“Artificial intelligence in X-ray imaging analysis – what is possible today?”
Dr. Dreischarf’s presentation titled “AI in X-ray image analysis – what is possible today?” rounded out a session highlighting the newest innovations in prosthetic technology.
In contrast to widely held notions by the orthopedics industry about AI being a future application, Dreischarf showed that AI-based based algorithms are already capable of automatically identifying anatomical entities, placing landmarks, and extracting and quantifying complex, clinically valuable information from standard clinical X-rays of the human musculoskeletal system.
Dr. Marcel Dreischarf explains the use of artificial intelligence in assessing orthopedic implant performance at the DKOU 2022. An accompanying article authored by RAYLYTIC appears in the event’s daily magazine.
With this automation comes speed and precision, promising clinicians relief from time-consuming, error-prone measurements – a critical component of surgical planning and quality control.
However, clinicians are not the only ones benefiting from these technological developments. Other industry players, such as orthopedic device manufacturers, may also use AI to assess key attributes of implant performance. Several of the algorithms presented by Dr. Dreischarf are already finding application in the long-term assessment of implant performance, for example, in clinical studies. Combining AI-based algorithms with 2D-3D registration, a method for estimating spatial relationships between 3D structures, enables the assessment of implant wear and migration within the micrometer range.
How reliable is AI?
Clinicians in the audience expressed lingering concerns about the reliability of AI, particularly in the case of images with abnormal patient positioning or post-operative instrumentation.
To ensure the reliability of the algorithms, RAYLYTIC – together with leading physicians – has developed robust validation procedures. Here, AI measurements are systematically compared to manual measurements conducted by experienced surgeons. The results of these studies, which have been published in academic journals, such as the European Spine Journal1, the Journal of Neurosurgery: Spine2 and, most recently, Diagnostics3, show that AI demonstrates comparable reliability to surgeons.
The demand for big data analysis
The common thread throughout the session was the pressing need for enhanced clinical data collection and analysis tools.
RAYLYTIC’s UNITY platform – containing modules for AI-based analysis of medical images and the capture of patient reported outcome measures (PROMs) – is providing a solution for the growing demand to correlate radiological and patient-reported outcomes. Together, these can aid physicians in drawing conclusions between treatment technique, implant performance, and outcomes to provide patients with the highest standard of care.
Predicting that the future of patient registries will reside in “registry-embedded studies,” or the use of extant registry data for clinical studies, the preceding presentation, for example, underscored the demand for tools that can automate big data analyses.
The DKOU 2022
This year the DKOU, one of Europe’s largest conventions for orthopedics and traumatology, carried the motto of “Passion for the Patient.” In halls of South Berlin’s convention center, industry experts, surgeons, and medical device manufacturers gathered to exchange technology and the latest developments in the field.
At the end of four days, one thing is clear: Better patient care requires better, more complete data.
1. Orosz, L., Bhatt, F., Jazini, E., Dreischarf, M., Grover, P., Grigorian, J., Roy, R., Schuler, T., Good, C., Haines, C. Novel Artificial Intelligence Algorithm Accurately and Independently Measures Spinopelvic Parameters. Journal of Neurosurgery: Spine (2022).
2. Grover, P., Siebenwirth, J., Caspari, C., Drange, S., Dreischarf, M., Le Huec, J.-C., Putzier, M., Franke, J. Can artificial intelligence support or even replace physicians in measuring sagittal balance? A validation study on preoperative and postoperative full spine images of 170 patients. European Spine Journal (2022).
3. Erne, F., Grover, P., Dreischarf, M., Reumann, M.K., Saul, D., Histing, T., Nüssler, A.K., Springer, F., Scholl, C. Automated Artificial Intelligence-Based Assessment of Lower Limb Alignment Validated on Weight-Bearing Pre- and Postoperative Full-Leg Radiographs. Diagnostics (2022).