AUTOMATIC MENISCUS ANALYSIS DEMONSTRATES REPAIR IS NOT SUPERIOR TO MENISCECTOMY IN IMPROVING MENISCAL UTE-T2* PROPERTIES 2-YEARS POST ACLR

Abstract

INTRODUCTION

Concomitant meniscus tear is common with ACL injury and amplifies OA risk. MRI ultrashort echo-time T2* (UTE-T2*) is sensitive to the compositional integrity of the meniscus and is histologically verified to associate with collagen fibril alignment. We implemented an automated pipeline to determine whether meniscal T2* composition 2-years after ACL reconstruction (ACLR) differs between patients with and without a meniscal tear at the time of surgery.

OBJECTIVE

To test whether menisci found to be torn at the time of ACLR exhibit, at 2-year follow-up, higher mean UTE-T2* reflecting greater compositional degeneration than intact menisci and whether meniscal repair demonstrates lower UTE-T2* than meniscectomy at 2-year follow-up.

METHODS

111 ACLR patients (53/111(48%) female; mean[SD] age: 32[10]yrs; BMI: 25[3]kg/m2) underwent 3T MRI 2 years after ACLR (2.0[0.9]years). UTE-T2* maps were generated by fitting a mono-exponential decay curve to sagittal T2*-weighted images using a Levenberg-Marquardt algorithm. Images were acquired at 8 TEs (32μs -16ms, non-uniform spacing) using a radial-out 3D Cones acquisition, TR = 22ms, in-plane resolution = 0.313 to 0.364 mm, and 3mm slice thickness. Menisci were automatically segmented using a U-Net pre-trained on >300 DESS volumes and fine-tuned to segment root-sum-of-squares images combining Cones echoes 2–6. Training labels were generated by registering DESS images to Cones and propagating the segmentation. Automated segmentation was evaluated in a validation cohort using the dice similarity coefficient (DSC) and average symmetric surface distance (ASSD). The menisci were subdivided into anterior, middle, and posterior thirds using an automated polar coordinate-based system (Fig 1). Meniscal tear and treatment at the time of ACLR was assessed from operative reports. UTE-T2* differences between torn and intact menisci, and between repair versus meniscectomy were assessed with t-tests (or Mann-Whitney U tests). Statistical analyses were performed with SPSS (IBM) and Excel (Microsoft).

RESULTS

Automated segmentation in the validation cohort (n=16) had median DSC = 0.71 and ASSD = 0.52 mm for the medial, and DSC = 0.68 and ASSD = 0.51 mm for the lateral meniscus. At the time of ACLR, meniscal tears were observed in 56/111(50%) patients: 24/111(22%) of medial and 45/111(41%) of lateral menisci. More tears were resected: 11/24(45%) medial, 25/45(56%) lateral than repaired: 10/24(42%) medial, 15/45(33%) lateral. Patients with any medial meniscal tear had higher mean UTE-T2* in the middle (14%, p<0.001) and posterior (20%, p=0.002) regions compared to those with intact menisci, (Fig 2). Patients with any lateral meniscal tear had 20% higher mean UTE-T2* in the middle region of the lateral menisci compared to those with intact menisci (p=0.001). Two years post-ACLR, no mean UTE-T2* differences were detected between repaired and resected menisci (p>0.181).

CONCLUSION

Menisci torn at the time of ACLR showed elevated UTE-T2* at 2-year follow-up, regardless of treatment. These data suggest that repair does not fully restore compositional properties. While meniscal repair is often favored over meniscectomy to preserve function and potentially reduce OA risk, repair is more costly and requires longer rehabilitation. Further research should identify whether specific tear patterns or patient factors influence compositional recovery, and whether UTE-T2* at 2-year follow-up predicts OA development and progression. This study demonstrates that automatic segmentation and analysis of sub-regional meniscal UTE-T2* is feasible and can provide standardized assessments across large data sets, including in knees with surgical hardware.