Can CT automatically quantify the location and volume of knee osteophytes?

Median osteophyte Dice was 0.913 and 0.909 in the external cohorts; volume ICC was 0.977 and 0.972.

Original workflow showing AI development, two external validation cohorts and surgical-cohort assessment.
Zheng et al., Journal of Robotic Surgery, 2026 · DOI: 10.1007/s11701-026-04015-y · Figure 2.
Evidence & Original Source Image use terms

Original photographs of accessible osteophyte resection and water-displacement volumetry; the recovered volume is not a complete reference for all CT-visible osteophytes.
Zheng et al., Journal of Robotic Surgery, 2026 · DOI: 10.1007/s11701-026-04015-y · Figure 3.
Evidence & Original Source Image use terms

Participants and Methods

A human-in-the-loop 3D U-Net was developed and frozen. It was tested against independent expert-consensus segmentations at two external centers and water-displacement volumes of osteophytes recovered during surgery. Development: 387 knee CTs; two external cohorts: 100 knees each; prospective robotic TKA cohort: 66 knees. Exploratory structural analysis included 382 development-cohort knees.

Main Findings

Median osteophyte Dice was 0.913 and 0.909 in the external cohorts; volume ICC was 0.977 and 0.972. Agreement with recovered osteophyte volume showed ICC 0.876, MAE 1.224 cc, mean difference +0.401 cc and 95% limits of agreement from −2.785 to +3.586 cc.

The point of reading on surgery

This is a research measure of preoperative osteophyte burden. Effects on resection, gap balance and patient outcomes require prospective testing.

Limitations

Recovered osteophytes do not represent all CT-visible osteophytes. Water displacement was measured once at approximately 1-cc resolution, without repeatability testing. Individual-knee precision for automated surgical decisions and improvement in surgical outcomes were not established.

Let's put it briefly.

What did this study confirm?

Median osteophyte Dice was 0.913 and 0.909 in the external cohorts; volume ICC was 0.977 and 0.972.

Can it be applied directly to individual patients' treatment decisions?

Recovered osteophytes do not represent all CT-visible osteophytes. Water displacement was measured once at approximately 1-cc resolution, without repeatability testing.

Keywords

AI
CT
osteophyte
external validation
robotic TKA

Abstract · Original text

Zhiqian Zheng, Do Weon Lee, Sewon Jeon, Hong Yeol Yang, Jong Keun Seon, Ho-Jung Jung, Joong Il Kim, Byung Sun Choi, Hyuk-Soo Han, Du Hyun Ro. Automated CT-based quantification of knee osteophytes for robotic-assisted total knee arthroplasty: a multi-cohort study. Journal of Robotic Surgery. 2026. doi:10.1007/s11701-026-04015-y.

Excerpt from the original abstract

Preoperative computed tomography (CT) is routinely acquired for planning robotic-assisted total knee arthroplasty (TKA), yet osteophyte burden is not quantitatively characterized.

Read abstract at source Terms of use

The original paper.

Automated CT-based quantification of knee osteophytes for robotic-assisted total knee arthroplasty: a multi-cohort study

Zhiqian Zheng, Do Weon Lee, Sewon Jeon, Hong Yeol Yang, Jong Keun Seon, Ho-Jung Jung, Joong Il Kim, Byung Sun Choi, Hyuk-Soo Han, Du Hyun Ro
Journal of Robotic Surgery · 2026

DOI · 10.1007/s11701-026-04015-y Based data

Reviewed academic journal articles