Deep Learning Based Automatic Ankle Tenosynovitis Quantification from MRI in Patients with Psoriatic Arthritis: A Feasibility Study

Abstract

Background/Objectives: Tenosynovitis is a common feature of psoriatic arthritis (PsA) and is typically assessed using semi-quantitative magnetic resonance imaging (MRI) scoring. However, visual scoring s variability. This study evaluates a fully automated, deep-learning approach for ankle tenosynovitis segmentation and volume-based quantification from MRI in psoriatic arthritis (PsA) patients. Methods: We analyzed 364 ankle 3T MRI scans from 71 PsA patients. Four tenosynovitis pathologies were manually scored and used to create ground truth segmentations through a human-machine workflow. For each pathology, 30 annotated scans were used to train a deep-learning segmentation model based on the nnUNet framework, and 20 scans were used for testing, ensuring patient-level disjoint sets. Model performance was evaluated using Dice scores. Volumetric pathology measurements from test scans were compared to radiologist scores using Spearman correlation. Additionally, 218 serial MRI pairs were assessed to analyze the relationship between changes in pathology volume and changes in visual scores. Results: The segmentation model achieved promising performance on the test set, with mean Dice scores ranging from 0.84 to 0.92. Pathology volumes correlated with visual scores across all test MRIs (Spearman ρ = 0.52-0.62). Volume-based quantification captured changes in inflammation over time and identified subtle progression not reflected in semi-quantitative scores. Conclusions: Our automated segmentation tool enables fast and accurate quantification of ankle tenosynovitis in PsA patients. It may enhance sensitivity to disease progression and complement visual scoring through continuous, volume-based metrics.

Keywords: MRI; deep learning; tenosynovitis.

Figure 1

Workflow of the study.

Figure 2

Human–machine workflow of creating ground truth pathology segmentation dataset. (a–d): (a): tendons segmented (colors), (b): ROIs around tendons defined (green), (c): clustering highlights regions (lighter regions), (d): highlighted regions inside ROIs selected (red).

Figure 3

Automatic segmentation of tendons. a: Tibialis posterior tendon, b: flexor digitorum tendon, c: flexor halluces longus tendon, d: peroneal longus/brevis tendon.

Figure 4

Example image overlaid with automatic segmentation of 4 tenosynovitis pathologies. (a) Tibialis posterior tenosynovitis (score: 3), (b) flexor digitorum tenosynovitis (score: 1), (c) flexor halluces longus tenosynovitis (score: 1), (d) peroneal longus/brevis tenosynovitis (score: 1).

Figure 5

Fully manual and semi-automatic (human–machine) tibialis posterior tenosynovitis segmentation performance. Rij denotes reader i in j-th repetition.

Figure 6

The changes in volume along with scores in all patients. Blue crosses indicate score changes over time as scored by a radiologist, while green dots indicate volume changes as automatically segmented. This figure also illustrates one image that shows negative change in the volume of pathology between the two time-points while the score does not change. Also another example where score negatively changes, but the change in volume is not big.