Dr. Peter Obid // University Clinic Freiburg

Artificial intelligence in radiology: Measuring coronal parameters in adolescent idiopathic scoliosis

Challenge

Before the initiation of this study, no thoroughly validated AI-based algorithms existed for the automated measurement of radiographic parameters in the coronal plane of full spinal radiographs

Solution

Development of novel AI-algorithms for the fully automated measurement of coronal plane parameters.

Outcome

The algorithms underwent a successful validation study that will be presented at the 37th meeting of the North American Spine Society in 2022. This study created the scientific basis for the future use of the algorithms in routine clinical care. 

artificial intelligence in radiology

Excerpt

Dr. Peter Obid, in cooperation with RAYLYTIC’s Research & Development Team, worked to develop artificial intelligence applications  the fully automated measurement of these parameters.

During their development, our UNITY platform served as the electronic data capture (EDC) and data management system

Different applications for artificial intelligence in radiology

Artificial intelligence in radiology has undergone rapid expansion in the last decade.

Scoliosis is the most common spinal deformity in children and teenagers.1 The accurate and objective measurement of spinal balance (SB) parameters, particularly in the sagittal and coronal planes, is critical for diagnosing scoliosis and planning its surgical correction.  

Manual evaluation of SB parameters often suffers from being imprecise and time-consuming. Improving the diagnosis and treatment requires massive amounts of data from diverse sources, like large registry data and routine checkups. Even if this data is accessible, the subjective placement of anatomical landmarks can skew measurements. Departments often lack the resources to conduct these kinds of analyses. There, big data projects are often abandoned early on. 

“The effort and time associated with compiling and analyzing registry data with manual techniques is staggering,” Obid remarked. 

However, artificial intelligence (AI) in radiology has begun to prevail as a rapid and precise method to diagnose and measure scoliosis. Before Obid’s study, only parameters in the sagittal plane had benefited from the speed and objectivity of AI-powered analyses.  

Novel AI-algorithms for the automated measurement of radiographic parameters in the coronal plane

Dr. Peter Obid, at the time of the study chief resident of orthopedic and spinal surgery at Greifswald University Hospital and now head of spinal surgery at Freiburg University clinic, was looking to expand the capabilities of AI in the diagnosis and treatment of scoliosis. To achieve this, he cooperated with RAYLYTIC to develop AI-algorithms for automatically measuring geometric parameters in the coronal plane of full spinal radiographs.  

The goal of the project, which was supported by the AO Foundation, was to develop artificial intelligence algorithms for use in radiology that measure the following coronal parameters with objectivity and precision: Cobb angles of major and minor curves, shoulder balance, T1 tilt, coronal balance, and lumbar modifier.   

RAYLYTIC’s UNITY platform served as the electronic data capture (EDC) and data management system. Here, UNITY’s different modules were used to upload and process and anonymize the imaging data in accordance with regulatory and data protection requirements.  

Outcomes and benefits

More time, more data

Dr. Obid revealed that, by automating his measurements, he can save 10 to 20 minutes per patient, depending on the complexity and extent of the analysis. Therefore, he has more time to analyze additional data from different sources. Artificial intelligence in radiology will likely continue to alleviate radiologists from time-consuming and error-prone measurements.

In this regard, Obid praised the symbiotic working relationship he was able to forge with RAYLYTIC: Providing RAYLYTIC with data gives him more time to access and analyze data, which he uses to improve his treatment methods. 

Obid’s optimistic prognosis for artificial intelligence in radiology

To scientifically corroborate their precision, the algorithms successfully underwent a validation study. The results will be presented at the 37th North American Spine Society meeting in Chicago this fall.  

Obid’s project is ongoing. He expressed his satisfaction with the precision of the already developed algorithms but noted that the most challenging part of the project is upcoming: the algorithms’ postoperative usage. Here he predicts that the continuing cooperation between with RAYLYTIC will result in postoperatively applicable algorithms that are just as precise as the ones in preoperative use.   

The straightforwardness of working with the RAYLYTIC team has also made Obid optimistic about AI’s potential in the future of scoliosis treatments. With RAYLYTIC’s help, he hopes to “augment the existing algorithms with the capability to assess parameters from different viewing angles.” Specifically, automatically measuring parameters in lateral and AP radiographs would represent a major development in the field. 

¹ Théroux, J., Stomski, N., Hodgetts, C.J. et al. Prevalence of low back pain in adolescents with idiopathic scoliosis: a systematic review. Chiropr Man Therap 25, 10 (2017). 

Are you facing a similar challenge and need support?

Discover how automation can transform your research.

Arrange a free and non-binding demo appointment

UNITY Research: Powerful analysis and controlling tools to conduct clinical studies efficiently.