The interobserver reliability of the diagnosis and classification of scaphoid fractures using high-resolution peripheral quantitative CT

Deep Convolutional Neural Networks Provide Motion Grading for High-Resolution Peripheral Quantitative Computed Tomography of the Scaphoid

In vivo high-resolution peripheral quantitative computed tomography (HR-pQCT) studies on bone characteristics are limited, partly due to the lack of standardized and objective techniques to describe motion artifacts responsible for lower-quality images. This study investigates the ability of such deep-learning techniques to assess image quality in HR-pQCT datasets of human scaphoids. In total, 1451 stacks of 482 scaphoid images from 53 patients, each with up to six follow-ups within one year, and each with one non-displaced fractured and one contralateral intact scaphoid, were independently graded by three observers using a visual grading scale for motion artifacts. A 3D-CNN was used to assess image quality. The accuracy of the 3D-CNN to assess the image quality compared to the mean results of three skilled operators was between 92% and 96%. The 3D-CNN classifier reached an ROC-AUC score of 0.94. The average assessment time for one scaphoid was 2.5 s. This study demonstrates that a deep-learning approach for rating radiological image quality provides objective assessments of motion grading for the scaphoid with a high accuracy and a short assessment time. In the future, such a 3D-CNN approach can be used as a resource-saving and cost-effective tool to classify the image quality of HR-pQCT datasets in a reliable, reproducible and objective way.

What Is the Diagnostic Performance of Conventional Radiographs and Clinical Reassessment Compared With HR-pQCT Scaphoid Fracture Diagnosis?

BACKGROUND

Conventional radiographs and clinical reassessment are considered guides in managing clinically suspected scaphoid fractures. This is a unique study as it assessed the value of conventional radiographs and clinical reassessment in a cohort of patients, all of whom underwent additional imaging, regardless of the outcome of conventional radiographs and clinical reassessment.

QUESTIONS/PURPOSES

(1) What is the diagnostic performance of conventional radiographs in patients with a clinically suspected scaphoid fracture compared with high-resolution peripheral quantitative CT (HR-pQCT)? (2) What is the diagnostic performance of clinical reassessment in patients with a clinically suspected scaphoid fracture compared with HR-pQCT? (3) What is the diagnostic performance of conventional radiographs and clinical reassessment combined compared with HR-pQCT?

METHODS

Between December 2017 and October 2018, 162 patients with a clinically suspected scaphoid fracture presented to the emergency department (ED). Forty-six patients were excluded and another 25 were not willing or able to participate, which resulted in 91 included patients. All patients underwent conventional radiography in the ED and clinical reassessment 7 to 14 days later, together with CT and HR-pQCT. The diagnostic performance characteristics and accuracy of conventional radiographs and clinical reassessment were compared with those of HR-pQCT for the diagnosis of fractures since this was proven to be superior to CT scaphoid fracture detection. The cohort included 45 men and 46 women with a median (IQR) age of 52 years (29 to 67). Twenty-four patients with a median age of 44 years (35 to 65) were diagnosed with a scaphoid fracture on HR-pQCT.

RESULTS

When compared with HR-pQCT, conventional radiographs alone had a sensitivity of 67% (95% CI 45% to 84%), specificity of 85% (95% CI 74% to 93%), positive predictive value (PPV) of 62% (95% CI 46% to 75%), negative predictive value (NPV) of 88% (95% CI 80% to 93%), and a positive and negative likelihood ratio (LR) of 4.5 (95% CI 2.4 to 8.5) and 0.4 (95% CI 0.2 to 0.7), respectively. Compared with HR-pQCT, clinical reassessment alone resulted in a sensitivity of 58% (95% CI 37% to 78%), specificity of 42% (95% CI 30% to 54%), PPV of 26% (95% CI 19% to 35%), NPV of 74% (95% CI 62% to 83%), as well as a positive and negative LR of 1.0 (95% CI 0.7 to 1.5) and 1.0 (95% CI 0.6 to 1.7), respectively. Combining clinical examination with conventional radiography produced a sensitivity of 50% (95% CI 29% to 71%), specificity of 91% (95% CI 82% to 97%), PPV of 67% (95% CI 46% to 83%), NPV of 84% (95% CI 77% to 88%), as well as a positive and negative LR of 5.6 (95% CI 2.4 to 13.2) and 0.6 (95% CI 0.4 to 0.8), respectively.

CONCLUSION

The accuracy of conventional radiographs (80% compared with HR-pQCT) and clinical reassessment (46% compared with HR-pQCT) indicate that the value of clinical reassessment is limited in diagnosing scaphoid fractures and cannot be considered directive in managing scaphoid fractures. The combination of conventional radiographs and clinical reassessment does not increase the accuracy of these diagnostic tests compared with the accuracy of conventional radiographs alone and is therefore also limited in diagnosing scaphoid fractures.

LEVEL OF EVIDENCE

Level II, diagnostic study.

Prospective comparison of magnetic resonance imaging and computed tomography in diagnosing occult scaphoid fractures

BACKGROUND

Suspected scaphoid fracture (SF) after a fall on an outstretched hand is a common presentation in the emergency department. Magnetic resonance imaging (MRI) or computed tomography (CT) has been suggested to assist in the diagnosis or exclusion of SF.

PURPOSE

To compare MRI and CT at diagnosing occult SFs.

MATERIAL AND METHODS

We routinely perform CT scans in patients with clinically suspected occult SF, after 7-10 days of injury following two negative radiographs. All eligible patients with a clinically suspected SF, but negative radiographs and a negative CT, underwent an MRI scan to assess further for evidence of occult fracture.

RESULTS

A total of 100 patients were included in our study. MRI showed fractures in 16% of the time (in 15 patients) when plain radiographs and CT did not. Of these fractures, 8% were SFs. In addition to fractures, 10% had bone bruising. A total of 25% of patients with fractures and bone bruising were referred to the hand surgery team for further follow-up.

CONCLUSION

The study demonstrated that MRI would identify a radiographically occult SF more often than CT. This supports NICE guidelines which recommend MRI as the best early diagnostic tool for occult SFs.

Assessment of the healing of conservatively-treated scaphoid fractures using HR-pQCT

Improving the clinical outcome of scaphoid fractures may benefit from adequate monitoring of their healing in order to for example identify complications such as scaphoid nonunion at an early stage and to adjust the treatment strategy accordingly. However, quantitative assessment of the healing process is limited with current imaging modalities. In this study, high-resolution peripheral quantitative computed tomography (HR-pQCT) was used for the first time to assess the changes in bone density, microarchitecture, and strength during the healing of conservatively-treated scaphoid fractures. Thirteen patients with a scaphoid fracture (all confirmed on HR-pQCT and eleven on CT) received an HR-pQCT scan at baseline and three, six, twelve, and 26 weeks after first presentation at the emergency department. Bone mineral density (BMD) and trabecular microarchitecture of the scaphoid bone were quantified, and failure load (FL) was estimated using micro-finite element analysis. Longitudinal changes were evaluated with linear mixed-effects models. Data of two patients were excluded due to surgical intervention after the twelve-week follow-up visit. In the eleven fully evaluable patients, the fracture line became more apparent at 3 weeks. At 6 weeks, individual trabeculae at the fracture region became more difficult to identify and distinguish from neighboring trabeculae, and this phenomenon concerned a larger region around the fracture line at 12 weeks. Quantitative assessment showed that BMD and FL were significantly lower than baseline at all follow-up visits with the largest change from baseline at 6 weeks (-13.6% and - 23.7%, respectively). BMD remained unchanged thereafter, while FL increased. Trabecular thickness decreased significantly from baseline at three (-3.9%), six (-6.7%), and twelve (-4.4%) weeks and trabecular number at six (-4.5%), twelve (-7.3%), and 26 (-7.9%) weeks. Trabecular separation was significantly higher than baseline at six (+13.3%), twelve (+19.7%), and 26 (+16.3%) weeks. To conclude, this explorative HR-pQCT study showed a substantial decrease in scaphoid BMD, Tb.Th, and FL during the first 6 weeks of healing of conservatively-treated scaphoid fractures, followed by stabilization or increase in these parameters. At 26 weeks, BMD, trabecular microarchitecture, and FL were not returned to baseline values.

Association between bone shape and the presence of a fracture in patients with a clinically suspected scaphoid fracture

Scaphoid fractures are difficult to diagnose with current imaging modalities. It is unknown whether the shape of the scaphoid bone, assessed by statistical shape modeling, can be used to differentiate between fractured and non-fractured bones. Therefore, the aim of this study was to investigate whether the presence of a scaphoid fracture is associated with shape modes of a statistical shape model (SSM). Forty-one high-resolution peripheral quantitative computed tomography (HR-pQCT) scans were available from patients with a clinically suspected scaphoid fracture of whom 15 patients had a scaphoid fracture. The scans showed no motion artefacts affecting bone shape. The scaphoid bones were semi-automatically contoured, and the contours were converted to triangular meshes. The meshes were registered, followed by principal component analysis to determine mean shape and shape modes describing shape variance. The first five out of the forty shape modes cumulatively explained 87.8% of the shape variance. Logistic regression analysis was used to study the association between shape modes and fracture presence. The regression models were used to classify the 41 scaphoid bones as fractured or non-fractured using a cut-off value that maximized the sum of sensitivity and specificity. The classification of the models was compared with fracture diagnosis on HR-pQCT. A regression model with four shape modes had an area under the ROC-curve of 72.3% and correctly classified 75.6% of the scaphoid bones (fractured: 60.0%, non-fractured: 84.6%). To conclude, fracture presence in patients with a clinically suspected scaphoid fracture appears to be associated with the shape of the scaphoid bone.

Reliability of Radiographs and Computed Tomography in Diagnosing Scaphoid Union After Internal Fixation

PURPOSE

To evaluate the reliability of radiographs (XR) alone versus the combination of XR and computed tomography (CT) in determining scaphoid union following open reduction internal fixation (ORIF) with a headless compression screw.

METHODS

We used our imaging database to identify 32 XR and corresponding CTs over a 6-year period (from 2012 to 2018) that were performed to evaluate scaphoid healing following ORIF. Three hand surgeons evaluated the studies to assess (1) healing, (2) if partially healed, the percentage of healing, and (3) the certainty of healing. Initially, XR were reviewed alone. Three weeks later, the same XR were reviewed with the corresponding CTs. Each reviewer performed a similar 2-stage evaluation 4 weeks later. We measured interobserver and intraobserver reliabilities using linearly weighted kappa (κ) coefficients for healing status and the percentage of healing.

RESULTS

The interobserver reliability for healing (healed vs partially healed vs not healed) was moderate both with XR alone and with the combination of XR and CT. The intraobserver reliability for healing was substantial with XR alone compared to moderate with the combination of XR and CT. For the percentage of healing, both interobserver and intraobserver reliabilities were fair with XR alone or with the combination of XR and CT. Reviewers reported significantly greater certainty with the combination of XR and CT compared with XR alone.

CONCLUSIONS

Following ORIF, surgeons are more certain in their evaluation of scaphoid healing with the combination of CT and XR. However, the reliability of assessing scaphoid union may not be improved by the addition of CT to XR.

TYPE OF STUDY/LEVEL OF EVIDENCE

Diagnostic III.

Improved Detection of Scaphoid Fractures with High-Resolution Peripheral Quantitative CT Compared with Conventional CT

BACKGROUND

Computed tomography (CT), magnetic resonance imaging, and bone scintigraphy are second-line imaging techniques that are frequently used for the evaluation of patients with a clinically suspected scaphoid fracture. However, as a result of varying diagnostic performance results, no true reference standard exists for scaphoid fracture diagnosis. We hypothesized that the use of high-resolution peripheral quantitative CT (HR-pQCT) in patients with a clinically suspected scaphoid fracture could improve scaphoid fracture detection compared with conventional CT in the clinical setting.

METHODS

The present study included 91 consecutive patients (≥18 years of age) who presented to the emergency department with a clinically suspected scaphoid fracture between December 2017 and October 2018. All patients were clinically reassessed within 14 days after first presentation, followed by CT and HR-pQCT. If a scaphoid fracture was present, the fracture type was determined according to the Herbert classification system and correlation between CT and HR-pQCT was estimated with use of the Kendall W statistic or coefficient of concordance (W) (the closer to 1, the higher the correlation).

RESULTS

The cohort included 45 men and 46 women with a median age of 52 years (interquartile range, 29 to 67 years). HR-pQCT revealed a scaphoid fracture in 24 patients (26%), whereas CT revealed a scaphoid fracture in 15 patients (16%). Patients with a scaphoid fracture were younger and more often male. The correlation between CT and HR-pQCT was high for scaphoid fracture type according to the Herbert classification system (W = 0.793; 95% confidence interval [CI], 0.57 to 0.91; p < 0.001) and very high for scaphoid fracture location (W = 0.955; 95%, CI 0.90 to 0.98; p < 0.001).

CONCLUSIONS

In the present study, the number of patients diagnosed with a scaphoid fracture was 60% higher when using HR-pQCT as compared with CT. These findings imply that a substantial proportion of fractures-in this study, more than one-third-will be missed by the current application of CT scanning in patients with a clinically suspected scaphoid fracture.

LEVEL OF EVIDENCE

Diagnostic Level II. See Instructions for Authors for a complete description of levels of evidence.