Artificial intelligence based intervertebral discs grading: From unsupervised detection to explainable Pfirrmann classification

This study aimed to explore the feasibility of the fractal method used in decoding disk heterogeneity, hoping to find a reliable imaging biomarker for the quantitative and continuous grading of intervertebral disks (IVDs). Totally, 180 IVDs in 65 low back pain patients (29 males, 36 female, 28-69years) were examined with MRI. Each IVD was manually segmented on axial slice (at the mid-height layer of the disk). All disks were visually evaluated regarding degeneration grade, using Pfirrmann classification, by two experienced radiologists. Fractal dimension (FD) of the IVD was calculated from the defined regions of interest and correlated with Pfirrmann grade. Fractal dimension differed significantly between any two groups (P < 0.01). The mean FDs for the four grades were as follows: Pfirrmann 1: 1.13 ± 0.02; Pfirrmann 2: 1.30 ± 0.05; Pfirrmann 3: 1.50 ± 0.05; and Pfirrmann 4: 1.65 ± 0.02. The well-hydrated IVDs displayed low fractal dimension. Degenerated IVDs displayed increased fractal dimension caused by disk heterogeneity, where the fractal dimension was shown to correlate strongly with Pfirrmann grade. Fractal dimension associated well with IVD degeneration, determined with Pfirrmann grading, suggesting that the IVD fractal analysis was a suitable detection tool for the objective and continuous classification of IVD degeneration. These slides can be retrieved under Electronic Supplementary Material.

The Pfirrmann classification is a commonly utilized system for assessing lumbar disc degeneration, primarily concentrating on the hydration of the nucleus pulposus (NP). However, it overlooks the significance of annulus fibrosus (AF) tears. This study aims to challenge the conventional perspective by reinterpreting the magnetic resonance imaging (MRI)-detected "horizon gray band" as a normal anatomical characteristic of condensed NP cells instead of a sign of degeneration. To fill this gap and improve clinical applicability, we propose a new grading system that emphasizes the integrity of the AF. We carefully assessed intervertebral disc (IVD) structures in different species, including human, with MRI examination and histological analysis. The "horizon gray band" and surrounding "high signal" area were isolated from goat and evaluated by proteomic analysis to identify respective compositions. We examined the volume of each compartment of IVDs with PACS software in 224 Chinese patients (144 healthy individuals and 80 with IVDD) using T2 MRI and factors influencing IVD volume were analyzed. A new grading system for lumbar disc degeneration, incorporating AF tears, was developed. The reliability of this grading system was tested on MRI scans of 500 lumbar IVDs from 100 low back pain patients, with three independent observers. Intra - and interobserver reliabilities were assessed using kappa statistics. The clinical symptoms and prognosis of 100 patients with this grading system were analyzed through 1 year of follow-up. The result of T2 mapping MRI is that each NP consists of central "horizon gray band" and surrounding "high signal" area. The histology and proteomic revealed the tissues from "horizon band" area represent NP cell and "high signal" represents water-rich NP cell-extracellular matrix (ECM) tissues. The NP cell volumes across various lumbar segments (L1-L2 to L4-L5) were 2488.2, 3238.44, 3638.34, 3752.83, and 2997.02mm 3 , respectively. The ratio of NP cell volume to total IVD volume was consistent across segments (0.167-0.184), with no significant variation. Further analysis found that NP cell volume, NP volume (NP cell and surrounding water-rich ECM), and IVD volume were correlated with segmental position, as well as with height, weight, and age. Using the new grading system, we identified 62% Grade I discs, 6% Grade IIA discs , 5% Grade IIB discs , 18% Grade III discs , and 9% Grade IV discs when examined the patients with low back pain. Intra-observer and inter-observer agreements were substantial to excellent, with complete agreement in 91.8%-93.2% of the 500 discs. The new grading system is closely related with patient' symptom and prognosis. T2-weighted imaging identified the "horizon gray band" as normal NP cell condense, not degeneration signal. Our findings redefine the anatomy of the IVD, statistically analyze its component volumes, and introduce a classification system that can be used for surgical purposes. This study is in line with recent translational innovations in spine care, effectively bridging the gap between MRI biomarkers and clinical practice.

Degenerative disc disease is the most common cause of low back pain. Intervertebral disc abnormalities are commonly evaluated by magnetic resonance imaging (MRI), and Pfirrmann’s system involves the use of T2-weighted images (T2WI) to classify disc degeneration. However, as this classification is based on visual evaluation, it is not possible to quantify degeneration using this method. The present study was performed to establish an MRI-based intervertebral disc classification system using diffusional kurtosis imaging (DKI), to quantify intervertebral disc water content according to the Pfirrmann classification. Sagittal mean diffusional kurtosis (MK) mapping was performed for the L3/4, L4/5, and L5/S1 intervertebral discs in 32 patients (15 female, 17 male; age range, 24 - 82 years; mean age, 57.7 years). The degree of disc degeneration was assessed in the midsagittal section on T2WI according to the Pfirrmann classification (grade I - V). The relationships between MK values, which are correlated with intervertebral disc composition changes, and grade of degeneration determined using the Pfirrmann classification were analyzed. The MK values tended to decrease with increasing grade of degeneration, and differed significantly between grades I and IV, but not between grade IV and V (P < 0.05, Mann-Whitney U test). DKI is an effective means of detecting the early stages of disc degeneration. Therefore, DKI may be a useful diagnostic tool for quantitative assessment of intervertebral disc degeneration.

OBJECTIVE: To evaluate the relationship of sex and age with Modic and Pfirrmann classifications and verify the relationship between the two classifications in the group of patients studied. METHODS: 300 magnetic resonance scans (MRI) of the lumbar segment of the patients were evaluated; each lumbar segment (L1-L2, L2-L3, L3-L4, L4-L5 and L5-S1) was assessed according to Modic and Pfirrmann classifications. RESULTS: The type III of Pfirrmann was the most prevalent, whereas Modic was absent on most levels analyzed. The Modic signal was present mainly in the lower levels (L4-L5 and L5-S1) and in females. CONCLUSION: There is a strong correlation between the changes of the endplate and intervertebral discs, evidenced by changes in MRI. The relationship between Pfirrmann and Modic classifications is statistically higher the greater the disc degeneration.

This study aimed to develop a graph neural network (GNN) for automated three-dimensional (3D) magnetic resonance imaging (MRI) visualization and Pfirrmann grading of intervertebral discs (IVDs), and benchmark it against manual classifications. Lumbar IVD MRI data from 300 patients were retrospectively analyzed. Two clinicians assessed the manual segmentation and grading for inter-rater reliability using Cohen's kappa. The IVDs were then processed and classified using an automated convolutional neural network (CNN)–GNN pipeline, and their performance was evaluated using F1 scores. Manual Pfirrmann grading exhibited moderate agreement (κ = 0.455–0.565) among the clinicians, with higher exact match frequencies at lower lumbar levels. Single-grade discrepancies were prevalent except at L5/S1. Automated segmentation of IVDs using a pretrained U-Net model achieved an F1 score of 0.85, with a precision and recall of 0.83 and 0.88, respectively. Following 3D reconstruction of the automatically segmented IVD into a 3D point-cloud representation of the target intervertebral disc, the GNN model demonstrated moderate performance in Pfirrmann classification. The highest precision (0.81) and F1 score (0.71) were observed at L2/3, whereas the overall metrics indicated moderate performance (precision: 0.46, recall: 0.47, and F1 score: 0.46), with variability across spinal levels. The integration of CNN and GNN offers a new perspective for automating IVD analysis in MRI. Although the current performance highlights the need for further refinement, the moderate accuracy of the model, combined with its 3D visualization capabilities, establishes a promising foundation for more advanced grading systems.

Comparison of Pfirrmann classification and objective T2 signal intensity of cervical disc-cisterna magna ratio measurements in cervical intervertebral disc degeneration

Spondylolysis is a common entity, a minority of people affected by this disease need medical care, and only a few require surgery. Reconstruction of the pars interarticularis is an interesting alternative to segmental fusion; this technique has the advantage of preserving segmental motion. Most authors report good results for young patients without intervertebral disk or facet degenerative changes. Moreover Louis also showed good to excellent results with his technique carried out among people who presented a satisfactory disk height (equal to two thirds of normal height). This could extend the number of patients for whom pars interarticularis repair could be proposed. In this study, the limit of reconstruction was set at grade 3 of the Pfirrmann's classification. The fixation of the isthmus was done with a new kind of pedicle screw hook system. This stable and strong device is easy to use, allows an anatomic pars interarticularis reconstruction of spondylolysis and avoids a postoperative bracing. Twenty-three patients were assessed in this study, the mean age at operation was 34 (range 16-52 years) and the average follow-up was for 59 months (range 6-113 months). Eight patients showed moderate degenerative disk disease before the surgery and 12 patients had a grade 1 spondylolisthesis. The visual analogical scale, the Oswestry disability index (ODI) and the modified Prolo score were used for assessment of pain and clinical outcome before and after surgery. The results were from "excellent" to "good" for twenty patients (87%) and "fair" for three of them (13%). The consolidation of the isthmus was assessed at the end of the study (CT-scan); the fusion rate was observed in 91%. Among patients aged less than 30 years results are from "good", to "excellent" in all cases and consolidation was always observed. All of them showed normal disc signal before the surgery. In the group aged more than 30 years, the results varied from "good" to "excellent' in 73% and fusion of the defect was discovered in 82% of cases. Eight of them (73%) had moderate disk signal modification before the surgery. All people with fair results displayed moderate disk degeneration signs at MRI before surgery; but two of those three patients had a failure of defect consolidation too and it is also associated with poor results by several authors. No complication was found in this series. According to the good results reported by Louis and upto the current finding, the authors believe that pars interarticularis repair can be carried out on patients with moderate degenerative disk disease; the stage 3 of Pfirrmann's classification seems a good limit. The Bone and joint research (B.J.R. system) is readily usable by any surgeon using pedicle screw systems and having a short learning curve. No device failure has been observed in this series.

Abdominal aortic calcification (AAC) is associated with lower back pain, reduced bone mineral density of the spine. Vascular changes could also affect the already sparsely perfused intervertebral endplate and intervertebral disc. Lumbar MRIs and lateral radiographs of patients with lower back pain were retrospectively analyzed. AAC was assessed on lateral lumbar radiographs according to the Kauppila score, with a maximum score of 24. Patients were grouped into no (AAC = 0), moderate (AAC 1 to ≤ 4), and severe AAC (AAC ≥ 5). Endplate and disc degeneration were classified according to the total endplate score (TEPS) and Pfirrmann classification. The associations between AAC and degenerative changes was analyzed with a generalized mixed model and was adjusted for age, sex, body mass index as well as diabetes mellitus, and smoking status. A total of 217 patients (47.9% female) were included in the analysis, totaling 1085 intervertebral levels. Of those, 45 (20.7%) patients had moderate, and 39 (18%) had severe AAC. The results of the generalized mixed model showed no significant association between AAC and disc degeneration (p > 0.05). In contrast, a significant positive association between AAC and the severity of TEPS (β: 0.51, 95% CI: 1.92-2.12, p = 0.004) was observed in the multivariable analysis. This study demonstrates an independent association between AAC and endplate degeneration. These findings expand our knowledge about the degenerative cascade of the lumbar spine and suggest that AAC might be a modifiable risk factor for endplate changes.

ObjectivesTo evaluate the changes of Hounsfield units (HU) value in different types of Modic changes (MCs) and to analyze the correlation between the change of HU value and area ratio of MCs region, bone mineral density (BMD), and degree of intervertebral disc degeneration.MethodsOne hundred fifty-eight endplates with MCs were included and analyzed. HU values of MCs regions and adjacent vertebral corresponding regions without MCs were measured. The area ratio of MCs region was defined as the area of MCs divided by the area of endplate or the vertebral sagittal plane. BMD was measured by Dual-energy x-ray absorptiometry (DXA). Degree of intervertebral disc degeneration was evaluated based on Pfirrmann classification. According to the types of variables, descriptive statistics, Kolmogorove-Smirnov test, paired t-test, Wilcoxon signed-rank test, Independent-Samples T Test, and Pearson correlation analysis were used.ResultsThe HU values in any types of MCs are significantly higher than that of adjacent vertebral corresponding regions without MCs (P < 0.001). The HU value of the type III MCs is higher than that of the type I and type II MCs. HU value was positively correlated with BMD. In the levels with Grade V disc degeneration, the area ratio of MCs region was significant increased.ConclusionsHU values of the vertebral endplate and bone marrow were increased in most MCs regions with all types of MCs. HU value of endplates had a significantly positive correlation with BMD. Higher area ratio of MCs region is associated with more severe intervertebral disc degeneration.

Background: Lumbar prolapsed intervertebral disc (PIVD) is a common musculoskeletal disorder that is routinely diagnosed clinically and confirmed by magnetic resonance imaging (MRI). Although there have been various artificial intelligence (AI) models studied as diagnostic options, they are mostly nonspecific for lumbar PIVD and do not incorporate clinical and physical examination parameters, which may contribute to misdiagnosis. This planned study aims to develop an AI-driven diagnosis tool using clinical and physical parameters and to compare its findings with MRI-based diagnoses.Methods: The research involves two concurrent phases. Phase one includes finding applicable clinical and physical diagnostic parameters from interviews with patients, literature, and expert questionnaires. These parameters are to be used to construct a dataset from chosen participants for developing the AI model. In phase two, there will be a prospective cross-sectional study comparing the diagnostic output of the AI tool with MRI results. The performance of the AI model will be measured using standard metrics and iteratively optimized according to the results.Results: Phase one started in October 2023 to determine diagnostic parameters. AI development data collection started in February 2024 and took 12–14 months. The future cross-sectional validation phase will follow the development of the AI model. Final study findings are expected by mid-2025.Conclusion: Upon the development of an efficient AI tool, the model can serve as a diagnostic substitute for MRI in diagnosing lumbar PIVD, thereby improving accessibility and reducing diagnostic delays, particularly in areas with limited imaging capacity.

Native vertebral osteomyelitis, also termed spondylodiscitis, is an antibiotic-resistant disease that requires long-term treatment. Without proper treatment, NVO can lead to severe nerve damage or even death. Therefore, it is important to accurately diagnose the cause of NVO, especially in spontaneous cases. Infectious NVO is characterized by the involvement of 2 adjacent vertebrae and intervertebral discs, and common infectious agents include Staphylococcus aureus, Mycobacterium tuberculosis, Brucella abortus, and fungi. Clinical symptoms are generally nonspecific, and early diagnosis and appropriate treatment can prevent irreversible sequelae. Advances in pathologic histologic imaging have led physicians to look more forward to being able to differentiate between tuberculous and septic spinal discitis. Therefore, research in identifying and differentiating the imaging features of these 4 common NVOs is essential. Due to the diagnostic difficulties, clinical and radiologic diagnosis is the mainstay of provisional diagnosis. With the advent of the big data era and the emergence of convolutional neural network algorithms for deep learning, the application of artificial intelligence (AI) technology in orthopedic imaging diagnosis has gradually increased. AI can assist physicians in imaging review, effectively reduce the workload of physicians, and improve diagnostic accuracy. Therefore, it is necessary to present the latest clinical research on NVO and the outlook for future AI applications.

12. Assessment of age-related intervertebral disc degenerations of the cervical spine in different age groups: a kinematic MRI study

This study aimed to evaluate the clinical applicability of 23Na-MRI for the quantitative assessment of lumbar intervertebral disc degeneration (IDD) by comparing 23Na signal intensities with T2-weighted imaging (T2WI), Pfirrmann classification, and clinical findings. A phantom experiment using NaCl solutions was conducted to validate the relationship between 23Na signal intensity and sodium concentration. Lumbar 23Na-MRI and T2WI were performed on 40 patients. Apparent tissue sodium concentration (aTSC) was calculated using cerebrospinal fluid (CSF) as a reference. Correlations between aTSC and T2 ratio, disc thickness, and Pfirrmann grade were assessed using linear mixed-effects models. Associations between mean aTSC across three discs and age and body mass index (BMI) were evaluated using Pearson's correlation, while differences by self-reported low back pain severity were evaluated using Welch's ANOVA. Differences across Pfirrmann grades were tested using Wald tests with FDR-corrected pairwise models. The phantom experiment demonstrated a strong linear relationship between sodium concentration and 23Na signal intensity (R2 = 0.996). In the clinical study, aTSC was significantly associated with T2 ratio (β = 154.5, P < 0.001) and disc thickness (β = 3.38, P = 0.020), inversely correlated with age (r = -0.630, P < 0.001), and not significantly correlated with BMI. aTSC significantly differed across Pfirrmann grades (P < 0.001). Patients with severe back pain had significantly lower aTSC than those without pain (P = 0.005). 23Na-MRI enables quantitative, clinically feasible evaluation of lumbar IDD and may serve as a non-invasive biomarker for disc degeneration severity.

Outcome of Anterior Lumbar Interbody Fusion: A Retrospective Study of Clinical and Radiologic Parameters.

BackgroundThere is a need to assess how commonly used classification systems of intervertebral disc degeneration reflect the compromise of neural elements. This study aims to explore the relationship between lumbar discs degenerative diseases using the Pfirrmann and the Combined Task Forces (CTF) of the North American Spine Society (NASS) grading systems as well as qualitative and quantitative grades of lumbar spinal stenosis. This retrospective cohort study included adult patients undergoing non-contrast MR imaging of lumbosacral spine. The radiological assessment included the Pfirrmann grading system, Van Rijn classification, Combined Task Force (CTF) classification, measurement of the cross-sectional area of the dural sac, mid-sagittal antero-posterior diameter of the thecal sac, the degree of dural sac compression at disc level, lateral recesses heights, and intervertebral foramina diameters. The degree of stenosis of the spinal canal and intervertebral foramina was assessed.ResultsOne hundred patients were included in the study. At all levels, Pfirrmann grades had a moderate, significant, positive correlation with the severity of stenosis of the central and lateral spinal canals as well as foraminal stenosis. The grades of lumbar spinal canal and foraminal stenosis had a significant positive correlation with degree of disc displacement as assessed by CTF classification and had a significant negative correlation with the quantitative lumbar spinal canal and foraminal measures.ConclusionsThere is a good correlation between Pfirrmann classification, CTF classification of NASS, qualitative grading and quantitative measures of lumbar spinal canal that reflects the severity of lumbar spinal canal stenosis and nerve root compression.