Could automated machine-learned MRI grading aid epidemiological studies of lumbar spinal stenosis? Validation within the Wakayama spine study

AI-Based Measurement of Lumbar Spinal Stenosis on MRI: External Evaluation of a Fully Automated Model

OBJECTIVES

The aim of this study was to clinically validate a fully automated AI model for magnetic resonance imaging (MRI)-based quantifications of lumbar spinal canal stenosis.

MATERIALS AND METHODS

This retrospective study included lumbar spine MRI of 100 consecutive clinical patients (56 ± 17 years; 43 females, 57 males) performed on clinical 1.5 (51 examinations) and 3 T MRI scanners (49 examinations) with heterogeneous clinical imaging protocols. The AI model performed segmentations of the thecal sac on axial T2-weighted sequences. Based on these segmentations, the anteroposterior (AP) and mediolateral (ML) distance, and the area of the thecal sac were measured in a fully automated manner. For comparison, 2 fellowship-trained musculoskeletal radiologists performed the same segmentations and measurements independently. Statistics included 1-sample t tests, the intraclass correlation coefficient (ICC), Bland-Altman plots, and Dice coefficients. A P value of <0.05 was considered statistically significant.

RESULTS

The average measurements of the AI model, reader 1, and reader 2 were 194 ± 72 mm 2 , 181 ± 71 mm 2 , and 179 ± 70 mm 2 for thecal sac area, 13 ± 3.3 mm, 12.6 ± 3.3 mm, and 12.6 ± 3.2 mm for AP distance, and 19.5 ± 3.9 mm, 20 ± 4.3 mm, and 19.4 ± 4 mm for ML distance, respectively. Significant differences existed for all pairwise comparisons, besides reader 1 versus AI model for the ML distance and reader 1 versus reader 2 for the AP distance ( P = 0.1 and P = 0.21, respectively). The pairwise mean absolute errors among reader 1, reader 2, and the AI model ranged from 0.59 mm and 0.75 mm for the AP distance, from 1.16 mm to 1.37 mm for the ML distance, and from 7.9 mm 2 to 15.54 mm 2 for the thecal sac area. Pairwise ICCs among reader 1, reader 2, and the AI model ranged from 0.91 and 0.94 for the AP distance and from 0.86 to 0.9 for the ML distance without significant differences. For the thecal sac area, the pairwise ICC between both readers and the AI model of 0.97 each was slightly, but significantly lower than the ICC between reader 1 and reader 2 of 0.99. Similarly, the Dice coefficient and Hausdorff distance between both readers and the AI model were significantly lower than the values between reader 1 and reader 2, overall ranging from 0.93 to 0.95 for the Dice coefficients and 1.1 to 1.44 for the Hausdorff distances.

CONCLUSIONS

The investigated AI model is reliable for assessing the AP and the ML thecal sac diameters with human level accuracies. The small differences for measurement and segmentation of the thecal sac area between the AI model and the radiologists are likely within a clinically acceptable range.

Practical Applications of Artificial Intelligence in Spine Imaging: A Review

Artificial intelligence (AI), a transformative technology with unprecedented potential in medical imaging, can be applied to various spinal pathologies. AI-based approaches may improve imaging efficiency, diagnostic accuracy, and interpretation, which is essential for positive patient outcomes. This review explores AI algorithms, techniques, and applications in spine imaging, highlighting diagnostic impact and challenges with future directions for integrating AI into spine imaging workflow.

Understanding the role and adoption of artificial intelligence techniques in rheumatology research: An in-depth review of the literature

The major and upward trend in the number of published research related to rheumatic and musculoskeletal diseases, in which artificial intelligence plays a key role, has exhibited the interest of rheumatology researchers in using these techniques to answer their research questions. In this review, we analyse the original research articles that combine both worlds in a five- year period (2017-2021). In contrast to other published papers on the same topic, we first studied the review and recommendation articles that were published during that period, including up to October 2022, as well as the publication trends. Secondly, we review the published research articles and classify them into one of the following categories: disease identification and prediction, disease classification, patient stratification and disease subtype identification, disease progression and activity, treatment response, and predictors of outcomes. Thirdly, we provide a table with illustrative studies in which artificial intelligence techniques have played a central role in more than twenty rheumatic and musculoskeletal diseases. Finally, the findings of the research articles, in terms of disease and/or data science techniques employed, are highlighted in a discussion. Therefore, the present review aims to characterise how researchers are applying data science techniques in the rheumatology medical field. The most immediate conclusions that can be drawn from this work are: multiple and novel data science techniques have been used in a wide range of rheumatic and musculoskeletal diseases including rare diseases; the sample size and the data type used are heterogeneous, and new technical approaches are expected to arrive in the short-middle term.

Automatic classification of the vertebral endplate lesions in magnetic resonance imaging by deep learning model

Introduction

A novel classification scheme for endplate lesions, based on T2-weighted images from magnetic resonance imaging (MRI) scan, has been recently introduced and validated. The scheme categorizes intervertebral spaces as "normal," "wavy/irregular," "notched," and "Schmorl's node." These lesions have been associated with spinal pathologies, including disc degeneration and low back pain. The exploitation of an automatic tool for the detection of the lesions would facilitate clinical practice by reducing the workload and the diagnosis time. The present work exploits a deep learning application based on convolutional neural networks to automatically classify the type of lesion.

Methods

T2-weighted MRI scans of the sagittal lumbosacral spine of consecutive patients were retrospectively collected. The middle slice of each scan was manually processed to identify the intervertebral spaces from L1L2 to L5S1, and the corresponding lesion type was labeled. A total of 1,559 gradable discs were obtained, with the following types of distribution: "normal" (567 discs), "wavy/irregular" (485), "notched" (362), and "Schmorl's node" (145). The dataset was divided randomly into a training set and a validation set while preserving the original distribution of lesion types in each set. A pretrained network for image classification was utilized, and fine-tuning was performed using the training set. The retrained net was then applied to the validation set to evaluate the overall accuracy and accuracy for each specific lesion type.

Results

The overall rate of accuracy was found equal to 88%. The accuracy for the specific lesion type was found as follows: 91% (normal), 82% (wavy/irregular), 93% (notched), and 83% (Schmorl's node).

Discussion

The results indicate that the deep learning approach achieved high accuracy for both overall classification and individual lesion types. In clinical applications, this implementation could be employed as part of an automatic detection tool for pathological conditions characterized by the presence of endplate lesions, such as spinal osteochondrosis.

Convolutional Neural Networks in Spinal Magnetic Resonance Imaging: A Systematic Review

OBJECTIVE

Convolutional neural networks (CNNs) are being increasingly used in the medical field, especially for image recognition in high-resolution, large-volume data sets. The study represents the current state of research on the application of CNNs in image segmentation and pathology detection in spine magnetic resonance imaging.

METHODS

For this systematic literature review, the authors performed a systematic initial search of the PubMed/Medline and Web of Science (Core collection) databases for eligible investigations. The authors limited the search to observational studies. Outcome parameters were analyzed according to the inclusion criteria and assigned to 3 groups: 1) segmentation of anatomical structures, 2) segmentation and evaluation of pathologic structures, and 3) specific implementation of CNNs.

RESULTS

Twenty-four retrospectively designed articles met the inclusion criteria. Publication dates ranged from 2017 to 2021. In total, 14,065 patients with 113,110 analyzed images were included. Most authors trained their network with a training-to-testing ratio of 80/20, while all but 2 articles used 5- to 10-fold cross-validation. Nine articles compared their performance results with other neural networks and algorithms, and all 24 articles described outcomes as positive.

CONCLUSIONS

State-of-the-art CNNs can detect and segment-specific anatomical landmarks and pathologies across a wide range, comparable to the skills of radiologists and experienced clinicians. With rapidly evolving network architectures and growing medical image databases, the future is likely to show growth in the development and refinement of these capable networks. However, the aid of automated segmentation and classification by neural networks cannot and should not be expected to replace clinical experts.

External validation of the deep learning system "SpineNet" for grading radiological features of degeneration on MRIs of the lumbar spine

BACKGROUND

Magnetic resonance imaging (MRI) is used to detect degenerative changes of the lumbar spine. SpineNet (SN), a computer vision-based system, performs an automated analysis of degenerative features in MRI scans aiming to provide high accuracy, consistency and objectivity. This study evaluated SN's ratings compared with those of an expert radiologist.

METHOD

MRIs of 882 patients (mean age, 72 ± 8.8 years) with degenerative spinal disorders from two previous trials carried out in our spine center between 2011 and 2019, were analyzed by an expert radiologist. Lumbar segments (L1/2-L5/S1) were graded for Pfirrmann Grades (PG), Spondylolisthesis (SL) and Central Canal Stenosis (CCS). SN's analysis for the equivalent parameters was generated. Agreement between methods was analyzed using kappa (κ), Spearman correlation (ρ) and Lin's concordance correlation (ρ_c) coefficients and class average accuracy (CAA).

RESULTS

4410 lumbar segments were analyzed. κ statistics showed moderate to substantial agreement in PG between the radiologist and SN depending on spinal level (range κ 0.63-0.77, all levels together 0.72; range CAA 45-68%, all levels 55%), slight to substantial agreement for SL (range κ 0.07-0.60, all levels 0.63; range CAA 47-57%, all levels 56%) and CCS (range κ 0.17-0.57, all levels 0.60; range CAA 35-41%, all levels 43%). SN tended to record more pathological features in PG than did the radiologist whereas the contrary was the case for CCS. SL showed an even distribution between methods.

CONCLUSION

SN is a robust and reliable tool with the ability to grade degenerative features such as PG, SL or CCS in lumbar MRIs with moderate to substantial agreement compared to the current gold-standard, the radiologist. It is a valuable alternative for analyzing MRIs from large cohorts for diagnostic and research purposes.

Improved Productivity Using Deep Learning-assisted Reporting for Lumbar Spine MRI

Background Lumbar spine MRI studies are widely used for back pain assessment. Interpretation involves grading lumbar spinal stenosis, which is repetitive and time consuming. Deep learning (DL) could provide faster and more consistent interpretation. Purpose To assess the speed and interobserver agreement of radiologists for reporting lumbar spinal stenosis with and without DL assistance. Materials and Methods In this retrospective study, a DL model designed to assist radiologists in the interpretation of spinal canal, lateral recess, and neural foraminal stenoses on lumbar spine MRI scans was used. Randomly selected lumbar spine MRI studies obtained in patients with back pain who were 18 years and older over a 3-year period, from September 2015 to September 2018, were included in an internal test data set. Studies with instrumentation and scoliosis were excluded. Eight radiologists, each with 2-13 years of experience in spine MRI interpretation, reviewed studies with and without DL model assistance with a 1-month washout period. Time to diagnosis (in seconds) and interobserver agreement (using Gwet κ) were assessed for stenosis grading for each radiologist with and without the DL model and compared with test data set labels provided by an external musculoskeletal radiologist (with 32 years of experience) as the reference standard. Results Overall, 444 images in 25 patients (mean age, 51 years ± 20 [SD]; 14 women) were evaluated in a test data set. DL-assisted radiologists had a reduced interpretation time per spine MRI study, from a mean of 124-274 seconds (SD, 25-88 seconds) to 47-71 seconds (SD, 24-29 seconds) (P < .001). DL-assisted radiologists had either superior or equivalent interobserver agreement for all stenosis gradings compared with unassisted radiologists. DL-assisted general and in-training radiologists improved their interobserver agreement for four-class neural foraminal stenosis, with κ values of 0.71 and 0.70 (with DL) versus 0.39 and 0.39 (without DL), respectively (both P < .001). Conclusion Radiologists who were assisted by deep learning for interpretation of lumbar spinal stenosis on MRI scans showed a marked reduction in reporting time and superior or equivalent interobserver agreement for all stenosis gradings compared with radiologists who were unassisted by deep learning. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Hayashi in this issue.

Severity and location of lumbar spine stenosis affects the outcome of total knee arthroplasty

Background

Recent studies have noted that patients with pre-existing lumbar spinal stenosis (LSS) have lower functional outcomes after total knee arthroplasty (TKA). Given that LSS manifests heterogeneously in location and severity, its influence on knee replacement merits a radiographically targeted analysis. We hypothesize that patients with more severe LSS will have diminished knee mobility before and after TKA.

Methods

This retrospective case series assessed all TKAs performed at our institution for primary osteoarthritis from 2017–2020. Preoperative lumbar magnetic resonance image (MRI) with no prior lumbar spine surgery was necessary for inclusion. Stenosis severity was demonstrated by (1) anterior–posterior (AP) diameter of the thecal sac and (2) morphological grade. TKA outcomes in 103 cases (94 patients) were assessed by measuring preoperative and postoperative arc of motion (AOM), postoperative flexion contracture, and need for manipulation under anesthesia.

Results

Patients with mild stenosis did significantly better in terms of postoperative knee AOM. As AP diameter decreased at levels L1–2, L2–3, L3–4, and L4–5, there was a significant reduction in preoperative-AOM (p < 0.001 for each), with a 16 degree decrease when using patients’ most stenotic level (p < 0.001). The same was noted with respect to increased morphological grade (p < 0.001), with a 5 degree decrease for patients’ most stenotic level (p < 0.001).

Conclusion

Severe LSS, which is readily demonstrated by a reduction in the AP diameter of the thecal sac or increased morphological grade on MRI, correlated with a significant reduction in preoperative AOM that was not improved after TKA. Persistent postoperative reductions in AOM may contribute to reduced patient satisfaction and recovery.

Level of evidence: Level 4

Development of a standardized histopathology scoring system for human intervertebral disc degeneration: an Orthopaedic Research Society Spine Section Initiative

BACKGROUND

Histopathological analysis of intervertebral disc (IVD) tissues is a critical domain of back pain research. Identification, description, and classification of attributes that distinguish abnormal tissues form a basis for probing disease mechanisms and conceiving novel therapies. Unfortunately, lack of standardized methods and nomenclature can limit comparisons of results across studies and prevent organizing information into a clear representation of the hierarchical, spatial, and temporal patterns of IVD degeneration. Thus, the following Orthopaedic Research Society (ORS) Spine Section Initiative aimed to develop a standardized histopathology scoring scheme for human IVD degeneration.

METHODS

Guided by a working group of experts, this prospective process entailed a series of stages that consisted of reviewing and assessing past grading schemes, surveying IVD researchers globally on current practice and recommendations for a new grading system, utilizing expert opinion a taxonomy of histological grading was developed, and validation performed.

RESULTS

A standardized taxonomy was developed, which showed excellent intra-rater reliability for scoring nucleus pulposus (NP), annulus fibrosus (AF), and cartilaginous end plate (CEP) regions (interclass correlation [ICC] > .89). The ability to reliably detect subtle changes varied by IVD region, being poorest in the NP (ICC: .89-.95) where changes at the cellular level were important, vs the AF (ICC: .93-.98), CEP (ICC: .97-.98), and boney end plate (ICC: .96-.99) where matrix and structural changes varied more dramatically with degeneration.

CONCLUSIONS

The proposed grading system incorporates more comprehensive descriptions of degenerative features for all the IVD sub-tissues than prior criteria. While there was excellent reliability, our results reinforce the need for improved training, particularly for novice raters. Future evaluation of the proposed system in real-world settings (eg, at the microscope) will be needed to further refine criteria and more fully evaluate utility. This improved taxonomy could aid in the understanding of IVD degeneration phenotypes and their association with back pain.

Deep Learning Model for Automated Detection and Classification of Central Canal, Lateral Recess, and Neural Foraminal Stenosis at Lumbar Spine MRI

Background Assessment of lumbar spinal stenosis at MRI is repetitive and time consuming. Deep learning (DL) could improve -productivity and the consistency of reporting. Purpose To develop a DL model for automated detection and classification of lumbar central canal, lateral recess, and neural -foraminal stenosis. Materials and Methods In this retrospective study, lumbar spine MRI scans obtained from September 2015 to September 2018 were included. Studies of patients with spinal instrumentation or studies with suboptimal image quality, as well as postgadolinium studies and studies of patients with scoliosis, were excluded. Axial T2-weighted and sagittal T1-weighted images were used. Studies were split into an internal training set (80%), validation set (9%), and test set (11%). Training data were labeled by four radiologists using predefined gradings (normal, mild, moderate, and severe). A two-component DL model was developed. First, a convolutional neural network (CNN) was trained to detect the region of interest (ROI), with a second CNN for classification. An internal test set was labeled by a musculoskeletal radiologist with 31 years of experience (reference standard) and two subspecialist radiologists (radiologist 1: A.M., 5 years of experience; radiologist 2: J.T.P.D.H., 9 years of experience). DL model performance on an external test set was evaluated. Detection recall (in percentage), interrater agreement (Gwet κ), sensitivity, and specificity were calculated. Results Overall, 446 MRI lumbar spine studies were analyzed (446 patients; mean age ± standard deviation, 52 years ± 19; 240 women), with 396 patients in the training (80%) and validation (9%) sets and 50 (11%) in the internal test set. For internal testing, DL model and radiologist central canal recall were greater than 99%, with reduced neural foramina recall for the DL model (84.5%) and radiologist 1 (83.9%) compared with radiologist 2 (97.1%) (P < .001). For internal testing, dichotomous classification (normal or mild vs moderate or severe) showed almost-perfect agreement for both radiologists and the DL model, with respective κ values of 0.98, 0.98, and 0.96 for the central canal; 0.92, 0.95, and 0.92 for lateral recesses; and 0.94, 0.95, and 0.89 for neural foramina (P < .001). External testing with 100 MRI scans of lumbar spines showed almost perfect agreement for the DL model for dichotomous classification of all ROIs (κ, 0.95-0.96; P < .001). Conclusion A deep learning model showed comparable agreement with subspecialist radiologists for detection and classification of central canal and lateral recess stenosis, with slightly lower agreement for neural foraminal stenosis at lumbar spine MRI. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Hayashi in this issue.

Augmented Reality, Virtual Reality and Artificial Intelligence in Orthopedic Surgery: A Systematic Review

Background: The application of virtual and augmented reality technologies to orthopaedic surgery training and practice aims to increase the safety and accuracy of procedures and reducing complications and costs. The purpose of this systematic review is to summarise the present literature on this topic while providing a detailed analysis of current flaws and benefits. Methods: A comprehensive search on the PubMed, Cochrane, CINAHL, and Embase database was conducted from inception to February 2021. The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were used to improve the reporting of the review. The Cochrane Risk of Bias Tool and the Methodological Index for Non-Randomized Studies (MINORS) was used to assess the quality and potential bias of the included randomized and non-randomized control trials, respectively. Results: Virtual reality has been proven revolutionary for both resident training and preoperative planning. Thanks to augmented reality, orthopaedic surgeons could carry out procedures faster and more accurately, improving overall safety. Artificial intelligence (AI) is a promising technology with limitless potential, but, nowadays, its use in orthopaedic surgery is limited to preoperative diagnosis. Conclusions: Extended reality technologies have the potential to reform orthopaedic training and practice, providing an opportunity for unidirectional growth towards a patient-centred approach.

Advancing frontiers in rheumatic and musculoskeletal imaging

In recent years, technological improvements allowed imaging modalities to become increasingly essential in achieving early and precise diagnoses in the field of rheumatic and musculoskeletal diseases (RMDs). To date, imaging examinations are routinely used in all steps of diagnostic and therapeutic care pathways of patients affected by RMDs. The articles published in this Article Collection clearly show the efforts of researchers to find innovative applications of musculoskeletal imaging in clinical practice and to face the crucial challenges that remain in the interpretation and quality control of images. Highly performing diagnostic technologies are currently available to early diagnose and accurately monitor several musculoskeletal disorders, but also to guide personalized interventional therapeutic procedures tailored to the individual patients in the emerging process of precision medicine. Among these new modalities, some are particularly promising and thereby subject to several lines of research in RMDs, including SPECT-CT and dual-energy CT, MRI sequences, high and ultra-high frequency ultrasound with effective tools like shear wave elastography.

Accurate prediction of lumbar microdecompression level with an automated MRI grading system

OBJECTIVE

Lumbar spine MRI interpretations have high variability reducing utility for surgical planning. This study evaluated a convolutional neural network (CNN) framework that generates automated MRI grading for its ability to predict the level that was surgically decompressed.

MATERIALS AND METHODS

Patients who had single-level decompression were retrospectively evaluated. Sagittal T2 images were processed by a CNN (SpineNet), which provided grading for the following: central canal stenosis, disc narrowing, disc degeneration, spondylolisthesis, upper/lower endplate morphologic changes, and upper/lower marrow changes. The grades were used to calculate an aggregate score. The variables and the aggregate score were analyzed for their ability to predict the surgical level. For each surgical level subgroup, the surgical level aggregate scores were compared with the non-surgical levels.

RESULTS

A total of 141 patients met the inclusion criteria (82 women, 59 men; mean age 64 years; age range 28-89 years). SpineNet did not identify central canal stenosis in 32 patients. Of the remaining 109, 96 (88%) patients had a decompression at the level of greatest stenosis. The higher stenotic grade was present only at the surgical level in 82/96 (85%) patients. The level with the highest aggregate score matched the surgical level in 103/141 (73%) patients and was unique to the surgical level in 91/103 (88%) patients. Overall, the highest aggregate score identified the surgical level in 91/141 (65%) patients. The aggregate MRI score mean was significantly higher for the L3-S1 surgical levels.

CONCLUSION

A previously developed CNN framework accurately predicts the level of microdecompression for degenerative spinal stenosis in most patients.