Axial T2-weighted Magnetic Resonance Imaging Research Articles

Magnetic resonance image segmentation of the compressed spinal cord in patients with degenerative cervical myelopathy using convolutional neural networks.

Spinal cord segmentation is the first step in atlas-based spinal cord image analysis, but segmentation of compressed spinal cords from patients with degenerative cervical myelopathy is challenging. We applied convolutional neural network models to segment the spinal cord from T2-weighted axial magnetic resonance images of DCM patients. Furthermore, we assessed the correlation between the cross-sectional area segmented by this network and the neurological symptoms of the patients. The CNN architecture was built using U-Net and DeepLabv3 + and PyTorch. The CNN was trained on 2762 axial slices from 174 patients, and an additional 517 axial slices from 33 patients were held out for validation and 777 axial slices from 46 patients for testing. The performance of the CNN was evaluated on a test dataset with Dice coefficients as the outcome measure. The ratio of CSA at the maximum compression level to CSA at the C2 level, as segmented by the CNN, was calculated. The correlation between the spinal cord CSA ratio and the Japanese Orthopaedic Association score in DCM patients from the test dataset was investigated using Spearman's rank correlation coefficient. The best Dice coefficient was achieved when U-Net was used as the architecture and EfficientNet-b7 as the model for transfer learning. Spearman's rs between the spinal cord CSA ratio and the JOA score of DCM patients was 0.38 (p = 0.007), showing a weak correlation. Using deep learning with magnetic resonance images of deformed spinal cords as training data, we were able to segment compressed spinal cords of DCM patients with a high concordance with expert manual segmentation. In addition, the spinal cord CSA ratio was weakly, but significantly, correlated with neurological symptoms. Our study demonstrated the first steps needed to implement automated atlas-based analysis of DCM patients.

Prediction of anterior tibiotalar ligament injury: measurement of the angle between the deep posterior tibiotalar ligament and talus on MRI.

Identification of anterior tibiotalar ligament (aTTL) injury is essential because it influences the surgeon's treatment option and patient prognosis. To assess the diagnostic accuracy of the angle measurement between the talus and posterior tibiotalar ligament (talus-pTTL) on magnetic resonance imaging (MRI) in patients with arthroscopically proven aTTL injuries. Ankle MRI scans of 67 patients who underwent arthroscopic examination were retrospectively reviewed. The talus-pTTL angle on axial T2-weighted MRI and the medial clear space (MCS) on mortise ankle radiograph were measured. Inter-observer agreement of the measurements was calculated. Also, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were the metrics of diagnostic accuracy. AUC was 0.90 for observer 1 with 78.6% sensitivity, 97.4% specificity, 88% accuracy, and 54.7° cutoff value for the talus-pTTL angle. AUC was 0.87 for observer 2 with 85.7% sensitivity, 84.6% specificity, 85.2% accuracy, and 53.7° cutoff value for the talus-pTTL angle. AUC was 0.86 with 82.1% sensitivity, 79.5% specificity, and 80.8% accuracy for observer 1 and 0.79 with 57.1% sensitivity, 92.3% specificity, and 74.7% accuracy for observer 2 for the MCS. Different MCS values and additional capabilities when complemented with the angle measurement showed an increase in diagnostic performances. Intra-observer reliability of MCS and talus-pTTL angle of the two radiologists was excellent. Inter-observer reliability of the two radiologists was excellent for both the talus-pTTL angle (0.95) and the MCS (0.85). Measurement of the talus-pTTL angle showed good sensitivity, specificity, and accuracy for the evaluation of aTTL injury with excellent inter-observer reliability.

Deep Learning for Multi-Tissue Segmentation and Fully Automatic Personalized Biomechanical Models from BACPAC Clinical Lumbar Spine MRI.

In vivo retrospective study of fully automatic quantitative imaging feature extraction from clinically acquired lumbar spine magnetic resonance imaging (MRI). To demonstrate the feasibility of substituting automatic for human-demarcated segmentation of major anatomic structures in clinical lumbar spine MRI to generate quantitative image-based features and biomechanical models. Previous studies have demonstrated the viability of automatic segmentation applied to medical images; however, the feasibility of these networks to segment clinically acquired images has not yet been demonstrated, as they largely rely on specialized sequences or strict quality of imaging data to achieve good performance. Convolutional neural networks were trained to demarcate vertebral bodies, intervertebral disc, and paraspinous muscles from sagittal and axial T1-weighted MRIs. Intervertebral disc height, muscle cross-sectional area, and subject-specific musculoskeletal models of tissue loading in the lumbar spine were then computed from these segmentations and compared against those computed from human-demarcated masks. Segmentation masks, as well as the morphological metrics and biomechanical models computed from those masks, were highly similar between human- and computer-generated methods. Segmentations were similar, with Dice similarity coefficients of 0.77 or greater across networks, and morphological metrics and biomechanical models were similar, with Pearson R correlation coefficients of 0.69 or greater when significant. This study demonstrates the feasibility of substituting computer-generated for human-generated segmentations of major anatomic structures in lumbar spine MRI to compute quantitative image-based morphological metrics and subject-specific musculoskeletal models of tissue loading quickly, efficiently, and at scale without interrupting routine clinical care.

Evaluation of iliotibial band volume with 3T MRI

Aim: The aim of our study is to evaluate the iliotibial band thickness (ITBT) and volume (ITBV) according to age and gender with 3 T MRI in a healthy Turkish population.
 Material and Method: In the current study, 150 patients who had knee MRI were retrospectively evaluated. They were divided into the following groups to examine the effects of age: group 1: 18–30 years old; group 2: 31–40 years old; group 3: 41–50 years old; and group 4: 51–60 years old. ITBV measurements were performed on axial T2-weighted knee MR images.
 Results: The mean age was 42.2±29.6 (range 18-60) years. The mean ITBT was 1.76±0.22 mm and the mean ITBV was 20,24±1,44 mm3 in all patients. There was statistically significant difference in mean iliotibial band thickness and volume between genders (p=0.001; p=0.001). There were no statistically significant differences in mean iliotibial band thickness and volume values between the groups in the one-way ANOVA test (p >0.05). 
 Conclusion: Radiological studies on iliotibial band thickness and band area in healthy individuals are new in the literature. This study is the first radiological study in which the volume of the iliotibial band was measured in healthy individuals and was performed on a 3 tesla MR device. Also, it is the first study that is used artificial intelligence for iliotibial band evaluation and the technic is effective and more rapid according to our experience. ITBV and ITBT are statistically significantly higher in males than females. Most thick ITB was detected between the ages of 31-40. The values of our study, especially ITBT, differ from previous studies and the values are altered in a wide range. Therefore, standardization in the calculation is necessary. This topic is open to future research.

Preoperative Association Between Quantitative Lumbar Muscle Parameters and Spinal Sagittal Alignment in Lumbar Fusion Patients.

A retrospective cross-sectional study. To assess the association between spinal muscle morphology and spinopelvic parameters in lumbar fusion patients, with a special emphasis on lumbar lordosis (LL). Maintenance of sagittal alignment relies on muscle forces, but the basic association between spinal muscles and spinopelvic parameters is poorly understood. Patients operated between 2014 and 2017 who had both lumbar magnetic resonance imaging scan and standing whole-spine radiographs within six months before surgery were included. Muscle measurements were conducted on axial T2-weighted magnetic resonance images at the superior endplate L3-L5 for the psoas and L3-S1 for combined multifidus and erector spinae (paraspinal) muscles. A pixel intensity threshold method was used to calculate the total cross-sectional area (TCSA) and the functional cross-sectional area (FCSA). Spinopelvic parameters were measured on lateral standing whole-spine radiographs and included LL, pelvic incidence (PI), PI-LL mismatch, pelvic tilt, sacral slope, thoracic kyphosis, and sagittal vertical axis. Analyses were stratified by biological sex. Multivariable linear regression analyses with adjustments for age and body mass index (BMI) were performed. A total of 104 patients (62.5% female) were included in the analysis. The patient population was 90.4% White with a median age at surgery of 69 years and a median BMI of 27.8kg/m 2 . All muscle measurements were significantly smaller in women. PI, pelvic tilt, and thoracic kyphosis were significantly greater in women. PI-LL mismatch was 6.1° (10.6°) in men and 10.2° (13.5°) in women ( P =0.106), and sagittal vertical axis was 45.3 (40.8)mm in men and 35.7 (40.8)mm in women ( P =0.251). After adjusting for age and BMI, paraspinal TCSA at L3-L5, and paraspinal FCSA at L4 showed significant positive associations with LL in women. In men, psoas TCSA at L5 and psoas FCSA at L5 showed significant negative associations with LL, but none of the paraspinal muscle measurements. Our findings indicate that psoas and lumbar spine extensor muscles interact differently on LL among men and women, creating a unique mechanical environment. Level 4.

Effects of Expression of Matrix Metalloproteinases and Discoidin Domain Receptors in Ligamentum Flavum Fibrosis in Patients with Degenerative Lumbar Canal Stenosis.

This is a retrospective cohort study. This study aimed to clarify the role of crosstalk between discoidin domain receptors (DDRs) and matrix metalloproteinases (MMPs) in the ligamentum flavum (LF) fibrosis obtained from patients with degenerative lumbar canal stenosis (DLCS). The DDRs, DDR1 and DDR2, are cell surface receptors and have an essential role in collagen fiber accumulation in several fibrotic diseases. MMPs are one of the critical factors in extracellular matrix remodeling and elastic fiber degradation in LF tissues. However, the crosstalk between DDRs and MMPs and the role of this molecular signal in LF fibrosis remain unclear. A total of 35 patients were divided into two groups in this study. Spinal surgery was performed in 23 of these patients with the diagnosis of DLCS. Twelve patients with lumbar disk herniation (LDH) were included in the control group. On axial T2-weighted magnetic resonance imaging, LF thickness was measured bilaterally at the level of the facet joint. Histology, quantitative real-time polymerase chain reaction (qRT-PCR), and Western blot analyses were performed on LF tissue samples. LF tissues were stained with hematoxylin and eosin. In addition, the grade of fibrosis was histologically assessed using Masson trichrome triple staining. DDR1 and DDR2 Western blot analyses were performed. DDR1, DDR2, MMP2, MMP3, MMP9, and MMP13 expression levels were measured using qRT-PCR analysis. The grade of fibrosis and LF thickness were significantly higher in the DLCS patients than in the LDH patients. DDR1 and DDR2 gene expression and protein levels in LF tissues are significantly greater in DLCS samples than in control samples, according to both qRT-PCR and Western blot analyses. In addition, we detected a significant expression of the MMP3, MMP9, and MMP13, which are known to have important roles in extracellular matrix remodeling in DLCS. Furthermore, we discovered a link between DDR protein levels and LF thickness, fibrosis, and MMP3/MMP9. Our results indicate that DDR1, DDR2, and MMP3 and MMP9 signals can be correlated with each other in LF tissues and be promoted LF fibrosis leading to spinal canal narrowing in patients with DLCS.

Gluteal muscle fatty infiltration, fall risk, and mobility limitation in older women with urinary incontinence: a pilot study.

Gluteal muscle quality influences risk of falling and mobility limitation. We sought (1) to compare gluteal muscle fatty infiltration (FI) between groups of older women with urinary incontinence (UI) at risk for falling (at-risk group) and not at risk for falling (not-at-risk group), and (2) to determine correlation of gluteal FI with Timed Up and Go (TUG) and Short Physical Performance Battery (SPPB) performance. Prospective pilot study of gluteal FI on pelvis MRI for 19 women with UI, aged ≥ 70years. A musculoskeletal radiologist selected axial T1-weighted MR images; then, two blinded medical student research assistants analyzed gluteal FI by quantitative fuzzy C-means segmentation. TUG and SPPB tests were performed. TUG ≥ 12s defined participants as at risk for falling. Descriptive, correlation, and reliability analyses were performed. Mean age, 76.3 ± 4.8years; no difference for age or body mass index (BMI) between the at-risk (n = 5) versus not-at-risk (n = 14) groups. SPPB score (p = 0.013) was lower for the at-risk group (6.4 ± 3.1) than for the not-at-risk group (10.2 ± 1.9). Fuzzy C-means FI-%-estimate differed between the at-risk group and the not-at-risk group for bilateral gluteus medius/minimus (33.2% ± 15.6% versus 19.5% ± 4.1%, p = 0.037) and bilateral gluteus maximus (33.6% ± 15.6% versus 19.7% ± 6.9%, p = 0.047). Fuzzy C-means FI-%-estimate for bilateral gluteus maximus had significant (p < 0.050) moderate correlation with age (rho = - 0.64), BMI (rho = 0.65), and TUG performance (rho = 0.52). Fuzzy C-means FI-%-estimates showed excellent inter-observer and intra-observer reliability (intraclass correlation coefficient, ≥ 0.892). Older women with UI at risk for falling have greater levels of gluteal FI and mobility limitation as compared to those not at risk for falling.

A novel MRI index for paraspinal muscle fatty infiltration: reliability and relation to pain and disability in lumbar spinal stenosis: results from a multicentre study

BackgroundFatty infiltration of the paraspinal muscles may play a role in pain and disability in lumbar spinal stenosis. We assessed the reliability and association with clinical symptoms of a method for assessing fatty infiltration, a simplified muscle fat index (MFI).MethodsPreoperative axial T2-weighted magnetic resonance imaging (MRI) scans of 243 patients aged 66.6 ± 8.5 years (mean ± standard deviation), 119 females (49%), with symptomatic lumbar spinal stenosis were assessed. Fatty infiltration was assessed using both the MFI and the Goutallier classification system (GCS). The MFI was calculated as the signal intensity of the psoas muscle divided by that of the multifidus and erector spinae. Observer reliability was assessed in 102 consecutive patients for three independent investigators by intraclass correlation coefficient (ICC) and 95% limits of agreement (LoA) for continuous variables and Gwet’s agreement coefficient (AC1) for categorical variables. Associations with patient-reported pain and disability were assessed using univariate and multivariate regression analyses.ResultsInterobserver reliability was good for the MFI (ICC 0.79) and fair for the GCS (AC1 0.33). Intraobserver reliability was good or excellent for the MFI (ICC range 0.86–0.91) and moderate to almost perfect for the GCS (AC1 range 0.55–0.92). Mean interobserver differences of MFI measurements ranged from -0.09 to -0.04 (LoA -0.32 to 0.18). Adjusted for potential confounders, none of the disability or pain parameters was significantly associated with MFI or GCS.ConclusionThe proposed MFI demonstrated high observer reliability but was not associated with preoperative pain or disability.

Prediction of Motor Outcome of Stroke Patients Using a Deep Learning Algorithm with Brain MRI as Input Data

Background: Deep learning techniques can outperform traditional machine learning techniques and learn from unstructured and perceptual data, such as images and languages. We evaluated whether a convolutional neural network (CNN) model using whole axial brain T2-weighted magnetic resonance (MR) images as input data can help predict motor outcomes of the upper and lower limbs at the chronic stage in stroke patients. Methods: We collected MR images taken at the early stage of stroke in 1,233 consecutive stroke patients. We categorized modified Brunnstrom classification (MBC) scores of ≥5 and functional ambulatory category (FAC) scores of ≥4 at 6 months after stroke as favorable outcomes in the upper and lower limbs, respectively, and MBC scores of <5 and FAC scores of <4 as poor outcomes. We applied a CNN to train the image data. Of the 1,233 patients, 70% (863 patients) were randomly selected for the training set and the remaining 30% (370 patients) were assigned to the validation set. Results: In the prediction of upper limb motor function on the validation dataset, the area under the curve (AUC) was 0.768, and for lower limb motor function, the AUC was 0.828. Conclusion: We showed that a CNN model trained using whole-brain axial T2-weighted MR images of stroke patients would help predict upper and lower limb motor function at the chronic stage.

The Effects of Reconditioning Exercises Following Prolonged Bed Rest on Lumbopelvic Muscle Volume and Accumulation of Paraspinal Muscle Fat

Reduced muscle size and accumulation of paraspinal muscle fat content (PFC) have been reported in lumbopelvic muscles after spaceflights and head-down tilt (HDT) bed rest. While some information is available regarding reconditioning programs on muscle atrophy recovery, the effects on the accumulation of PFC are unknown. Recently, a device (the Functional Re-adaptive Exercise Device—FRED) has been developed which aims to specifically recruit lumbopelvic muscles. This study aimed to investigate the effects of a standard reconditioning (SR) program and SR program supplemented by FRED (SR + FRED) on the recovery of the lumbopelvic muscles following 60-day HDT bed rest. Twenty-four healthy participants arrived at the facility for baseline data collection (BDC) before the bed rest period. They remained in the facility for 13-day post-HDT bed rest and were randomly allocated to one of two reconditioning programs: SR or SR + FRED. Muscle volumes of the lumbar multifidus (LM), lumbar erector spinae (LES), quadratus lumborum (QL), and psoas major (PM) muscles were measured from axial T1-weighted magnetic resonance imaging (MRI) at all lumbar intervertebral disc levels. PFC was determined using a chemical shift-based lipid/water Dixon sequence. Each lumbopelvic muscle was segmented into four equal quartiles (from medial to lateral). MRI of the lumbopelvic region was conducted at BDC, Day-59 of bed rest (HDT59), and Day-13 after reconditioning (R13). Comparing R13 with BDC, the volumes of the LM muscle at L4/L5 and L5/S1, LES at L1/L2, and QL at L3/L4 had not recovered (all—p < 0.05), and the PM muscle remained larger at L1/L2 (p = 0.001). Accumulation of PFC in the LM muscle at the L4/L5 and L5/S1 levels remained higher in the centro-medial regions at R13 than BDC (all—p < 0.05). There was no difference between the two reconditioning programs. A 2-week reconditioning program was insufficient to fully restore all volumes of lumbopelvic muscles and reverse the accumulation of PFC in the muscles measured to BDC values, particularly in the LM muscle at the lower lumbar levels. These findings suggest that more extended reconditioning programs or alternative exercises may be necessary to fully restore the size and properties of the lumbopelvic muscles after prolonged bed rest.

Deep learning models for triaging hospital head MRI examinations.

The growing demand for head magnetic resonance imaging (MRI) examinations, along with a global shortage of radiologists, has led to an increase in the time taken to report head MRI scans in recent years. For many neurological conditions, this delay can result in poorer patient outcomes and inflated healthcare costs. Potentially, computer vision models could help reduce reporting times for abnormal examinations by flagging abnormalities at the time of imaging, allowing radiology departments to prioritise limited resources into reporting these scans first. To date, however, the difficulty of obtaining large, clinically-representative labelled datasets has been a bottleneck to model development. In this work, we present a deep learning framework, based on convolutional neural networks, for detecting clinically-relevant abnormalities in minimally processed, hospital-grade axial T2-weighted and axial diffusion-weighted head MRI scans. The models were trained at scale using a Transformer-based neuroradiology report classifier to generate a labelled dataset of 70,206 examinations from two large UK hospital networks, and demonstrate fast (< 5s), accurate (area under the receiver operating characteristic curve (AUC) > 0.9), and interpretable classification, with good generalisability between hospitals (ΔAUC ≤ 0.02). Through a simulation study we show that our best model would reduce the mean reporting time for abnormal examinations from 28 days to 14 days and from 9 days to 5 days at the two hospital networks, demonstrating feasibility for use in a clinical triage environment.

MRI-Based Radiomic Features Help Identify Lesions and Predict Histopathological Grade of Hepatocellular Carcinoma.

Hepatocellular carcinoma (HCC) is the most common form of liver cancer. Radiomics is a promising tool that may increase the value of magnetic resonance imaging (MRI) in the management of HCC. The purpose of our study is to develop an MRI-based radiomics approach to preoperatively detect HCC and predict its histological grade. Thirty-eight HCC patients at staging who underwent axial T2-weighted and dynamic contrast-enhanced MRI (DCE-MRI) were considered. Three-dimensional volumes of interest (VOIs) were manually placed on HCC lesions and normal hepatic tissue (HT) on arterial phase post-contrast images. Radiomic features from T2 images and arterial, portal and tardive post-contrast images from DCE-MRI were extracted by using Pyradiomics. Feature selection was performed using correlation filter, Wilcoxon-rank sum test and mutual information. Predictive models were constructed for HCC differentiation with respect to HT and HCC histopathologic grading used at each step an imbalance-adjusted bootstrap resampling (IABR) on 1000 samples. Promising results were obtained from radiomic prediction models, with best AUCs ranging from 71% to 96%. Radiomics MRI based on T2 and DCE-MRI revealed promising results concerning both HCC detection and grading. It may be a suitable tool for personalized treatment of HCC patients and could also be used to develop new prognostic biomarkers useful for HCC assessment without the need for invasive procedures.

The Relationship Between Wild-Type Transthyretin Amyloid Load and Ligamentum Flavum Thickness in Lumbar Stenosis Patients

One key contributor to lumbar stenosis is thickening of the ligamentum flavum (LF), a process still poorly understood. Wild-type transthyretin amyloid (ATTRwt) has been found in the LF of patients undergoing decompression surgery, suggesting that amyloid may play a role. However, it is unclear whether within patients harboring ATTRwt, the amount of amyloid is associated with LF thickness. From an initial cohort of 324 consecutive lumbar stenosis patients whose LF specimens from decompression surgery were sent for analysis (2018-2019), 33 patients met the following criteria: 1) Congo red-positive amyloid in the LF, 2) ATTRwt by mass spectrometry-based proteomics, and 3) an available preoperative magnetic resonance imaging. Histological specimens were digitized, and amyloid load was quantified through Trainable Weka Segmentation machine learning. LF thicknesses were manually measured on axial T2-weighted preoperative magnetic resonance imaging scans at each lumbar level, L1-S1. The sum of thicknesses at every lumbar LF level (L1-S1) equals "lumbar LF burden". Patients had a mean age of 72.7years (range= 59-87), were mostly male (61%) and white (82%), and predominantly had surgery at L4-L5 levels (73%). Amyloid load was positively correlated with LF thickness (R= 0.345, P= 0.0492) at the levels of surgical decompression. Furthermore, amyloid load was positively correlated with lumbar LF burden (R= 0.383, P= 0.0279). Amyloid load is positively correlated with LF thickness and lumbar LF burden across all lumbar levels, in a dose-dependent manner. Further studies are needed to validate these findings, uncover the underlying pathophysiology, and pave the way toward using therapies that slow LF thickening.

Magnetic Resonance Imaging Documentation of Approach Trauma With Lumbar Endoscopic Interlaminar, Translaminar, Compared to Open Microsurgical Discectomy.

Endoscopic spine surgery is associated with less approach trauma than conventional open translaminar surgery. However, objective evidence to corroborate this empiric observation is scarce. Preservation of the anatomic attachment and sensory function of multifidus muscles have been stipulated to be critical to maintaining the normal function of the lumbar spinal motion segment. The authors were interested in comparing the postoperative approach trauma between traditional open translaminar microsurgical and interlaminar endoscopic discectomy. The approach trauma to the paraspinal muscles due to interlaminar or open microsurgical discectomy was evaluated on T2-weighted axial magnetic resonance imaging (MRI) images of 39 consecutive patients who underwent lumbar disc surgery. Postoperative MRI images taken at 4 days and within 1 year after surgery were analyzed. Eleven patients underwent conventional open translaminar microdiscectomy surgery. Endoscopic discectomy was performed on 17 patients via the interlaminar and on another 11 patients via the transforaminal approach. The immediate surgical approach trauma was estimated as the defect zone by measuring the normalized relative cross-sectional area (rCSA) of muscle disruption in the surgical corridor 4 days postoperatively. The long-term effect of surgical approach trauma was assessed by measuring the area of the paraspinal muscles that had been replaced by fatty tissue 1 year postoperatively. The rCSA data showed diminished approach trauma with a smaller surgical defect zone in the interlaminar endoscopy group (17.6%) was smaller than in the microsurgical group 4 days postoperatively (41.2%). At 1 year postoperatively, the mean fatty replacement of the paraspinal muscles was 23.6% after microsurgery and 2.1% after the interlaminar endoscopy. Muscle recovery was substantially reduced in the interlaminar endoscopic group, with the muscle zone reducing from 20.3% to 2.1% when analyzed 1 year postoperatively. In the microsurgery group, the muscle damage by atrophy increased from 41.2% to 62.9% at 1 year postoperatively (P < 0.001). Fatty replacement of the multifidus muscle was seen on the ipsilateral and contralateral approach side. There was a negligible change in the muscle zone with the transforaminal approach. Tissue trauma was significantly reduced with endoscopic surgery techniques compared with the traditional translaminar microdiscectomy approach. There was a minor postoperative tissue trauma and hardly any long-term replacement of the multifidus and paraspinal muscles by fatty tissue 1 year postoperatively with the endoscopic technique. The transforaminal approach has the least effect on the paraspinal muscles of the surgical motion segment. Further study is needed to investigate whether these findings translate into decreased postoperative instability or low back pain following endoscopic discectomy surgery. MRI analysis of multifidus atrophy following various lumbar translaminar and transforaminal decompression techniques.

Automatic MRI segmentation of pectoralis major muscle using deep learning

To develop and validate a deep convolutional neural network (CNN) method capable of selecting the greatest Pectoralis Major Cross-Sectional Area (PMM-CSA) and automatically segmenting PMM on an axial Magnetic Resonance Imaging (MRI). We hypothesized a CNN technique can accurately perform both tasks compared with manual reference standards. Our method is based on two steps: (A) segmentation model, (B) PMM-CSA selection. In step A, we manually segmented the PMM on 134 axial T1-weighted PM MRIs. The segmentation model was trained from scratch (MONAI/Pytorch SegResNet, 4 mini-batch, 1000 epochs, dropout 0.20, Adam, learning rate 0.0005, cosine annealing, softmax). Mean-dice score determined the segmentation score on 8 internal axial T1-weighted PM MRIs. In step B, we used the OpenCV2 (version 4.5.1, https://opencv.org) framework to calculate the PMM-CSA of the model predictions and ground truth. Then, we selected the top-3 slices with the largest cross-sectional area and compared them with the ground truth. If one of the selected was in the top-3 from the ground truth, then we considered it to be a success. A top-3 accuracy evaluated this method on 8 axial T1-weighted PM MRIs internal test cases. The segmentation model (Step A) produced an accurate pectoralis muscle segmentation with a Mean Dice score of 0.94 ± 0.01. The results of Step B showed top-3 accuracy > 98% to select an appropriate axial image with the greatest PMM-CSA. Our results show an overall accurate selection of PMM-CSA and automated PM muscle segmentation using a combination of deep CNN algorithms.

Relationship between facet joint opening on CT and facet joint effusion on MRI in patients with lumbar spinal stenosis: analysis of a less invasive decompression procedure.

Both facet joint opening (FJO) on CT and facet joint effusion (FJE) on MRI are reportedly indicators of segmental instability in the lumbar facet joints of patients with lumbar spinal stenosis (LSS). However, no study has investigated both parameters simultaneously. Therefore, the association between these findings and which parameter is better for predicting clinical outcomes after surgical treatment remains unclear. The purpose of this study was to investigate the relationship between FJO and FJE in patients who underwent less invasive decompression procedures for LSS and to investigate the impact of these findings on clinical outcomes. This study included 1465 lumbar levels (L1-2 to L5-S1) in 293 patients who underwent less invasive surgery for LSS and had ≥ 5 years of follow-up. FJO was defined as joint space widening ≥ 2 mm on preoperative axial CT images. FJE was defined as fluid effusion in the facet joint on preoperative axial T2-weighted MR images. The characteristics and distributions of FJO and FJE were investigated with other preoperative radiological findings. The association between need for further surgery and FJO/FJE was analyzed according to intervertebral level. FJO was observed at 402 levels (27%), and FJE was found at 306 levels (21%). The correspondence rate between FJO and FJE was 70% (kappa 0.195, p < 0.01). One hundred thirty-seven levels (9%) had both FJO and FJE. Levels with both FJO and FJE more commonly had lateral olisthesis, lateral wedging, and axial intervertebral rotation than other levels (p < 0.001). Levels with both FJO and FJE were more likely than other levels to need further surgery (OR 2.42, p = 0.027). The correspondence rate between FJO and FJE was not high. However, multivariate analysis showed that levels with both FJO and FJE had a higher risk of requiring further surgery than those with other radiological findings, such as lateral olisthesis, lateral wedging, and axial intervertebral rotation. Patients with levels with both FJO and FJE need careful long-term follow-up after undergoing a less invasive decompression procedure.

External Validation of a Radiomics Model for the Prediction of Complete Response to Neoadjuvant Chemoradiotherapy in Rectal Cancer.

Simple SummaryIn locally advanced rectal cancer (LARC), a minority of patients presents a pathological complete response (pCR) after neoadjuvant chemoradiotherapy (CRT). In this sub-population, organ preservation could be proposed without compromising overall survival. Using a robust neural network based statistical approach, correction of imbalanced data and inter-center variability, a radiomics-based model was externally validated with a balanced accuracy of 85.5%. This model efficiently predicted the patients with a pCR in an external cohort and could be used to select the patients eligible for organ preservation.Objective: Our objective was to develop a radiomics model based on magnetic resonance imaging (MRI) and contrast-enhanced computed tomography (CE-CT) to predict pathological complete response (pCR) to neoadjuvant treatment in locally advanced rectal cancer (LARC). Material: All patients treated for a LARC with neoadjuvant CRT and subsequent surgery in two separate institutions between 2012 and 2019 were considered. Both pre-CRT pelvic MRI and CE-CT were mandatory for inclusion. The tumor was manually segmented on the T2-weighted and diffusion axial MRI sequences and on CE-CT. In total, 88 radiomic parameters were extracted from each sequence using the Miras© software, with a total of 822 features by patient. The cohort was split into training (Institution 1) and testing (Institution 2) sets. The ComBat and Synthetic Minority Over-sampling Technique (SMOTE) approaches were used to account for inter-institution heterogeneity and imbalanced data, respectively. We selected the most predictive characteristics using Spearman’s rank correlation and the Area Under the ROC Curve (AUC). Five pCR prediction models (clinical, radiomics before and after ComBat, and combined before and after ComBat) were then developed on the training set with a neural network approach and a bootstrap internal validation (n = 1000 replications). A cut-off maximizing the model’s performance was defined on the training set. Each model was then evaluated on the testing set using sensitivity, specificity, balanced accuracy (Bacc) with the predefined cut-off. Results: Out of the 124 included patients, 14 had pCR (11.3%). After ComBat harmonization, the radiomic and the combined models obtained a Bacc of 68.2% and 85.5%, respectively, while the clinical model and the pre-ComBat combined achieved respective Baccs of 60.0% and 75.5%. Conclusions: After correction of inter-site variability and imbalanced data, addition of radiomic features enhances the prediction of pCR after neoadjuvant CRT in LARC.

Lumbar Spine Computed Tomography to Magnetic Resonance Imaging Synthesis Using Generative Adversarial Network: Visual Turing Test.

(1) Introduction: Computed tomography (CT) and magnetic resonance imaging (MRI) play an important role in the diagnosis and evaluation of spinal diseases, especially degenerative spinal diseases. MRI is mainly used to diagnose most spinal diseases because it shows a higher resolution than CT to distinguish lesions of the spinal canals and intervertebral discs. When it is inevitable for CT to be selected instead of MR in evaluating spinal disease, evaluation of spinal disease may be limited. In these cases, it is very helpful to diagnose spinal disease with MR images synthesized with CT images. (2) Objective: To create synthetic lumbar magnetic resonance (MR) images from computed tomography (CT) scans using generative adversarial network (GAN) models and assess how closely the synthetic images resembled the true images using visual Turing tests (VTTs). (3) Material and Methods: Overall, 285 patients aged ≥ 40 years who underwent lumbar CT and MRI were enrolled. Based on axial CT and T2-weighted axial MR images from 285 patients, an image synthesis model using a GAN was trained using three algorithms (unsupervised, semi-supervised, and supervised methods). Furthermore, VTT to determine how similar the synthetic lumbar MR images generated from lumbar CT axial images were to the true lumbar MR axial images were conducted with 59 patients who were not included in the model training. For the VTT, we designed an evaluation form comprising 600 randomly distributed axial images (150 true and 450 synthetic images from unsupervised, semi-supervised, and supervised methods). Four readers judged the authenticity of each image and chose their first- and second-choice candidates for the true image. In addition, for the three models, structural similarities (SSIM) were evaluated and the peak signal to noise ratio (PSNR) was compared among the three methods. (4) Results: The mean accuracy for the selection of true images for all four readers for their first choice was 52.0% (312/600). The accuracies of determining the true image for each reader’s first and first + second choices, respectively, were as follows: reader 1, 51.3% and 78.0%; reader 2, 38.7% and 62.0%, reader 3, 69.3% and 84.0%, and reader 4, 48.7% and 70.7%. In the case of synthetic images chosen as first and second choices, supervised algorithm-derived images were the most often selected (supervised, 118/600 first and 164/600 second; semi-supervised, 90/600 and 144/600; and unsupervised, 80/600 and 114/600). For image quality, the supervised algorithm received the best score (PSNR: 15.987 ± 1.039, SSIM: 0.518 ± 0.042). (5) Conclusion: This was the pilot study to apply GAN to synthesize lumbar spine MR images from CT images and compare training algorithms of the GAN. Based on VTT, the axial MR images synthesized from lumbar CT using GAN were fairly realistic and the supervised training algorithm was found to provide the closest image to true images.

The prevalence of Schmorl's nodes in osteoporotic vs normal patients: a Middle Eastern population study.

This study aims to identify the prevalence of Schmorl's nodes (SNs) in a cohort of Omani nationals, and also to determine any relation between osteoporosis and Schmorl's nodes. This retrospective observational study was conducted on Omani nationals. One thousand three hundred and forty-eight DEXA scan patients were included. Of these, 545 patients had complete X-rays and MRI scans that would help determine the SN status. The X-rays and sagittal, coronal, and axial T2-weighted MR images were used to identify the presence and exact location of the Schmorl nodes by one orthopedic trainee and confirmed by the senior author. The correlation of each parameter with the presence of SN was analyzed by the independent-samples T test and one-way ANOVA. The overall prevalence of SN in this population sample appeared to be 41.1%. Over 88% of the SN-positive cases were either osteopenic or frankly osteoporotic by the WHO definition. Vast majority of SNs (87.1%) occurred in the lumbar spine and were central in location and mostly solitary. Statistical analysis of the data revealed significant correlation between osteopenia or osteoporosis and the presence of SNs. The prevalence of SN in the sample of Omanis studied was 41.1% and was most frequently seen in older men in the lumbar spine. It is strongly associated with osteoporosis/osteopenia (88.4%) and frequently presents as solitary central lesions.

Imaging of Spinal Bone Tumors: Principles and Practice.

Age, location of the tumor, and detailed patient history can narrow the differential diagnosis of spinal bone lesions, including metastasis and primary benign and malignant bone tumors. Computed tomography and magnetic resonance imaging are both crucial in evaluating the characteristics of spinal bone tumors. Growth speed and Lodwick margin description can differentiate malignant from benign tumors to a certain degree. Positron emission tomography has a limited ability to differentiate malignant from benign tumors. A biopsy is often required for a definitive diagnosis. To select the optimal treatment for spinal metastasis, neurological status by epidural spinal cord compression grade (axial T2-weighted magnetic resonance image), radiosensitivity of tumor histology, mechanical instability by Spine Instability Neoplastic Score (sagittal and axial computed tomography image), and systemic disease should be evaluated by a multidisciplinary team. This review article summarizes the role of imaging for diagnosis and treatment of spinal bone tumors.