A deep learning algorithm to identify cervical ossification of posterior longitudinal ligaments on radiography

Area of Science:

• Radiology
• Artificial Intelligence
• Orthopedics

Background:

• Cervical ossification of the posterior longitudinal ligament (cOPLL) is often misdiagnosed or missed on standard radiography.
• Accurate diagnosis of cOPLL is crucial for appropriate patient management and preventing neurological complications.

Purpose of the Study:

• To validate a deep learning algorithm for diagnosing cOPLL on cervical radiography.
• To compare the diagnostic accuracy of the algorithm against experienced spine physicians.

Main Methods:

• A deep learning algorithm was developed using radiographic data from 486 patients (243 with cOPLL, 243 controls).
• The algorithm's diagnostic performance was assessed (AUC 0.94, accuracy 0.88).
• The algorithm's diagnoses were compared to those of four spine physicians using 50 independent cases.

Main Results:

• The deep learning algorithm achieved a diagnostic accuracy of 0.88 (AUC 0.94).
• The algorithm correctly diagnosed 47 out of 50 cases, significantly outperforming spine physicians who correctly diagnosed 39 out of 50 cases (p=0.041).

Conclusions:

• The deep learning algorithm demonstrates significantly higher diagnostic accuracy for cOPLL than experienced spine physicians.
• This AI tool has the potential to improve the diagnostic accuracy of cOPLL using cervical radiography.