Advanced MRI techniques in hip cartilage assessment

Hip deformities are the main cause of hip pain and osteoarthritis in young adults, oftentimes necessitating hip-preserving surgery or, in more severe cases, a hip replacement. In order to successfully perform these surgeries and ensure the best possible outcome, surgeons rely on various imaging techniques. However, these techniques only produce 2D images which fail to accurately depict the complex 3D anatomy of the hip, and composing 3D models out of 2D images is a very time-consuming process. New technologies and techniques are applied in an attempt to automate 3D imaging and make it more convenient and reliable for hip cartilage assessment.

Challenges in joint preserving hipsurgery

Hip deformities, such as femoroacetabular impingement (FAI) and developmental dysplasia of the hip (DDH), are leading causes of hip pain and early osteoarthritis in young adults, often resulting in the need for total hip replacement.¹ FAI is caused by an early bony conflict between the pelvis and femur which results from osseous deformities and damages the labrum and cartilage over time. In DDH, inadequate coverage of the femoral head by the acetabulum leads to increased stress to the acetabular rim. Hip-preserving surgeries aim to correct these deformities and improve joint biomechanics, potentially delaying joint degeneration. However, challenges remain.

One challenge is to determine whether patients with early to moderate degeneration would benefit more from hip-preserving surgery or if hip replacement might be the favorable option based on already too severe joint deterioration.² The other challenge is to accurately diagnose hip deformities. Depending on the underlying pathology, i.e. instability or hip impingement, different surgical techniques need to be considered. In cam-type FAI the aspherical portion at the femoral head-neck junction is resected. In pincer-type FAI there is a range of different degrees of acetabular overcoverage with or without a retroverted but presumably normally sized acetabular cup. Surgical treatment of pincer FAI may be either resection of the too deep acetabulum or correcting the retroverted acetabulum with anteverting periacetabular osteotomy which is a more complex and invasive surgery. Thus accurate diagnosis of the underlying deformity is crucial to achieve good outcome. The two major morphologic pathologies leading to hip instability are hip dysplasia due to insufficient acetabular coverage and increased femoral version. In these patients a periacetabular osteotomy or femoral osteotomy can be considered. Identifying the underlying pathology with standard preoperative imaging remains a challenge, especially in combined or borderline pathologies such as borderline hip dysplasia with associated cam deformities. Advanced diagnostic tools and image postprocessing yield great potential for improving preoperative diagnosis and surgical outcome.

Current standard imaging in joint preserving hip surgery

Current standard imaging modalities for diagnosing structural hip deformities include radiographs, computed tomography (CT), and magnetic resonance imaging (MRI). Radiographs are the primary diagnostic tool and are used to classify the bony morphology based on acetabular coverage and orientation as well as initial assessment of femoral head-neck junction.³ However, they only provide a 2D view of the hip’s complex 3D anatomy, which can be affected by variations in imaging techniques and patient positioning.³ To visualize osseous hip anatomy in more detail, CT can be used to create fast 3D bone models for a thorough analysis of acetabular and femoral morphology but has limited use for visualization of soft tissue. By contrast MRI offers excellent soft tissue contrast and is the modality of choice in assessing cartilage and labrum damage.³ However, joint degeneration can only be detected at an advanced level with macroscopic structural damage to the hip cartilage and labrum. This assessment is more or less qualitative and is prone to inter-observer variability. Furthermore, even though 3D MRI-sequences are commonly performed for image reformation, the actual assessment of labrum and cartilage damage is performed on few selected 2D images since manual segmentation of 3D labrum and cartilage models is too time-consuming for clinical routine.

Traction MRI of the hip for improved detection of macroscopic cartilage damage

The hip is a ball-and-socket joint with the labrum dividing it into peripheral and central compartments. In standard MR arthrography, a contrast agent expands the hip capsule and flows into the central joint cavity through the acetabular notch. However, most of the contrast does not spread centrally due to the opposing femoral and acetabular cartilage layers, which typically prevent its accumulation unless there is significant cartilage damage. Therefore, visualizing acetabular cartilage delamination which represents a typical and frequent finding in patients with cam-type FAI is challenging with direct MR arthrography.⁴

To improve visualization of the central joint cavity during MR arthrography, axial leg traction can be applied, similar to the joint distraction needed during hip arthroscopy (Fig. 1). Traction MR arthrography of the hip has shown high accuracy in detecting chondrolabral lesions, ligamentum teres injuries, and intra-articular loose bodies and has demonstrated its value in predicting failure of FAI surgery.^5–8 It is particularly helpful for better visualizing the extent of cartilage damage in older patients with mild radiographic joint degeneration and thus aid in surgical decision making. Traction MR arthrography can also detect unstable labral tears in patients with hip dysplasia and may aid in planning treatment for femoral head necrosis in young patients with early collapse and preserved joint space.⁴

Biochemical MRI for compositional quantification of cartilage damage

Cartilage integrity is a crucial factor for the success of joint-preserving surgery. Morphologic MRI, the current gold standard, can only detect joint degeneration at advanced stages when severe structural changes in the hip cartilage and labrum are already present.⁹ However, biochemical deterioration of cartilage, marked by the loss of glycosaminoglycans (GAG), occurs before these changes become visible on MRI. To address this limitation, biochemically sensitive MRI techniques like delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) have been developed.^10,11

GAG, a key component of healthy cartilage, is negatively charged, and its loss reduces the cartilage’s overall negative charge. In dGEMRIC, a negatively charged gadolinium contrast agent penetrates the cartilage as GAG levels decrease, reducing the T1 relaxation time, which can be measured in MRI. Thus, dGEMRIC provides a quantifiable biomarker of cartilage health. Early studies at 1.5 Tesla showed dGEMRIC’s potential to predict the risk of early failure after joint-preserving surgery.^12,13 However, with the adoption of 3 Tesla MRI in clinical practice, field inhomogeneity issues emerged with the traditional dGEMRIC technique. To address this, more robust inversion-recovery-based method was developed for 3D dGEMRIC imaging (MP2RAGE).¹⁴ Despite these advancements, standard dGEMRIC analysis requires manual placement of regions of interest on select 2D images, making it time-consuming and user-dependent, which limits its routine clinical use.

AI-based 3D analysis of cartilage and labrum morphology and cartilage damage

As mentioned above, hip MRI assessments are based on selected 2D images. Creating 3D models requires manual segmentation, which involves dividing an image into multiple regions of interest based on pixel properties. This process is done slice by slice, making it very time-consuming and impractical for routine clinical use and is affected by inter-observer variability even when performed by an expert. Deep learning, a branch of machine learning, has shown great promise in automating image segmentation. Convolutional neural networks, particularly those with U-net architecture, are commonly used for this purpose. More recently, the nnU-net, an extension of the U-net, was introduced. It automatically configures itself for biomedical image segmentation and has produced excellent results across various medical imaging tasks.¹⁵ Compared to other joints such as the knee automatic segmentation of hip cartilage is more challenging due to its naturally thinner structure and oblique orientation. Previous attempts to automate hip cartilage segmentation were based on small sample sizes and did not include the labrum.^16–18 Ideally, combining a 3D analysis of chondro-labral morphology with cartilage composition would enable an improved analysis of chondro-labral dimensions and of the severity of joint damage (Fig. 2 & 3).

More recently, we introduced a deep-learning-based approach for automatic segmentation of hip cartilage and labrum using a 3D dGEMRIC sequence (MP2RAGE). The approach allows for automatic and reliable generation of 3D models of hip cartilage and labrum within seconds, and providing both 3D morphologic information of the hip joint as well as color-coded dGEMRIC indices for quantitative analysis of cartilage damage (Fig. 2 & 3).¹⁹

This new technique allows for the first time to investigate the 3D morphology of hip deformities based on MRI (Fig.4). Using this technology enabled us to quantify the size of the articular surface and demonstrate that the severely retroverted acetabulum does not actually represent a deep hip. Instead hips with acetabular retroversion present a smaller articulating cartilage surface area but a comparably large non-articulating fossa acetabularis (Fig. 4). This finding supports anteverting periacetabular osteotomy over acetabular rim trimming in order to prevent iatrogenic hip dysplasia. Furthermore, we could quantify the adaptive hypertrophy of the acetabular labrum in hip dysplasia compared to hips with a normal acetabulum and pincer-type hips. A fast and reliable individual analysis of cartilage and labrum surface area compared to surface area of the non-articulating fossa acetabularis yields unprecedented ways to characterize hip morphology with the goal to improve surgical planning on a routine basis. Furthermore, a fully automated volumetric analysis of cartilage composition using 3D dGEMRIC maps may aid surgeons when contemplating.

Literatur: