Development of a Deep Learning Model for Diagnosing Lumbar Spinal Stenosis Based on CT Images.

STUDY DESIGN: Retrospective study. OBJECTIVES: This study aimed to develop an initial deep learning model based on CT scans for diagnosing lumbar spinal stenosis. SUMMARY OF BACKGROUND DATA: MRI is commonly used for diagnosing lumbar spinal stenosis due to its high soft tissue resolution, but CT is more portable, cost-effective, and has wider regional coverage. Using deep learning models to improve the accuracy of CT diagnosis can effectively reduce missed diagnoses and misdiagnoses in clinical practice. METHODS: Axial lumbar spine CT scans obtained between March 2022 and September 2023 were included. The dataset was divided into a training set (62.3%), a validation set (22.9%), and a control set (14.8%). All data were labeled by two spine surgeons using the widely accepted grading system for lumbar spinal stenosis. The training and validation sets were used to annotate the ROIs by the two spine surgeons. First, an ROI detection model and a CNN classifier were trained using the training set. After training, the model was preliminarily evaluated using a validation set. Finally, the performance of the deep learning model was evaluated on the control set, and a comparison was made between the model and classification performance of specialists with varying levels of experience. RESULTS: The central stenosis grading accuracies of DL Model Version 1 and DL Model Version 2 were 88% and 83%, respectively. The lateral recess grading accuracies of DL Model Version 1 and DL Model Version 2 were 75% and 71%, respectively. CONCLUSIONS: Our preliminarily developed deep learning system for assessing the degree of lumbar spinal stenosis in CT, including the central canal and lateral recess, has shown similar accuracy to experienced specialist physicians. This holds great value for further development and clinical application.

Complications and radiographic changes after implantation of interspinous process devices: average eight-year follow-up.

PURPOSE: This study aims to evaluate complications, clinical outcomes, and radiographic results following Coflex implantation. METHODS: We retrospectively studied 66 patients who had decompressive surgery combined with Coflex implantation to treat lumbar spinal stenosis. All imaging data were collected and examined for imaging changes. Clinical outcomes, included Oswestry Disability Index (ODI), back and leg visual analog scale (VAS) scores, were evaluated before surgery, six months after surgery and at the last follow-up. The number of complications occurring after five years of follow-up was counted. All reoperation cases were meticulously recorded. RESULTS: 66 patients were followed up for 5-14 years. The VAS and ODI scores were significantly improved compared with baseline. Heterotopic Ossification (HO) was detectable in 59 (89.4%). 26 (39.4%) patients had osteolysis at the contact site of Coflex with the spinous process. Coflex loosening was detected in 39 (60%) patients. Spinous process anastomosis was found in 34 (51.5%) patients. There was a statistically significant difference in the VAS score of back pain between patients with and without spinous process anastomosis. Nine cases of lumbar spinal restenosis were observed, and prosthesis fracture was observed in one case. CONCLUSION: Our study identified various imaging changes after Coflex implantation, and majority of them did not affect clinical outcomes. The majority of patients had HO, but osteolysis and Coflex loosening were relatively rare. The VAS score for back pain of these patients was higher if they have spinous process anastomosis. After five-year follow-up, we found lumbar spinal restenosis and prosthesis fracture cases.

Serum CXCL12 levels are associated with stroke severity and lesion volumes in stroke patients.

OBJECTIVE: Previous studies had shown that CXC chemokine ligand-12 (CXCL12) might play a significant role in stroke. The aim of this study was to test the serum baseline CXCL12 levels in Chinese patients with acute ischemic stroke (AIS). METHODS: All consecutive patients with first-ever AIS from January 2013 to June 2014 were recruited to participate in the study. CXCL12 and National Institutes of Health Stroke Scale were measured at the time of admission. Logistic regression analysis was used to evaluate the risk of stroke according to serum CXCL12 levels. Receiver operating characteristic (ROC) curve was used to evaluate the accuracy of serum CXCL12 in diagnosing stroke. RESULTS: From 306 screened patients, a total of 239 patients with first-ever AIS were included in this study. The results indicated that the serum CXCL12 levels were significantly higher in AIS patients as compared to normal controls (P < 0.0001). Serum CXCL12 were positively correlated with infarct volume(r = 0.307, P < 0.0001) and stroke severity(r = 0.288, P < 0.0001). After adjusting for all other possible covariates, CXCL12 was a stroke predictor with an adjusted OR of 2.047 [95% confidence interval (CI), 1.781-2.352; P < 0.0001]. Based on the ROC curve, the optimal cutoff value of serum CXCL12 levels as an indicator for auxiliary diagnosis of AIS was projected to be 3.4 ng/ml, which yielded a sensitivity of 87.9% and a specificity of 72.0%, with the area under the curve at 0.902 (95% CI, 0.875-0.929). CONCLUSION: Our study demonstrated that serum CXCL12 levels increased significantly following AIS, and these changes in serum CXCL12 were positively correlated with infarct volume and stroke severity in Chinese sample.