deep learning–based approach to automatically measure the patellofemoral instability (PFI) indices related to patellar height and trochlear dysplasia in knee magnetic resonance imaging (MRI) scans. Methods A total of 763 knee MRI slices from 95 patients were included in the study, and 3393 anatomical landmarks were annotated for measuring sulcus angle (SA), trochlear facet asymmetry (TFA), trochlear groove depth (TGD) and lateral trochlear inclination (LTI) to assess trochlear dysplasia, and Insall-Salvati index (ISI), modified Insall-Salvati index (MISI), Caton Deschamps index (CDI) and patellotrochlear index (PTI) to assess patellar height. A U-Net based network was implemented to predict the landmarks’ locations. The successful detection rate (SDR) and the mean absolute error (MAE) evaluation metrics were used to evaluate the performance of the network. The intraclass correlation coefficient (ICC) was also used to evaluate the reliability of the proposed framework to measure the mentioned PFI indices. Results The developed models achieved good accuracy in predicting the landmarks’ locations, with a maximum value for the MAE of 1.38 ± 0.76 mm. The results show that LTI, TGD, ISI, CDI and PTI can be measured with excellent reliability (ICC > 0.9), and SA, TFA and MISI can be measured with good reliability (ICC > 0.75), with the proposed framework. Conclusions This study proposes a reliable approach with promising applicability for automatic patellar height and trochlear dysplasia assessment, assisting the radiologists in their clinical practice. Clinical relevance statement The objective knee landmarks detection on MRI images provided by artificial intelligence may improve the reproducibility and reliability of the imaging evaluation of trochlear anatomy and patellar height, assisting radiologists in their clinical practice in the patellofemoral instability assessment. Key Points • Imaging evaluation of patellofemoral instability is subjective and vulnerable to substantial intra and interobserver variability. • Patellar height and trochlear dysplasia are reliably assessed in MRI by means of artificial intelligence (AI). • The developed AI framework provides an objective evaluation of patellar height and trochlear dysplasia enhancing the clinical practice of the radiologists.

Detection of normal and abnormal brain tumor MRI images using machine learning approaches

<p>Without image processing, the modern technological world we live in today would be completely different. Different applications can be found in many different areas, including as medicine, remote sensing, computer vision, and more. Brain tumours occur because of the abnormal growth of brain tissue. Therefore, it is crucial to surgically remove the tumor’s component that has spread throughout the brain. The creation of images of the human body’s internal organs and structures depends on numerous factors, one of which is the use of Magnetic Resonance Imaging (MRI), one of the many medical imaging techniques. Segmenting images is a challenging task in the modern medical field. The tumor’s location and sizes can be detected with more accuracy with the help of a segmentation performed on MRI slices. In this paper, we propose the Adaptive Convex Region Contour (ACRC) algorithm, a novel method for achieving such requirements. Here, we use the Support Vector Machine, often known as SVM, to identify slices and determine whether they are normal or abnormal. Once the SVM results are in, the off-kilter slices will be included into the analysis. Since the human body already has a complex anatomy in 3D. It’s disappointing because MRI slices can only provide images in a plain, 2D plane. Because of its 3D structure, the tumor’s actual shape cannot be detected from a surface image. This means that the transformation from 2D to 3D is important, as it provides necessary data to the surgeons doing the operation. The Rapid Mode Image Matching (RMIM) algorithm should be used when constructing a 3D reconstruction model. Immediately after the 3D model is completed being created, a rough estimate of the tumor’s original volume is generated. The methodological consideration was conducted in the simulated MATLAB software. It was determined from the data that the proposed method is better to the current establishment methodologies in terms of accuracy. <strong>Objective:</strong> This paper aims to address the need for accurate segmentation of brain tumors in Magnetic Resonance Imaging (MRI) slices. The primary research question revolves around developing a novel segmentation method, the Adaptive Convex Region Contour (ACRC) algorithm, to improve the precision of tumor detection by incorporating a Support Vector Machine (SVM) for identifying abnormal slices.<strong> </strong><strong>Methods: </strong>The proposed approach involves utilizing the SVM to classify MRI slices as either normal or abnormal. The abnormal slices, detected by SVM, are subjected to further analysis. To overcome the limitation of MRI producing 2D images for the inherently 3D human anatomy, a transformation from 2D to 3D is crucial. The Rapid Mode Image Matching (RMIM) algorithm is adopted to construct a 3D reconstruction model of the brain and its tumor. This model allows surgeons to obtain a more accurate understanding of the tumor’s actual 3D shape. <strong>Results: </strong>The methodology was implemented and evaluated using simulated MATLAB software. The results demonstrate that the proposed ACRC algorithm, combined with SVM-based abnormal slice identification and subsequent 3D reconstruction using RMIM, outperforms existing methodologies in terms of accuracy. The 3D reconstruction provides valuable insights into the tumor’s shape, aiding surgical planning. <strong>Conclusions: </strong>In conclusion, the research contributes to the field of medical image processing by presenting a comprehensive approach for brain tumor segmentation and 3D reconstruction. The ACRC algorithm, in conjunction with SVM and RMIM, enhances the accuracy of tumor detection and provides critical 3D information for surgical procedures. This advancement holds the potential to improve the efficacy of brain tumor surgeries, underscoring the significance of innovative image processing techniques in modern medical applications.</p>

The automatic diagnosis artificial intelligence system for preoperative magnetic resonance imaging of uterine sarcoma.

Magnetic resonance imaging (MRI) is efficient for the diagnosis of preoperative uterine sarcoma; however, misdiagnoses may occur. In this study, we developed a new artificial intelligence (AI) system to overcome the limitations of requiring specialists to manually process datasets and a large amount of computer resources. The AI system comprises a tumor image filter, which extracts MRI slices containing tumors, and sarcoma evaluator, which diagnoses uterine sarcomas. We used 15 types of MRI patient sequences to train deep neural network (DNN) models used by tumor filter and sarcoma evaluator with 8 cross-validation sets. We implemented tumor filter and sarcoma evaluator using ensemble prediction technique with 9 DNN models. Ten tumor filters and sarcoma evaluator sets were developed to evaluate fluctuation accuracy. Finally, AutoDiag-AI was used to evaluate the new validation dataset, including 8 cases of sarcomas and 24 leiomyomas. Tumor image filter and sarcoma evaluator accuracies were 92.68% and 90.50%, respectively. AutoDiag-AI with the original dataset accuracy was 89.32%, with 90.47% sensitivity and 88.95% specificity, whereas AutoDiag-AI with the new validation dataset accuracy was 92.44%, with 92.25% sensitivity and 92.50% specificity. Our newly established AI system automatically extracts tumor sites from MRI images and diagnoses them as uterine sarcomas without human intervention. Its accuracy is comparable to that of a radiologist. With further validation, the system could be applied for diagnosis of other diseases. Further improvement of the system's accuracy may enable its clinical application in the future.

Femoral Head Remodeling After Surgical Reduction of Developmental Hip Dislocations.

Developmental dysplasia of the hip is prevalent and is associated with dysplasia of both the femoral head and acetabulum. There is a paucity of literature describing femoral head remodeling after surgical reduction of developmentally dislocated hips. The purpose of this study was to describe and quantify changes in femoral head sphericity after closed or open reduction for developmental dysplasia of the hip. A retrospective cohort study was performed including patients with typical developmental hip dislocations who underwent closed or open reduction from 2009 to 2022 at a single institution and had immediate postoperative and average 3-week follow-up magnetic resonance imaging (MRI) scans. A subset of patients also had 3-year follow-up MRI scans. Patients with insufficient imaging or bony procedures at the time of reduction were excluded. We developed a technique to quantify femoral head "sphericity" by comparing differences in measured radii of the femoral head on axial and coronal plane MRI slices. We then calculated the variance of the radii for each plane and averaged these to calculate a combined variance. The variance was used to represent "sphericity," with a larger variance indicating a wider distribution of radii and thus a less spherical shape. A total of 74 patients (69 females) with 96 hips were included in this series. The median age of the child at the time of reduction was 8.7 months [interquartile range (IQR): 2.2]. Over half (58.3%) of the hips had a closed reduction, whereas the remaining were open reduced (41.7%). Immediately postoperatively, at the 3-week time point, and at the 3-year time point the median combined variance was 1.1 (IQR: 3.93), 0.51 (IQR: 1.32), and 0.31 (IQR: 0.50), respectively, indicating improved sphericity over time. Sphericity of the femoral head in developmental hip dislocations improves in both the immediate postoperative period, as well as the first few years after reduction. Further research is needed to evaluate the mechanism of remodeling, the ideal timing of reduction, and the relationship between femoral head and acetabular remodeling. Level IV-case series, therapeutic study.

Rectus Femoris Ultrasound Echo Intensity Is a Valid Estimate of Percent Intramuscular Fat in Patients Following Anterior Cruciate Ligament Reconstruction.

The aim of the work described here was to evaluate quadriceps muscle ultrasound metrics and common demographic variables to create a conversion equation that validly predicts magnetic resonance imaging (MRI) percent intramuscular fat after anterior cruciate ligament reconstruction (ACLR). We recruited 15 participants between the ages of 18 and 35 y who were 1-5 y post-ACLR. For the MRI assessment, we used an iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL) sequence to assess the mid-thigh. A single reader manually segmented the rectus femoris on two consecutive MRI slices using ITK-Snap to estimate the percent intramuscular fat. For the ultrasound assessment, a single investigator captured transverse panoramic ultrasound images of the mid-thigh with the participant positioned supine and the knee flexed to 30°. A separate single reader used ImageJ to manually segment the rectus femoris ultrasound images. Ultrasound metrics included muscle cross-sectional area, echo intensity and subcutaneous fat thickness. A stepwise linear multiple regression was used to develop an equation to predict MRI percent intramuscular fat using the ultrasound metrics and common demographics (i.e., age, sex, height, mass). Additionally, intraclass correlation coefficients (ICC2,k) and Bland-Altman plots were used to assess the agreement between true and estimated percent intramuscular fat. Echo intensity and age significantly predicted MRI intramuscular fat percent (p=0.003, r2=0.62). When using the conversion equation, there was high agreement (ICC2,k=0.87, 95% confidence interval: 0.62-0.96) between the estimated and true percent intramuscular fat. Our patient population-specific conversion equation that uses quadriceps muscle ultrasound echo intensity and age is a valid estimate of MRI percent intramuscular fat.

Volumetric MRI with sparse sampling for MR-guided 3D motion tracking via sparse prior-augmented implicit neural representation learning.

Volumetric reconstruction of magnetic resonance imaging (MRI) from sparse samples is desirable for 3D motion tracking and promises to improve magnetic resonance (MR)-guided radiation treatment precision. Data-driven sparse MRI reconstruction, however, requires large-scale training datasets for prior learning, which is time-consuming and challenging to acquire in clinical settings. To investigate volumetric reconstruction of MRI from sparse samples of two orthogonal slices aided by sparse priors of two static 3D MRI through implicit neural representation (NeRP) learning, in support of 3D motion tracking during MR-guided radiotherapy. A multi-layer perceptron network was trained to parameterize the NeRP model of a patient-specific MRI dataset, where the network takes 4D data coordinates of voxel locations and motion states as inputs and outputs corresponding voxel intensities. By first training the network to learn the NeRP of two static 3D MRI with different breathing motion states, prior information of patient breathing motion was embedded into network weights through optimization. The prior information was then augmented from two motion states to 31 motion states by querying the optimized network at interpolated and extrapolated motion state coordinates. Starting from the prior-augmented NeRP model as an initialization point, we further trained the network to fit sparse samples of two orthogonal MRI slices and the final volumetric reconstruction was obtained by querying the trained network at 3D spatial locations. We evaluated the proposed method using 5-min volumetric MRI time series with 340ms temporal resolution for seven abdominal patients with hepatocellular carcinoma, acquired using golden-angle radial MRI sequence and reconstructed through retrospective sorting. Two volumetric MRI with inhale and exhale states respectively were selected from the first 30s of the time series for prior embedding and augmentation. The remaining 4.5-min time series was used for volumetric reconstruction evaluation, where we retrospectively subsampled each MRI to two orthogonal slices and compared model-reconstructed images to ground truth images in terms of image quality and the capability of supporting 3D target motion tracking. Across the seven patients evaluated, the peak signal-to-noise-ratio between model-reconstructed and ground truth MR images was 38.02±2.60dB and the structure similarity index measure was 0.98±0.01. Throughout the 4.5-min time period, gross tumor volume (GTV) motion estimated by deforming a reference state MRI to model-reconstructed and ground truth MRI showed good consistency. The 95-percentile Hausdorff distance between GTV contours was 2.41±0.77mm, which is less than the voxel dimension. The mean GTV centroid position difference between ground truth and model estimation was less than 1mm in all three orthogonal directions. A prior-augmented NeRP model has been developed to reconstruct volumetric MRI from sparse samples of orthogonal cine slices. Only one exhale and one inhale 3D MRI were needed to train the model to learn prior information of patient breathing motion for sparse image reconstruction. The proposed model has the potential of supporting 3D motion tracking during MR-guided radiotherapy for improved treatment precision and promises a major simplification of the workflow by eliminating the need for large-scale training datasets.

Artificial intelligence for the detection of sacroiliitis on magnetic resonance imaging in patients with axial spondyloarthritis.

Magnetic resonance imaging (MRI) is important for the early detection of axial spondyloarthritis (axSpA). We developed an artificial intelligence (AI) model for detecting sacroiliitis in patients with axSpA using MRI. This study included MRI examinations of patients who underwent semi-coronal MRI scans of the sacroiliac joints owing to chronic back pain with short tau inversion recovery (STIR) sequences between January 2010 and December 2021. Sacroiliitis was defined as a positive MRI finding according to the ASAS classification criteria for axSpA. We developed a two-stage framework. First, the Faster R-CNN network extracted regions of interest (ROIs) to localize the sacroiliac joints. Maximum intensity projection (MIP) of three consecutive slices was used to mimic the reading of two adjacent slices. Second, the VGG-19 network determined the presence of sacroiliitis in localized ROIs. We augmented the positive dataset six-fold. The sacroiliitis classification performance was measured using the sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). The prediction models were evaluated using three-round three-fold cross-validation. A total of 296 participants with 4,746 MRI slices were included in the study. Sacroiliitis was identified in 864 MRI slices of 119 participants. The mean sensitivity, specificity, and AUROC for the detection of sacroiliitis were 0.725 (95% CI, 0.705-0.745), 0.936 (95% CI, 0.924-0.947), and 0.830 (95%CI, 0.792-0.868), respectively, at the image level and 0.947 (95% CI, 0.912-0.982), 0.691 (95% CI, 0.603-0.779), and 0.816 (95% CI, 0.776-0.856), respectively, at the patient level. In the original model, without using MIP and dataset augmentation, the mean sensitivity, specificity, and AUROC were 0.517 (95% CI, 0.493-0.780), 0.944 (95% CI, 0.933-0.955), and 0.731 (95% CI, 0.681-0.780), respectively, at the image level and 0.806 (95% CI, 0.729-0.883), 0.617 (95% CI, 0.523-0.711), and 0.711 (95% CI, 0.660-0.763), respectively, at the patient level. The performance was improved by MIP techniques and data augmentation. An AI model was developed for the detection of sacroiliitis using MRI, compatible with the ASAS criteria for axSpA, with the potential to aid MRI application in a wider clinical setting.

Placenta segmentation in magnetic resonance imaging: Addressing position and shape of uncertainty and blurred placenta boundary

Automatic segmentation of the placenta in MRI facilitates quantitative measurements and further analysis. Automatic segmentation of the placenta in magnetic resonance imaging (MRI) is challenging due to the fact that (1) the position of the placenta varies greatly in different views of MRI. In addition, the shape of the placenta changes layer by layer in different MRI slices. The position and shape of the placenta are highly uncertain. (2) PAS results in a blurred placental boundary. To address the above issues, a refinement fusion based on the U-Net (RFU-Net) is proposed for the segmentation of the placenta with variable position and shape in MR images, and the boundary de-precision explores the distribution of blood vessels around the placental boundary. In RFU-Net, ResNet34 is used as a feature extractor, and a fusion multiscale feature (FMF) is proposed to extract multiscale features and their contextual information to solve the problem of variable placental shape. A refinement segmentation module (RSM) is designed to provide information about the position of the placenta to solve the problem of variable placental position, which effectively improves the segmentation of the placenta. The mean intersection over union (MIoU), hausdorff distance (HD), dice coefficient (Dice) and accuracy (Acc) are 0.8620, 4.3142, 0.9314 and 0.9987, respectively. Extensive experiments show that RFU-Net effectively improves the segmentation accuracy of the placenta. Boundary de-precision resolves the blurring of placental boundaries, effectively highlighting areas of the placental boundary that may contain blood vessels to aid in the physician's diagnosis.

K-strip: A novel segmentation algorithm in k-space for the application of skull stripping

Background and objectiveWe present a novel deep learning-based skull stripping algorithm for magnetic resonance imaging (MRI) that works directly in the information rich complex valued k-space. MethodsUsing four datasets from different institutions with a total of around 200,000 MRI slices, we show that our network can perform skull-stripping on the raw data of MRIs while preserving the phase information which no other skull stripping algorithm is able to work with. For two of the datasets, skull stripping performed by HD-BET (Brain Extraction Tool) in the image domain is used as the ground truth, whereas the third and fourth dataset comes with per-hand annotated brain segmentations. ResultsAll four datasets were very similar to the ground truth (DICE scores of 92 %–99 % and Hausdorff distances of under 5.5 pixel). Results on slices above the eye-region reach DICE scores of up to 99 %, whereas the accuracy drops in regions around the eyes and below, with partially blurred output. The output of k-Strip often has smoothed edges at the demarcation to the skull. Binary masks are created with an appropriate threshold. ConclusionWith this proof-of-concept study, we were able to show the feasibility of working in the k-space frequency domain, preserving phase information, with consistent results. Besides preserving valuable information for further diagnostics, this approach makes an immediate anonymization of patient data possible, already before being transformed into the image domain. Future research should be dedicated to discovering additional ways the k-space can be used for innovative image analysis and further workflows.

Femoral cartilage ultrasound echo-intensity is a valid measure of cartilage composition.

This study aimed to create a conversion equation that accurately predicts cartilage magnetic resonance imaging (MRI) T2 relaxation times using ultrasound echo-intensity and common participant demographics. We recruited 15 participants with a primary anterior cruciate ligament reconstruction between the ages of 18 and 35 years at 1-5 years after surgery. A single investigator completed a transverse suprapatellar scan with the ACLR limb in max knee flexion to image the femoral trochlea cartilage. A single reader manually segmented the femoral cartilage cross-sectional area to assess the echo-intensity (i.e., mean gray-scale pixel value). At a separate visit, a T2 mapping sequence with the MRI beam set to an oblique angle was used to image the femoral trochlea cartilage. A single reader manually segmented the cartilage cross-sectional area on a single MRI slice to assess the T2 relaxation time. A stepwise, multiple linear regression was used to predict T2 relaxation time from cartilage echo-intensity and common demographic variables. We created a conversion equation using the regression betas and then used an ICC and Bland-Altman plot to assess agreement between the estimated and true T2 relaxation time. Cartilage ultrasound echo-intensity and age significantly predicted T2 relaxation time (F = 7.33, p = 0.008, R2 = 0.55). When using the new conversion equation to estimate T2 relaxation time from cartilage echo-intensity and age, there was strong agreement between the estimated and true T2 relaxation time (ICC2,k = 0.84). This study provides promising preliminary data that cartilage echo-intensity combined with age can be used as a clinically accessible tool for evaluating cartilage composition.

Application of Sequential Thresholding-Based Automated Reconstruction of the Trigeminal Nerve in Trigeminal Neuralgia

High-resolution magnetic resonance imaging (MRI) of the trigeminal nerve is indispensable for workup of trigeminal neuralgia (TN) before microvascular decompression; however, the evaluation is often subjective and prone to variability. We aim to develop and assess sequential thresholding-based automated reconstruction ofthe trigeminal nerve (STAR-TN) as an algorithm for segmenting the trigeminal nerve and contacting structures that will allow for a structured method for assessing neurovascular conflict. A total of 42 patients with TN who underwent high-resolution MRI before microvascular decompression in 2022 were included in our study. Segmentation of the trigeminal nerve and contacting structures was performed on preoperative MRI scans using STAR-TN. The segmentations were then evaluated for neurovascular conflict and compared to the preoperative radiology and operative notes. Geometric features, including the area of contact and distance to conflict, were extracted. Of the 42 patients, 32 (76.2%) were found to show neurovascular conflict based solely on their STAR-TN segmentations and 10 (23.8%) were found to not show neurovascular conflict. Compared with the intraoperative findings, this resulted in a sensitivity of 78.0% and specificity of 100%. In contrast, assessments of neurovascular conflict by radiologists using only 2-dimensional MRI views had a sensitivity of 68.3% and specificity of 100%. Of the 32 patients with neurovascular conflict, 29 (90.9%) had conflict within the root entry zone. Overall, the patients had a median area of contact of 10.66 mm2. STAR-TN allows for 3-dimensional visualization and identification of neurovascular conflict with improved sensitivity compared with neuroradiologist assessments from MRI slices.

Time-series visual explainability for Alzheimer’s disease progression detection for smart healthcare

Artificial intelligence (AI)-based diagnostic systems provide less error-prone and safer support to clinicians, enhancing the medical decision-making process. This study presents a smart and reliable healthcare framework for detecting Alzheimer's disease (AD) progression. Early detection of AD before the onset of clinical symptoms is the most crucial step in starting timely treatment. To predict the conversion of cognitively normal patients to those with AD, three-dimensional (3D) magnetic resonance imaging (MRI) whole-brain neuroimaging methods have been extensively studied. However, depending on the 3D volume, this method is computationally expensive. To solve this problem, we used an approximate rank pooling method originally designed for video action recognition with a 3D MRI volume to obtain a compressed representation of multiple two-dimensional (2D) MRI slices. This study proposes a hybrid multimodal CNN-Bi-LSTM deep model for AD progression detection, in which the resulting dynamic 2D images are fused with cognitive features. Moreover, a novel explainable AI approach is proposed to provide visual explanations using the resulting longitudinal 2D dynamic images. Temporal explanations were provided by visualizing the affected brain regions captured using longitudinal 2D MRIs. By utilizing a sample of 1,692 subjects with multimodal data from the Alzheimer’s Disease Neuroimaging Initiative dataset, our method was assessed using a 10-fold cross-validation process. The model achieved an area under the receiver operating characteristics curve (AUC) of 94% using longitudinal 2D three-time-step dynamic image data. The fusion of 2D dynamic images with cognitive features enhanced the performance by 2% in terms of the AUC. Patients who gradually develop AD, show changes in various brain regions. For such patients, our system highlights the critical role of the hippocampus, medial amygdala, caudal hippocampus, and lateral amygdala at the initial time steps. In the late stages of AD, the system detects abnormalities in extra brain regions such as the medial temporal gyrus, superior temporal gyrus, fusiform gyrus, and caudal hippocampus; indicating that patients have completely progressed to AD.

Prostate Age Gap: An MRI Surrogate Marker of Aging for Prostate Cancer Detection.

Aging is the most important risk factor for prostate cancer (PC). Imaging techniques can be useful to measure age-related changes associated with the transition to diverse pathological states. However, biomarkers of aging from prostate magnetic resonance imaging (MRI) remain to be explored. To develop an aging biomarker from prostate MRI and to examine its relationship with clinically significant PC (csPC, Gleason score ≥7) risk occurrence. Retrospective. Four hundred and sixty-eight (65.97 ± 6.91 years) biopsied males, contributing 7243 prostate MRI slices. A deep learning (DL) model was trained on 3223 MRI slices from 81 low-grade PC (Gleason score ≤6) and 131 negative patients, defined as non-csPC. The model was tested on 90 negative, 52 low-grade (142 non-csPC), and 114 csPC patients. 3-T, axial T2-weighted spin sequence. Chronological age was defined as the age of the participant at the time of the visit. Prostate-specific antigen (PSA), prostate volume, Gleason, and Prostate Imaging-Reporting and Data System (PI-RADS) scores were also obtained. Manually annotated prostate masks were used to crop the MRI slices, and a DL model was trained with those from non-csPC patients to estimate the age of the patients. Following, we obtained the prostate age gap (PAG) on previously unseen csPC and non-csPC cropped MRI exams. PAG was defined as the estimated model age minus the patient's age. Finally, the relationship between PAG and csPC risk occurrence was assessed through an adjusted multivariate logistic regression by PSA levels, age, prostate volume, and PI-RADS ≥ 3 score. T-test, Mann-Whitney U test, permutation test, receiver operating characteristics (ROC), area under the curve (AUC), and odds ratio (OR). A P value <0.05 was considered statistically significant. After adjusting, there was a significant difference in the odds of csPC (OR = 3.78, 95% confidence interval [CI]: 2.32-6.16). Further, PAG showed a significantly larger bootstrapped AUC to discriminate between csPC and non-csPC than that of adjusted PI-RADS ≥ 3 (AUC = 0.981, 95% CI: 0.975-0.987). PAG may be associated with the risk of csPC and could outperform other PC risk factors. 3 TECHNICAL EFFICACY: Stage 3.

Fractal-Based Morphometrics of Glioblastoma.

Morphometrics have been able to distinguish important features of glioblastoma from magnetic resonance imaging (MRI). Using morphometrics computed on segmentations of various imaging abnormalities, we show that the average and range of lacunarity and fractal dimension values across MRI slices can be prognostic for survival. We look at the repeatability of these metrics to multiple segmentations and how they are impacted by image resolution. We speak to the challenges to overcome before these metrics are included in clinical care, and the insight that they may provide.

Prediction of Progression After Cervix Cancer Radiotherapy Using a Machine-Learning Model on Pre-Treatment MRI

MRI may be useful to identify women with cervical cancer at high risk of disease progression to test strategies of treatment intensification. The purpose of this study was to determine the value of a machine-learning model built on pre-treatment MRI for prediction of risk of progression after radiation therapy. MagneticResonance Imaging (MRI) data for women with cervical cancer was collected from The Cancer Genome Atlas Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma Collection (TCGA-CESC) on the Cancer Imaging Archive (TCIA), which reported clinical, treatment, and imaging data from a single institution. 27 patients who had received radiation for cervical cancer were selected for input into a custom 3-D Residual Neural Network (ResNet) model with added custom layers specific to DICOM data in tensorflow python package. One T2 MRI per patient was used to predict recurrence free survival after radiation treatment, where patients were predicted to be "high risk" or "low risk" for disease recurrence as the output of the model. All slices of the T2 MRI were used. The model was validated using five-fold cross validation; 80% of the data was used to train each fold and 20% was used for testing. Final model statistical significance was confirmed through shuffle test at the p < 0.01 level. The clinical outcomes of patients and the model's "low-risk" and "high-risk" prediction were compared. There were 27 patients in the study with mean age of 51 years (range 29-79). 20 patients had squamous cell carcinoma and 7 patients had adenocarcinoma. The stage breakdown consisted of 9 women IB, 2 IIA, 9 IIB, 2 IIIA, 2 IIIB, and 3 stage IV. 10 women were treated with radiation alone and 17 with chemo-radiation. 5 women received surgery in addition to radiation or chemoradiation. 21 patients received brachytherapy. Median follow-up of patients was 29 months (range 3-64). The model predicted 7 patients as "high risk" for recurrence; all 7 developed a recurrence during follow up. None of the 20 patients predicted to be "low risk" developed disease recurrence. Among all patients in the study, the two-year progression free survival (PFS) was 82.0%. Patients identified as "low risk" and "high risk" by model had two-year PFS of 100% and 43%, respectively. Among patients with recurrence, 3 developed local recurrence and 4 developed distant metastases. The ResNet model achieved cross-validated accuracy of 92% for prediction of progression-free survival (p<0.01). A 3-D ResNet machine-learning model using pretreatment MRI image data can accurately predict clinical outcomes for cervical cancer following radiation therapy. Future work to confirm generalizability should focus on validation with a larger clinical dataset.

Small Samples and Low-Cost Auto-Segmentation Method for Pelvic Organ-at-Risk Segmentation in Magnetic Resonance Images Using Deep-Learning

In radiotherapy, magnetic resonance (MR) imaging has higher contrast of soft tissue, and no radiation compared with computed tomography (CT) scanning. Due to the high-cost of manual annotation, the deep-learning based automatic organ-at-risk (OAR) and target delineation algorithms are in high-demand, but the collecting of large amounts of high-quality annotated datasets remains difficulty. In this paper, we proposed a low-cost OAR segmentation method with semi-supervised annotation using small annotation samples of pelvic MR images. This study consisted of 94 patients diagnosed with rectal cancer from April 2018 to March 2021 at Peking University People's Hospital. We used 17 slices of MR images with annotation and 78 slices without annotation to train a deep-learning based segmentation model. The bladder, femoral heads, rectum and small intestine were selected as OAR. Semi-supervised method and ensemble learning were used for generating training set using small sample with annotation. Post-processing algorithm was used to correct the self-annotation data. Two of 14 annotation samples were set as test set. As for un-labeled images, 40 of them were set as semi-supervised annotation train set, the rest were test set. Besides, both 2D and 3D auto-segmentation networks were evaluated. The dice of bladder, femoral head left and right, rectum and small intestine between segmentation results and reference masks is 0.947, 0.983, 0.981, 0.900, 0.845 only using self-annotation and post-processing method of 2D segmentation model. And the dice of corresponding OAR is 0.871, 0.975, 0.975, 0.783, 0.724 using 3D segmentation network, 0.885,0.982, 0.982, 0.882, 0,814 using 2D segmentation network with supervised method (nnUNet). The 2D model outperformed 3D model with better segmentation performance, shorter inference time and fewer parameters. The results proved that we can train a multi-OAR segmentation model only using small annotation samples and other unlabeled samples. Ensemble learning and post-processing methods are necessary for semi-supervised data annotation. For anisotropy data, 2D model shows better performance than 3D models.

Automatic segmentation of brain glioma based on XY-Net.

Glioma is a malignant primary brain tumor, which can easily lead to death if it is not detected in time. Magnetic resonance imaging is the most commonly used technique to diagnose gliomas, and precise outlining of tumor areas from magnetic resonance images (MRIs) is an important aid to physicians in understanding the patient's condition and formulating treatment plans. However, relying on radiologists to manually depict tumors is a tedious and laborious task, so it is clinically important to investigate an automated method for outlining glioma regions in MRIs. To liberate radiologists from the heavy task of outlining tumors, we propose a fully convolutional network, XY-Net, based on the most popular U-Net symmetric encoder-decoder structure to perform automatic segmentation of gliomas. We construct two symmetric sub-encoders for XY-Net and build interconnected X-shaped feature map transmission paths between the sub-encoders, while maintaining the feature map concatenation between each sub-encoder and the decoder. Moreover, a loss function composed of the balanced cross-entropy loss function and the dice loss function is used in the training task of XY-Net to solve the class unevenness problem of the medical image segmentation task. The experimental results show that the proposed XY-Net has a 2.16% improvement in dice coefficient (DC) compared to the network model with a single encoder structure, and compare with some state-of-the-art image segmentation methods, XY-Net achieves the best performance. The DC, HD, recall, and precision of our method on the test set are 74.49%, 10.89mm, 78.06%, and 76.30%, respectively. The combination of sub-encoders and cross-transmission paths enables the model to perform better; based on this combination, the XY-Net achieves an end-to-end automatic segmentation of gliomas on 2D slices of MRIs, which can play a certain auxiliary role for doctors in grasping the state of illness.