Axial MRI Research Articles (Page 1)

Introduction: Quantifying adiposity, a key biomarker of metabolic health, typically requires imaging that involves radiation, high costs, and manual effort. We developed an AI framework to segment external white adipose tissue (EWAT) from routine non-contrast MRI, offering a radiation-free, low-effort alternative. Hypothesis: We hypothesized that combining classical image processing with deep learning would enable accurate, robust EWAT segmentation from routine T1/T2-weighted MRI, without specialized sequences or manual labeling. Methods: In 105 Type 1 diabetes patients, T1/T2-weighted axial abdominal MRI scans at the aortic bifurcation were used to develop three segmentation approaches: Region Growing with automatic seed selection, iterative pixel aggregation and adaptive thresholds; UNet CNN trained on 52 masks from region-growing results, with Dice&Binary Cross-Entropy loss; and, Fine-Tuned UNet, optimized on 48 complex cases using extensive augmentations (flips, crops, brightness shifts, Gaussian noise) to enhance robustness and generalizability. Key challenges like artifacts, low fat volume, and anatomical overlap were addressed via local adjustments and hyperparameter tuning. Three independent clinicians scored segmentation quality (0–3) for anatomical alignment (Accuracy), circumference capture (Completeness), target area segmentation (Coverage), and boundary continuity (Smoothness). Results: Table 1 summarizes the mean clinical evaluation scores across all patients and metrics. The UNet and Fine-Tuned UNet consistently outperformed Region Growing in all four metrics, with mean accuracy of 2.81 and 2.80, respectively, versus 2.16 for Region Growing. Figure 1 shows example segmentations for each method. Figures 2 and 3 visualize model performance, highlighting mean scores in complex cases and the percentage of perfect (3/3) segmentations. The Fine-Tuned UNet had the highest mean accuracy (2.80) in challenging images, while the base UNet had the most perfect scores overall (73.3%). Conclusion: This unsupervised AI framework enables accurate, radiation-free EWAT segmentation from routine MRI. All methods, including deep learning, were trained without manual labeling, using region-growing outputs as pseudo ground truth. Clinical evaluations confirmed that the UNets achieved superior accuracy, completeness, coverage, and smoothness, particularly in complex cases. This scalable, cost-effective approach supports broader validation in cardiometabolic populations.

Goutallier grading of psoas major and lumbar extensor muscles as a predictor of cage subsidence and reoperation following transforaminal and posterior lumbar interbody fusion.

Is There a Correlation between the position of conus medullaris and the clinical presentation and surgical outcomes in primary tethered cord Syndrome?

BackgroundThoracic spinal stenosis (TSS) is a rare yet debilitating condition, often requiring surgical decompression. Prognostic assessments traditionally rely on single clinical or imaging features, limiting prediction accuracy. This study explores whether radiomics-based models enhance outcome prediction in TSS.MethodsWe retrospectively enrolled 106 surgically treated TSS patients (2012–2022), collecting clinical data and T2 axial MRI scans. Radiomics features were extracted from the most stenotic level, followed by rigorous feature selection (ICC > 0.9, U-test, Spearman, mRMR, and LASSO). Six machine learning classifiers were trained using radiomics and/or clinical data. Model performance was evaluated using AUC on an independent test set.ResultsRadiomics models outperformed clinical models (SVM AUC: 0.824 vs. 0.731). The combined radiomics–clinical model achieved the highest test-set AUC of 0.867, offering improved sensitivity and specificity.ConclusionIn this preliminary exploratory study, integrating MRI radiomics with clinical data appeared to improve prediction of neurological recovery in TSS. These findings suggest that radiomics may enable objective, high-dimensional assessment of spinal cord pathology and potentially support individualized surgical decision-making, although further validation in larger, multicenter prospective cohorts is required.

Imaging parameter study. This study aims to evaluate iliac crest height and slope on plain radiographs and assess their correlation with the positions of the safe working zone (SWZ), lumbar plexus, and iliac great vessels on MRIs. Despite the advantages of lateral lumbar interbody fusion (LLIF), complications related to patient anatomy and the surgical approach can occur. Variations in iliac crest morphology may affect the location of neurovascular structures. We reviewed lumbar plain films and MRIs of 98 patients. Iliac crest height was classified into high and low iliac crest height (HICH and LICH), and iliac crest slope was classified into high and low iliac crest slope (HICS and LICS). The SWZ, lumbar plexus position, and L4-5 oblique corridor were assessed from the MRIs. The LICH group had statistically significant wider SWZ on the left side compared with HICH (P=0.007). For the lumbar plexus, the HICH group had significantly more patients in Moro zone II (P=0.024) and significantly fewer patients in Moro zone IV (P=0.002). For the iliac slope showed similar results, there were statistically significant wider SWZ (P=0.003) and more posterior left-sided lumbar plexus position in the HICS group (P=0.002). The mean L4-5 oblique corridor showed no significant difference between the high and low iliac crest height or slope groups. The LICH and HICS had a correlation with more widening SWZ and more posterior lumbar plexus position on axial MRI. These iliac crest types are suitable to perform LLIF of L4-5.

Astrocytoma is the most common type of brain glioma and is classified by the World Health Organization into four grades, providing prognostic insights and guiding treatment decisions. The accurate determination of astrocytoma grade is critical for patient management, especially in high-malignancy-grade cases. This study proposes a fully optimized Convolutional Neural Network (CNN) for the classification of astrocytoma MRI slices across the three malignant grades (G2–4). The training dataset consisted of 1284 pre-operative axial 2D MRI slices from T1-weighted contrast-enhanced and FLAIR sequences derived from 69 patients. To provide the best possible model performance, an extensive hyperparameter tuning was carried out through the Hyperband method, a variant of Successive Halving. Training was conducted using Repeated Hold-Out Validation across four randomized data splits, achieving a mean classification accuracy of 98.05%, low loss values, and an AUC of 0.997. Comparative evaluation against state-of-the-art pre-trained models using transfer learning demonstrated superior performance. For validation purposes, the proposed CNN trained on an altered version of the training set yielded 93.34% accuracy on unmodified slices, which confirms the model’s robustness and potential use for clinical deployment. Model interpretability was ensured through the application of two Explainable AI (XAI) techniques, SHAP and LIME, which highlighted the regions of the slices contributing to the decision-making process.

BackgroundSignal-to-noise ratio (SNR) is a key metric for evaluating MRI image quality, but conventional measurement methods are time-consuming and operator-dependent. Deep learning offers potential for automating this process.PurposeTo develop and validate a deep learning-based method for automatic SNR measurement from single MRI images.Material and methodsA Pix2Pix framework with a U-Net++ generator and GAN-based discriminator was trained using axial brain MRI images (T1WI, T2WI, and FLAIR) from a 3T scanner. The model generated signal and noise maps from a single image, and SNR maps were computed by pixel-wise division. Whole-brain, white matter (WM), and cerebrospinal fluid (CSF) regions were automatically segmented for regional SNR measurement. The subtraction-map method served as the reference. Structural similarity index (SSIM), correlation coefficients, and Bland–Altman analyses were used to evaluate agreement.ResultsAcross all sequences, the mean SSIM was 0.95 ± 0.02. SNR values showed strong correlations with the reference method (r > 0.86) and low relative errors (<7%) for whole-brain, WM, and CSF. Bland–Altman analysis demonstrated a small paired bias and narrow 95% limits of agreement across sequences.ConclusionThe proposed deep learning method enables automatic, accurate, and observer-independent SNR quantification from single MR images, supporting clinical and research image quality evaluation.

A decade-long trends in ligamentum flavum hypertrophy among spinal stenosis patients: A comparative analysis of incidence and patterns.

Glenoid bony anatomy in patients with epilepsy is influenced by their tonic-clonic seizure burden.

Accurate segmentation in short-axis cardiac MRI is crucial for extracting pa-rameters to diagnose cardiovascular diseases. However, deep learning models often use Max or Average pooling to reduce feature map size, risking loss of important information. In addition, most current research focuses mainly on the segmentation accuracy of machine learning models, while lacking inter-pretation of the segmentation results- a critical factor for clinical applica-tions. In this study, we propose a fuzzy deep learning model called 3DMFL-Net, which combines Explainable AI (XAI) techniques to segment cardiac structures including the Left Ventricle, Myocardium, and Right Ventricle from 3D short axis MRI images, while also providing interpretation of the segmentation results. The model adopts a 3D fuzzy pooling method to re-place traditional pooling methods. The 3D fuzzy pooling leverages fuzzy logic along with a Gaussian membership function to identify important features within the pooling window, allowing reduction of the feature map size while preserving essential information. To visualize and interpret the model’s deci-sions, we apply XAI with the 3D Grad-CAM method to generate heatmaps that highlight the important image regions involved in the segmentation process of the proposed model. The proposed model is evaluated on the M&amp;Ms-2020 dataset, which includes multiple 3D cardiac MRI scans across various disease groups and is provided by different device vendors. The re-sults show that the Dice coefficient reaches 84.6% for the left ventricle, 76.6% for the myocardium, and 79.5% for the right ventricle. The proposed method suits intelligent healthcare deployment and helps clinicians trust segmentation outcomes.

The goal of this study was to investigate the patterns of functional connectivity (FC) in patients with asymptomatic cervical spinal cord compression and determine how the patterns differ from those in healthy controls and correlate with spinal compression and Neck Disability Index (NDI) scores. This cross-sectional study consisted of 45 patients with asymptomatic spinal cord compression (ASCC) and 35 healthy controls (HCs) with resting-state functional MRI (rs-fMRI) scans. The patients with ASCC also had sagittal and axial T2-weighted cervical spine MRI scans. The rs-fMRI scans were used for region of interest to region of interest analyses that generated brain networks of FC that could be compared between and within groups. The patients with ASCC had stronger FC between visual and motor regions than the HCs, with the intracalcarine cortex (occipital cortex) as the largest hub of connection strength differences. Within the ASCC cohort, the cerebellar region associated with attention (multi-domain task battery [MDTB] region 5) was the hub of functional changes related to the severity of spinal compression. However, the NDI scores of patients covaried most with functional connections of the left superior parietal lobule. This study indicated that functional brain changes are evident before neurological symptoms appear. These alterations in FC patterns reflect a systematic reorganization of neural dynamics, suggesting that the brain adaptively reconfigures its computational architecture to compensate for compromised signal transmission through the compressed spinal cord. Patients with ASCC appear to rely more on visual information to maintain normal sensorimotor function, as proprioception information is likely compromised due to spinal compression. Their functional changes in the subregion of the cerebellum involved in attention indicate possible strain on multitasking and working memory. Finally, connectivity differences related to NDI scores support the idea that the superior parietal lobule helps to compensate for motor difficulties. These early adaptations in brain computation could serve as crucial biomarkers for disease progression, potentially enabling more precise timing of clinical interventions in this challenging patient population.

Background/Objectives: Adolescent idiopathic scoliosis (AIS) can lead to significant chest deformations. The quantification of chest deformity and spinal length could provide additional insights for monitoring during follow-up and treatment. This study proposes a 3D U-Net convolutional neural network (CNN) for automatic thoracic and spinal segmentations of chest MRI scans. Methods: In this proof-of-concept study, axial chest MRI scans from 19 girls aged 8–10 years at risk for AIS development and 19 asymptomatic young adults were acquired (n = 38). The thoracic volume and spine were manually segmented as the ground truth (GT). A 3D U-Net CNN was trained on 31 MRI scans. The seven remaining MRI scans were used for validation, reported by the Dice similarity coefficient (DSC), the Hausdorff distance (HD), precision, and recall. From these segmentations, the thoracic volume and 3D spinal length were calculated. Results: Automatic chest segmentation was possible for all chest MRIs. For the chest volume segmentations, the average DSC was 0.91, HD was 51.89, precision was 0.90, and recall 0.99. For the spinal segmentation, the average DSC was 0.85, HD was 25.98, precision was 0.74, and recall 0.99. Chest volumes and 3D spinal lengths differed by on average 11% and 12% between automatic and GT, respectively. Qualitative analysis showed agreement between the automatic and manual segmentations in most cases. Conclusions: The proposed 3D U-Net CNN shows a high accuracy and good predictions in terms of HD, DSC, precision, and recall. This suggested 3D U-Net CNN could potentially be used to monitor the progression of chest deformation in scoliosis patients in a radiation-free manner. Improvement can be made by training the 3D U-net with more data and improving the GT data.

OBJECTIVES:To use 3D imaging modalities to obtain precise measurements of the proximal tibia in pediatric patients and assess the safety of current intraosseous needle lengths (15 and 25 mm).DESIGN:Retrospective descriptive study.SETTING:University of Minnesota and MHealth Fairview System, Minneapolis, MN.PATIENTS:Pediatric patients (≤ 16 yr) who underwent full-body positron emission tomography-CT or axial MRI scans of the lower extremities between January 2014 and December 2023.INTERVENTIONS:None.MEASUREMENTS AND MAIN RESULTS:A total of 912 scans were initially retrieved; 232 scans were excluded due to osseous diseases, tibial fractures, suboptimal scan quality, or soft-tissue abnormalities, leaving 680 scans for analysis. Scans were stratified into 1-year age groups. Measurements at the proximal tibia included soft-tissue thickness, cortical bone thickness, and medullary canal diameter. Other values, such as the pre-intraosseous space (sum of cortical thickness and soft-tissue depth) and total distance to deep cortex, were calculated. Simulated needle insertions demonstrated that 31.62% of the 15 mm needles were too shallow, failing to reach the medullary canal, whereas 34.85% of the 25 mm needles were too deep, both of which could cause severe complications. A cutoff analysis for needle size based on age rather than weight was also calculated. For the 15 mm needle, 95% CI was not found in any age range, and the highest confidence cutoff was for using the needle in the age range of 0–8 years (91.9%). The 25 mm needle had a 97.8% CI from ages 10–16.CONCLUSIONS:The study reveals significant age-related variability in the proximal tibia’s anatomical dimensions, suggesting that standard 15 and 25 mm intraosseous needles may not reliably achieve optimal placement in pediatric patients. Our findings indicate that the current intraosseous needles may not be as safe as previously thought and support the need to develop improved intraosseous needle designs to enhance safety and therapeutic effectiveness in pediatric emergency care.

Objectives:Anterior cruciate ligament (ACL) injuries are reliably diagnosed through MRI and clinical examination. However, concomitant injuries such as anterolateral ligament (ALL) tears pose challenges for accurate assessment. Additionally, while the pivot-shift test effectively evaluates anterolateral rotatory laxity, its assessment may be hindered by patient discomfort during medical office evaluations, particularly in the acute phase. This study aimed to evaluate whether tibiofemoral rotation was associated with a concurrent ALL injury and pivot-shift grading in patients with a primary ACL tear.Methods:In this multicenter cross-sectional study, constituting a secondary analysis of previous studies, medical records and MRI scans of patients with unilateral primary ACL injury were reviewed. Demographics and pivot-shift grading were collected. Anterolateral ligament was identified on MRI coronal images and classified as intact or injured. Tibiofemoral rotation angle (TFA) was measured on axial MRI. Optimal TFA cutoff associated with ALL injury was identified by a receiver operating characteristic (ROC) curve.Results:A total of 206 patients were included, with a mean age of 28.3 ± 11.3 years. Among them, 152 (73.8%) exhibited signs of ALL injury. Pivot-shift tests were predominantly graded as 2 (71.4%), and notably, all grade 3 pivot-shift assessments were associated with ALL injury. The mean TFA was 4.5 ± 3.8 degrees, significantly higher in cases with ALL injury (5.2 ± 3.6 degrees) compared to intact ALL cases (2.7 ± 3.5 degrees; p < 0.001). A positive correlation was observed between pivot-shift grading and TFA (r = 0.204, p = 0.003). Optimal TFA cutoff value to predict ALL injury was 2.5 degrees (sensitivity 0.77; specificity 0.55). Patients with TFA angles at or above 2.5 degrees exhibited an increased likelihood of ALL injury compared to those below it (OR 3.34 - 95% CI: 1.74 to 6.42, p < 0.001). Interestingly, when TFA equal to or greater than 2.5 is combined with pivot-shift grade 2 or 3, the likelihood of ALL injury increased substantially to 13.68 (95% CI: 6.29 to 29.84, p < 0.001).Conclusions:Higher TFA was associated with an increased prevalence of ALL injuries and a high-grade pivot-shift in ACL-deficient patients. Patients with a TFA ≥ 2.5 degrees showed a 3-fold higher likelihood of ALL injuries at the time of ACL injury, and this risk further escalated with a higher-grade pivot-shift. Thus, assessing TFA in ACL reconstruction patients can guide decisions for concomitant anterolateral reconstruction, particularly in the presence of a high-grade pivot-shift.

Full-endoscopic extraforaminal lumbar discectomy: Use of 3-D image-guidance can mitigate risks and overcome steep learning curve.

Objectives:Femoroacetabular impingement (FAI) is a common hip condition that affects an estimated 54.4 persons per 100,000 with a higher prevalence in females. FAI is defined as abnormal contact between the femur and acetabulum leading to impingement, often as a result of bone deformities of the acetabulum or femoral neck. Diagnosis of FAI is often suspected based on history and physical exam and is typically confirmed with advanced imaging. Computed tomography (CT) has long been considered the gold standard for three-dimensional (3D) osseous morphology evaluation, while MRI has been used for soft tissue evaluation. Diagnosis often involves both modalities. While effective, this method is logistically burdensome, expensive, and may expose patients to ionizing CT scan radiation. Our institution incorporated a novel axial MRI sequence using in- and out-of-phase (ioMRI) sequences into an FAI-specific MRI to improve evaluation of osseous morphology.The aim of this study was to compare the ioMRI protocol to CT imaging with regard to FAI osseous evaluation. We performed a retrospective case-control series that compared musculoskeletal (MSK) radiologist dictated FAI measurements between ipsilateral hip CT and ioMRI studies. We hypothesize that the FAI measurements taken from the ioMRI will be similar to those obtained from CT.Methods:We conducted a retrospective electronic medical record (EMR) review of two attending sports fellowship-trained orthopedic surgeons to identify FAI patients who underwent both hip CT scan and ipsilateral ioMRI between May 2014 and May 2024. Each included patient must have had both imaging studies and had formal imaging interpretations from our institution’s fellowship-trained musculoskeletal radiologists. All reports contained each of the following hip morphometrics: midcoronal angle, midsagittal angle, acetabular version at 1, 2, and 3 o’clock, maximum femoral α-angle, femoral neck angle, and femoral neck version. Patient demographic data were also recorded.A two-way mixed model with absolute agreement was used to calculate interclass correlation coefficients (ICC’s) to compare ioMRI and CT scan measurement agreement. Results were interpreted as follows: minimal correlation < 0.2, poor correlation 0.2 to < 0.4, moderate correlation 0.4 to < 0.6, strong correlation 0.6 to ≤ 0.8, and almost perfect correlation > 0.8 (Koo 2016). Paired t-tests were conducted to compare means between CT and MRI hip measurements. All statistical testing was performed in RStudio (Version 2024.04.2).Results:Our initial chart review yielded 178 patients. After inclusion and exclusion criteria were applied, our final population studied included 95 patients (68 female). Flow diagram of patient data is depicted in Figure 1). Average patient age was 28.6 years and average BMI was 25.3 (Table 1).Average results for each hip morphometric measurement are reported in Table 2. ICC’s comparing modalities are reported in Table 3. An almost perfect correlation was found for the acetabular version measurement at 3 o’clock. Strong correlation between CT and MR imaging was found for femoral neck version measurement and acetabular version at 1 and 2 o’clock. Moderate correlation was found for the midcoronal angle, midsagittal angle, maximum α-angle, and femoral neck angle measurements.Our inter-method ICC’s were notable for moderate to almost perfect correlation across the board. Acetabular version at 3 o’clock had an ICC of 0.801, which is nearly indistinguishable. Likewise, acetabular version at 1 and 2 o’clock as well as femoral neck version all had strong ICC values of 0.617, 0.665, and 0.731 respectively.Conclusions:Our study found moderate to strong inter-method agreement between CT and ioMRI for all FAI evaluation hip measurements. This represents statistically significant correlation. This finding also demonstrates that ioMRI is able to provide osseous imaging that is fairly comparable to CT. Surgeons may anticipate that ioMRI measurements are reasonably reliable at estimating these measurements. However, several ioMRI measurements only had moderate ICC values, and CT imaging is likely still necessary for a thorough evaluation. The best performing measurements, acetabular version at 1, 2, and 3 o’clock, were all derived from the novel axial sequence of the ioMRI. Future work may consider applying this imaging technique to the other measurements. Our study is limited by its retrospective nature, and we did not assess inter-rater agreement between the many (>10) reading radiologists. However, the large number of reading radiologists extends real world applicability of our findings. Presently, ioMRI protocol studies likely cannot be used in place of CT scans for FAI osseous evaluation, but can provide a similar evaluation of hip osseous morphology. Surgeons may consider using isolated ioMRI in circumstances where CT may be contraindicated (cost, pregnancy, young age).

To propose and elucidate a novel MRI imaging feature, termed "Greta oto" sign, which is a specific sign for detecting axial myopathy patients with low back pain as the major clinical manifestation. From 2020 to 2024, four patients with axial myopathy with low back pain as the main clinical manifestation were retrospectively studied. The MRI findings and clinical features were analyzed. There were three females and one male, aged from 38 to 55 years. MRI examinations showed significant atrophy and fat infiltration of the bilateral erector spinae muscles, as well as varying degrees of atrophy observed in the multifidus muscles. The fatty degeneration in the erector spinae and multifidus muscles, together with the vertebral bodies and spinous processes, produced a distinctive imaging manifestation resembling a "Greta oto" sign on axial T2-weighted MRI scans in all patients. Ultimately, all four patients were referred to the neurology department, where the diagnosis of myopathy was confirmed by muscle biopsy (one myopathy with early-onset Paget disease and frontotemporal dementia type 1 (IBMPFD), one HMGCR antibody-positive myopathy, one myotonic dystrophy and one limb-girdle muscular dystrophy. The "Greta Otto" sign may be a specific MRI manifestation of axial myopathy with low back pain as the primary clinical manifestation. This sign may help clinicians identify axial myopathy and reduce the likelihood of misdiagnosis as nonspecific low back pain.

Background: Chondromalacia of patella is a common disease characterized by softening and degeneration of the patellar articular cartilage and is a frequent cause of anterior knee pain in young adults. Objectives: The purpose of this study is to assess sensitivity, specificity and the accuracy of axial proton density –spectral attenuated inversion recovery MRI sequence in detecting and grading patellar cartilage in patients with chondromalacia patellae in comparison to axial proton density sequence. Patients and methods: Thirty patients with chondromalacia patellae would be included in our study, 18 patients were women and 12 were men, their ages ranged from 12-30 years old. Another thirty-one patients that underwent knee MRI for another knee problem will be included also as a standard control group, 23were males and 8 were females, their ages ranging from 13-36 years old. All these patients will be examined by MR imaging with 1.5 tesla imaging system using both the axial proton density- spectral attenuated inversion recovery (PD-SPAIR) and axial proton density (PD) sequences. The study was done in AL Emamain Alkadhmain medical city. All the results will be reviewed 2 radiologists; final grading of patellar chondromalacia was made with mutual agreement. Results: Sensitivity of PD-SPAIR and PD sequences was 86.7% and 70% respectively. Specificity was for PD-SPAIR 93.5% and 77.4% for PD sequences. Accuracy for detecting these lesions in comparison to the control groupwas90.1% for PD-SPAIR and 73.8% for PD sequence Conclusion: The axial proton density spectrum attenuated inversion recovery sequence may accurately and quickly detect and grade cartilage defects in chondromalacia patellae patients. Fat saturation and proton density–weighted sequences are sensitive to cartilage lesions and intramedullary osseous oedematous changes, thus they should substitute the traditional proton density sequence in these patients.

The dural deviation ratio: a novel indicator for preoperative differentiation of intradural extension in spinal dumbbell schwannomas using Axial T2-weighted MRI.

MRI evaluation of peroneus brevis tendon position: Anatomical variants in individuals with normal peroneal tendons to improve recognition and prevent misdiagnosis.