Acute and Chronic Injury of the Knee Articular Cartilage: Prevalence, Injury mechanisms, MRI Assessment and Association with Osteoarthritis

Abstract

Knee articular cartilage lesions resulting from acute or chronic injury are frequently encountered in routine clinical magnetic resonance imaging (MRI). Acute lesions typically occur following recreational or athletic activity and are often associated with concomitant internal knee derangements. Chronic lesions, on the other hand, are thought to result from repetitive loading that exceeds physiological thresholds in both magnitude and frequency. Accurate characterization of cartilage lesion size, depth, and subchondral bone involvement is clinically relevant, as it directly influences therapeutic decision-making. In clinical practice, MRI evaluation of the knee cartilage is primarily performed using intermediate-weighted, two-dimensional fast spin-echo sequences, preferably with high spatial resolution and at 3 Tesla. Looking ahead, machine learning approaches hold promise for improving the prediction of osteoarthritis by integrating pathologic imaging findings with complementary sources of information such as clinical parameters, quantitative biomarkers, and patient-reported outcomes.

Introduction

Normal knee joint function relies on the integrity of a low-friction articular surface, which is maintained by healthy cartilage. However, due to the articular cartilage’s limited capacity for repair, cartilage injuries can lead to chronic pain and joint dysfunction [1]. Lesions of the articular cartilage are typically associated with acute traumatic injury, inflammatory arthropathy, and degenerative joint disease [2]. Athletes are particularly at risk of traumatic cartilage injury [3,4], which is often found with concomitant internal knee derangement [5]. Knee cartilage lesions may contribute to the progression of osteoarthritis (OA), a debilitating disease that affects millions of people worldwide and places a substantial burden on healthcare systems [6].

About 11% of the chondral lesions identified during arthroscopy are potentially repairable [7]. This underscores the importance of accurately detecting chondral lesions, which is critical not only for prevention of degenerative disease, but also for guiding surgical planning and optimizing postoperative outcomes. MRI is the current reference standard for noninvasive evaluation of articular cartilage, enabling assessment of both macroscopic morphology and microstructural composition [8]. Combined with the use of higher field strengths (≥3 Tesla [T]), continuous advances in MRI hardware and software development now permit high-resolution imaging of the cartilage within clinically feasible scan times, allowing accurate characterization of lesion size, depth, and subchondral bone involvement [9,10]. Over the past two decades, research on predicting incident radiographic OA or OA progression from cartilage lesions has largely relied on traditional regression models. However, with the advent of machine learning, this paradigm may shift, as machine learning offers a comprehensive framework for integrating diverse data sources and has the potential to further enhance the prediction of OA [11].

This review summarizes the prevalence and injury mechanisms of acute and chronic cartilage injury, highlights the role of MRI in cartilage assessment, and provides an overview of future directions in the use of imaging to predict OA outcomes.

Gross & microstructural anatomy

Articular cartilage is a highly specialized connective tissue that covers the articulating surfaces of bones within synovial joints, enabling load distribution and resisting compressive forces during joint motion.

Macroscopically, hyaline cartilage is characterized by its translucent, glass-like gross appearance. With aging, this appearance often becomes yellowish and opaque, a transformation which may be attributed to progressive dehydration and the accumulation of pigment deposits within the extracellular matrix [12]. Notably, adult articular cartilage is distinct among connective tissues in that it is aneural, avascular, alymphatic, and solely relies on passive diffusion of nutrients and metabolites from the synovial fluid and the subchondral bone. In contrast, immature cartilage may also receive nutritional support from the vascularized subchondral bone prior to full growth plate ossification [13].

Articular cartilage is predominantly composed of an extracellular matrix with a cellular component of only 2–5% of its wet weight. Chondrocytes, which are responsible for synthesis and composition of the matrix, are embedded at varying depths within this matrix. The extracellular matrix facilitates the mechanical properties of articular cartilage and is responsible for nutrient exchange. It is primarily composed of water (70–80% of its wet weight), type II collagen, and proteoglycans, along with additional non-collagenous proteins [14,15]. Nutrient transport to chondrocytes is facilitated by loading-induced fluid flow and the passive diffusion of solutes through the extracellular matrix. While the high water content of the matrix is a result of the polyanionic and hydrophilic nature of proteoglycans, type II collagen provides the matrix with tensile strength and structural integrity, counteracting the expansive osmotic pressure generated by the water within the matrix [16].

Histologically, adult articular cartilage is organized into four zones (Fig. 1), i.e., superficial zone, transitional zone, deep zone, and calcified cartilage zone; the interface between uncalcified and calcified cartilage can be identified by a calcified line referred to as the tidemark [12]. While the superficial aspect of articular cartilage interfaces with the joint cavity, its deep zone is anchored to the subchondral bone plate through a thin layer of calcified cartilage. The zonal architecture of cartilage determines its biomechanical properties: in the superficial zone, collagen fibers and elongated chondrocytes align parallel to the surface to resist shear forces, while in the transitional and deep zones, collagen fibers and chondrocytes align perpendicular to the surface to resist compression [16,17]. The calcified cartilage, in which chondrocytes are sparse and hypertrophic, serves as the attachment point to the subchondral bone. Proteoglycan distribution also varies by depth, being lowest at the surface and most concentrated around chondrocytes in the deep zone [16].

Fig. 1.

Hematoxylin and Eosin Stain of the articular cartilage. The articular cartilage is organized into superficial zone, transitional zone, deep zone, and calcified cartilage zone. The tidemark marks the boundary between the calcified and non-calcified zones of the cartilage. Proteoglycans, highlighted by blue hematoxylin staining, are predominantly found around chondrocytes in the deep zone. Image published with permission from [12].

Acute Cartilage Injury

Incidence and risk factors

The overall incidence of traumatic chondral injuries of the knee remains difficult to establish, as many are clinically silent and underdiagnosed. Large arthroscopy-based series consistently demonstrate a high burden of chondral pathology. Across cohorts which included thousands of procedures, 60–66% of knees exhibit at least one chondral lesion [3,5,18]. The medial femoral condyle is the most common site, followed by the patella [5]. Participation in competitive sports is linked to a higher incidence of knee injuries, particularly those affecting the articular cartilage. Among various disciplines, soccer, football, and skiing are most commonly associated with chondral injury [4,19–22]. Non-modifiable risk factors include male sex [5] and the presence of anterior cruciate ligament (ACL) tear, meniscal tear, or lateral patellar dislocation [3,23,24], which are discussed in more detail below. On the other hand, full-thickness cartilage defects may arise in otherwise healthy adults without identifiable risk factors. The potential underlying mechanism is suspected to be a chondro-osseous delamination of the cartilage at the tidemark, possibly driven by shear forces [25,26]. The precise etiology of cartilage delamination, however, remains poorly understood. Further research is warranted to clarify the underlying causes of why the tidemark fails, which would substantially advance our understanding of the pathophysiology of cartilage injury [27].

Injury mechanisms and clinical presentations

Traumatic injuries to specific regions of the knee cartilage can result from several distinct injury mechanisms. These are often located adjacent to bone marrow contusions with specific patterns, which serve as an indirect marker of underlying cartilage damage. Recognizing these contusion patterns on a static MR image can aid in reconstructing the dynamic mechanism of injury. Five primary mechanisms have been described [28,29]: the pivot shift mechanism, dashboard injury, hyperextension injury, clip injury, and lateral patellar dislocation, which are outlined in more detail in Table 1. However, bone marrow edema abnormalities may sometimes be difficult to classify [29].

Table 1.

Overview of knee injury mechanisms.

	Injury mechanism	Location of bone contusion	Associated ligamentous injuries
Pivot Shift Injury	Flexed knee, valgus load, internal femoral rotation on externally rotated tibia	Posterior lateral tibial plateau and lateral femoral condyle	ACL tear
Dashboard Injury	Direct anterior tibial impact on a flexed knee	Anterior tibia	PCL tear
Hyperextension Injury	Forceful hyperextension during landing or kicking	Anterior femoral condyle and anterior tibia	Posterolateral corner injury
Clip Injury	Lateral impact causing valgus stress	Lateral tibiofemoral compartment	MCL tear
Lateral Patellar Dislocation	Noncontact, flexed knee with lateral displacement of patella	Lateral femoral condyle and medial patellar facet	Medial retinaculum

Patients with traumatic chondral injuries often present with nonspecific symptoms, and many are unable to recall the precipitating event [26]. Up to 60% report reduced knee function compared to the contralateral side, with symptom severity correlating with lesion size [30,31]. The predominant clinical features include pain, recurrent swelling or effusion, which can mimic meniscal pathology. On the other hand, mechanical locking is relatively uncommon [26].

Patellofemoral and trochlear lesions typically cause anterior knee pain, exacerbated by activities such as jumping, deceleration, or kicking [32]. Lesions on the anterior femoral condyles may provoke pain during terminal extension, central condylar lesions during pivoting or lateral movements, and posterior condylar lesions during deep flexion or squatting [26].

Associated internal knee derangements

Anterior cruciate ligament tear

Articular cartilage damage is observed in up to 46% of cases of ACL injuries, increasing in prevalence with delayed surgical intervention [23,33,34]. MRI frequently shows the typical bone contusions at the posterior lateral tibial plateau and lateral femoral condyle. In some cases, an osteochondral depression fracture might be present (Fig. 2). Time to surgery has been identified as the strongest predictor of chondral injury severity; the risk of developing full-thickness cartilage lesions is more than four times higher when reconstruction is postponed beyond five years [33,35]. Moreover, sex-based differences in ACL injury risk are recognized, with women exhibiting a nearly fourfold increased likelihood of non-contact ACL tears compared to men [36]. However, associated chondral injuries occur twice as frequently in men with ACL tears compared to women [37].

Fig 2.

Intermediate-weighted fat-suppressed sagittal MR images of the right knee joint of a 30-year-old male patient with a complete ACL tear. The ACL tear (arrows) is shown in (A). The typical bone contusion pattern at the posterolateral tibial plateau and the weight-bearing portion of the lateral femoral condyle (arrowheads) are visible in (B). A disruption of the subchondral bone is noted in (B), suggestive of an osteochondral fracture (arrow). Cartilage hyperintensity adjacent to the posterolateral tibial plateau (arrow, C) is observed consistent with cartilage contusion. In addition, there is a deep cartilage fissure (>50%) adjacent to the contusion (arrowhead, C).

Meniscal tear

Meniscal tears are observed in approximately 36–40% of acute chondral injuries, and are frequently observed in patients with OA [3,18,38]. In the presence of an ACL injury, approximately three-quarters of chondral lesions are accompanied by meniscal tears [33]. Medial meniscal ramp lesions have been found in 16% of ACL tears and were associated with accelerated degeneration of cartilage composition of the medial knee compartment [39].

Meniscal extrusion, meniscectomy, and meniscal root tears have been shown to be associated with progressive cartilage loss [40] and with early onset of OA [41–43]. These associations are thought to result from repetitive joint stress and altered biomechanical load distribution, which contribute to chondral degeneration.

Lateral patellar dislocation

Lateral patellar dislocation is strongly associated with articular cartilage injury, with chondral lesions observed in 57–71% of patients undergoing arthroscopy (Fig. 3) [24,44]. Given this high prevalence, MRI has been recommended in all cases to accurately assess the extent of intra-articular damage [24]. The majority of cartilage lesions are located on the patella, and up to 26% of patients present with an osteochondral fracture [45]. On the other hand, the prevalence of cartilage injury in the lateral femoral condyle after patellar dislocation is reported to be variable [45,46].

Fig 3.

Intermediate-weighted fat-suppressed MR images of the right knee joint of a 29-year-old male patient with a history of patellar dislocation. MRI performed 3 months post-injury due to persistent pain demonstrates a near full-thickness cartilage defect of the medial patellar facet adjacent to the patellar ridge (arrows) on images in in axial orientation (A) and sagittal orientation (B). Adjacent signal abnormality of the cartilage in the sagittal image is compatible with cartilage contusion (arrowhead).

Chronic Cartilage Injury

Articular cartilage responds to mechanical stimuli by alterating matrix synthesis and cell metabolism [47]. The concept of chronic traumatization through mechanical overloading of the articular cartilage was initially described by Dye [48], who suggested that articular cartilage and subchondral bone become vulnerable to structural failure if the magnitude or frequency of applied forces surpasses physiologic thresholds.

A number of animal and in vitro studies have examined how mechanical overloading contributes to cartilage injury and disruption of the ECM. When cartilage is exposed to excessive load (either bei magnitude or frequency), it undergoes multiple structural and cellular alterations. These include chondrocyte apoptosis [49], softening [50,51], the development of fissures [52], and loss of collagen interconnectivity [53]. Notably, some authors have reported that collagen deterioration can begin deep within the tissue even before any visible surface defects appear. At higher loading rates, additional forms of cartilage damage may arise, such as loss of proteoglycans [54], microscopic cracks [55], and collagen damage [52,56]. However, these distinct yet interconnected pathways of cartilage degeneration, all of which culminate in varying manifestations of collagen network damage, is not yet fully understood [57].

On the other hand, immobilization and partial weight bearing has been associated with cartilage thinning in humans [58,59], along with ECM changes which were observed in animal models [56,60].

Classification of cartilage lesions

Multiple semiquantitative MRI grading systems derived from arthroscopic classifications exist for grading of cartilage lesions. The arthroscopic International Cartilage Repair Society (ICRS) classification, modified Outerbridge classification and the modified Noyes classification remain the most commonly used in clinical practice [8,61–64], and are summarized in Table 2. These classifications primarily evaluate the depth of the cartilage defect and if subchondral bone involvement is present. However, radiologists should apply cartilage grading systems with caution and explicitly state the grading system used, as identical grades from different systems describe a varying extent of cartilage injury. Moreover, cartilage lesions frequently exhibit diverse morphologies (Fig. 4), and even within the same joint compartment, multiple distinct defects may coexist. Consequently, a structured radiology report that directly describes lesion morphology, size, and depth is often more practical and clinically useful than forcing findings into a rigid grading framework [2]. In clinical trials focused on knee osteoarthritis, MRI-based morphologic assessment of cartilage is frequently performed using detailed semiquantitative scoring systems. Among the most widely adopted is the Whole-Organ Magnetic Resonance Imaging Score (WORMS), which provides a comprehensive framework for evaluating cartilage integrity alongside other knee structures relevant for knee osteoarthritis [65]. More recently, the MRI Osteoarthritis Knee Score (MOAKS) system has been developed to improve grading for cartilage, meniscus and bone marrow lesions [66].

Table 2.

MRI grading systems for articular cartilage lesions [8,61–64]

Grade	ICRS	Modified Outerbridge	Modified Noyes
0	Normal cartilage	Normal cartilage	Normal cartilage
I	Superficial lesions: soft indentation and/or superficial fissures and cracks	Superficial chondral softening, swelling, or blistering with intact cartilage surface	Normal cartilage with focally increased signal intensity on T2-weighted images
II	Lesions extending <50% of cartilage depth	Superficial cartilage fragmentation, ulceration, fibrillation, or fissuring involving an area 1⁄2 an inch or less in diameter and less than 50% of the cartilage depth	Partial-thickness lesion: 2a = <50% cartilage thickness; 2b = >50% cartilage thickness)
III	Lesions extending >50% of cartilage depth, down to calcified layer; deep fissures/fragmentation	Deep cartilage ulceration, fibrillation, or fissuring involving an area more than 50% or more of the cartilage depth but without exposure of the subchondral bone	Full-thickness cartilage defect
IV	Full-thickness cartilage defect with subchondral bone changes	Full-thickness cartilage defect with exposure of subchondral bone

Fig. 4.

Intermediate-weighted, fat-suppressed axial MR images of the patellofemoral joint showing the morphologic appearance of cartilage lesions with varying depths.. T2 hyperintensity of the patellar cartilage is shown in (A), low-grade partial-thickness lesion (<50% cartilage thickness) in (B), high-grade partial-thickness lesion (>50% cartilage thickness) in (C), and full-thickness cartilage loss in (D). Subchondral bone involvement (arrowheads) is evident in A–D.

MRI assessment of articular cartilage

MRI is currently considered as the reference standard for noninvasive cartilage evaluation [67]. While different field strengths from less than 1 T to 3 T are available in clinical practice, diagnostic performance, e.g. for the detection of low-grade cartilage defects or for evaluation of outcomes after cartilage repair procedures, can be improved with increased spatial resolution of MR images and higher field strength [68,69].

In routine clinical musculoskeletal MRI, two-dimensional (2D) fast spin echo (FSE) or turbo spin echo (TSE) sequences remain the cornerstone for morphological articular cartilage assessment due to their superior signal-to-noise ratio (SNR), high in-plane spatial resolution, and versatile tissue contrast. Among these, proton density- (PD) and intermediate-weighted (IW) 2D FSE/TSE sequences with or without fat suppression are most commonly used. PD-weighted sequences, characterized by shorter echo times (~20 ms), offer superior depiction of the tidemark, but are prone to magic angle artifacts [70]. In contrast, intermediate-weighted sequences (TE ~35–45 ms) are the preferred choice in clinical practice due to their improved fluid conspicuity and reduced susceptibility to magic angle artifacts compared to PD-weighted sequences [38]. T2-weighted images further increase contrast between the articular surface and intra-articular fluid, improving sensitivity for identifying superficial cartilage lesions and chondro-osseous cartilage delamination [10]. However, these sequences come at the cost of reduced overall contrast resolution between musculoskeletal soft tissues and the cartilage matrix [10,38].

Owing to hardware improvements and increased availability of MRI systems with field strengths ≥3 T and beyond, high-resolution imaging of articular cartilage has become feasible within clinically acceptable acquisition times. Notably, 7 T MRI systems could increase diagnostic confidence in detecting very small chondral lesions compared to lower field strengths due the inherently higher SNR [71]. The recent implementation of deep learning reconstructions in 7 T MRI knee imaging might accelerate its transition from research to clinical practice by providing high-resolution cartilage images within feasible scan times [72–74]. Limitations of 7 T imaging include a substantial artifact increase, specific absorption rate constraints, and a limited number of available dedicated coils [75].

Three-dimensional (3D) MRI techniques are useful for cartilage evaluation due to the acquisition of isotropic voxels, which allows multiplanar reformations in any orientation from a single volumetric dataset [9]. 3D fluid-bright gradient echo sequences, such as dual-echo steady-state (DESS) and steady-state free precession (SSFP), are combined with spectral fat suppression or water excitation, and provide adequate contrast between articular cartilage, subchondral bone, and joint fluid [76]. Currently, these sequences are predominantly applied in research settings to enable precise manual or automated cartilage segmentation [76,77]. However, gradient echo sequences have limited diagnostic accuracy for detecting other internal joint derangements compared to FSE/TSE sequences. Thus, these sequences are not a suitable replacement for FSE/TSE pulse sequences in clinical MRI protocols [10,76,78]. Current 3D FSE/TSE sequences demonstrate diagnostic performance for detection of cartilage lesions comparable to that of conventional 2D techniques [79]. Importantly, advances in acceleration strategies are mitigating the previously long acquisition times of these 3D sequences, increasing their feasibility for routine clinical use [9,80].

On routine clinical MRI, isolated focal cartilage defects with sharply angled margins are typically the result of traumatic injuries. However, cartilage fissures may also arise as a consequence of acute trauma [81]. A cartilage flap typically develops when a fissure progresses, partially detaches the cartilage, and results in subsequent elevation of the overlying cartilage (Fig. 5). Chondro-osseous cartilage delamination describes the separation of articular cartilage from the underlying subchondral bone occurring along the tidemark, with or without a surface fissure ([82]). On MRI, it is characterized by a linear high-signal intensity tracking between cartilage and bone, sometimes accompanied by buckling of the detached cartilage layer (Fig. 6). Cartilage fragments may remain attached, partially detach, or become fully loose within the joint space (Fig. 7). Accordingly, recognition of such defects should trigger a systematic search for intra-articular chondral bodies that may have been displaced from the donor site [2,83]. As previously discussed, chondro-osseous cartilage delamination may also arise in individuals without a preceding history of injury [27]. Furthermore, it is important to recognize that cartilage flaps, chondro-osseous delamination, and cartilage defects may actually represent different manifestations or stages of the same underlying pathological process [27,84].

Fig.5.

Coronal intermediate-weighted, fat-suppressed MR images of the left knee joint of two individuals acquired at 7 Tesla field strength. Cartilage lesion with obtuse margins of the medial femoral condyle (arrow) and associated subchondral bone marrow edema-like signal (arrowhead), consistent with cartilage loss due to degenerative joint disease is shown in (A). Subtle flap-like chondral lesion of the medial femoral condyle (arrow) is shown in (B). Image courtesy of Reto Sutter, MD (Balgrist University Hospital, Zurich, Switzerland)

Fig. 6.

Intermediate-weighted, fat-suppressed MR images of the right knee in a 32-year-old male with history of acute-on-chronic swelling following recreational sports activity. The sagittal image (A) and coronal image (B) reveal chondro-osseous cartilage delamination, demonstrated by T2 hyperintense signal interposed between the articular cartilage and the underlying subchondral bone (arrows).

Fig. 7.

Intermediate-weighted, fat-suppressed MR images of the right knee joint of a 27-year-old female patient with a history of atraumatic recurrent knee swelling and pain. The sagittal image (A) demonstrates a full-thickness cartilage defect at the weight-bearing portion of the lateral femoral condyle (arrow) and adjacent subchondral bone marrow edema–like signal (arrowheads), which may be the result of a preceding chondro-osseous delamination. The coronal image (B) shows a displaced chondral fragment within the intercondylar notch (arrow).

Early degenerative cartilage changes may present as subtle contour irregularities on MR images. Fissures or superficial fibrillations are an early structural change marked by vertical clefts between chondrocytes, forming irregular “fingerlike” projections into the joint space (Fig. 8), and occur especially in high-friction zones like the patellofemoral joint or load-bearing regions of the femorotibial joint [2].

Fig. 8.

Intermediate-weighted, fat-suppressed axial MR images of the patellofemoral joint showing fissuring of the lateral patellar facet extending to the subchondral bone (arrow) in (A), and superficial “fingerlike” fibrillations (arrow) along the patellar ridge in (B).

Additional early features of degenerative cartilage injury include localized variations in cartilage thickness, which may involve thinning or, in some cases, transient thickening before actual tissue loss occurs. Increased T₂ signal intensity changes within the cartilage may also appear, often reflecting microscopic collagen matrix disruption and elevated free-water content [83].

In more advanced stages of degenerative cartilage injury, MRI typically demonstrates widespread areas of cartilage thinning with varying depth and extent, frequently involving opposing joint surfaces. Defects generally exhibit broad, obtuse edges (Fig. 5), often accompanied by edema-like signal in the subchondral bone or cystic changes, or areas of low T2 signal intensity indicating fibrosis or trabecular sclerosis. Other findings seen with advanced cartilage disease include osteophytes, bone marrow lesions, meniscal pathology, and effusion-synovitis (Fig. 9) [8,83].

Fig. 9.

Intermediate-weighted, fat-suppressed MR images of the right knee joint of a 70-year-old male patient with typical MR imaging features of osteoarthritis: Full-thickness cartilage loss in the medial compartment (arrow) with adjacent edema-like signal in the tibial plateau (arrowhead) is shown in (A). Extrusion and maceration of the medial meniscus are also present (outlined arrow). The sagittal image (B) demonstrate osteophyte formation of the femoral condyle (arrow) and joint effusion (arrowhead). In (C), patellofemoral osteoarthritis with deep cartilage fissures at the trochlear groove is visible (arrow). Additionally, joint effusion is observed (arrowhead).

Association with knee osteoarthritis

OA is a multi-factorial painful disease of the synovial joints and affects more than 500 million individuals worldwide [85]. The knee is the most commonly involved joint, with radiographic evidence observed in nearly one-third of adults over 45 years of age [86]. Risk factors include age, female sex, obesity, biomechanical factors such as malalignment, history of trauma, and joint instability due to surgery [87–91]. Disease burden and progression are most often assessed using indirect radiographic measures of cartilage loss such as Kellgren-Lawrence grade or joint space width [92,93], but it should be noted that OA is a disorder of the entire synovial joint rather than the cartilage alone. To address the need for the assessment of other structural changes of the knee joint in clinical trials, MRI scoring systems such as WORMS and MOAKS were established to assess internal joint derangements beyond cartilage loss [65,66].

MRI findings that are associated with progression of radiographic OA or predict future knee arthroplasty include extent of cartilage damage, osteophytes, meniscal tears (especially root tears), meniscal extrusion, joint effusion and subchondral bone marrow lesions [43,94–97]. Beyond morphologic assessment, compositional MRI techniques such as T_1rho and T₂ mapping provide sensitive measures of cartilage matrix integrity, enabling detection of early cartilage damage prior to irreversible defects and offering predictive value for radiographic OA [100–102].

Novel developments to predict osteoarthritis progression

Future directions in predicting OA outcomes are increasingly centered on machine learning models, which has seen rapid adoption in recent years. These models can integrate different data sources for evaluation of disease progression beyond the scope of traditional regression models [11]. As an example, a recent study by Bonakdari et al. [103] found that knee bone curvature could predict cartilage volume loss over one year with a strong predictive performance (R ≥ 0.78). Another study by Jamshidi et al. [104] used >1000 baseline features of data, including 135 MRI variables, to predict cartilage volume loss in patients with knee OA. Using machine learning models, the authors found that the strongest predictors among MRI-based measures were cartilage thickness and cartilage surface area in the medial tibial plateau. More recently, Joseph et al. [105] described that two machine learning models, one with 112 predictors and a second one with a reduced number of 10 key predictors selected by feature importance achieved a comparable prediction (AUC = 0.77) of incident radiographic OA after eight years follow-up. These results suggest that machine learning models can capture complex data and feature interactions, uncovering prognostic patterns that often remain undetected with traditional regression approaches [11].

Conclusion

Cartilage lesions of the knee resulting from acute or chronic injury are clinically significant findings that can be noninvasively diagnosed with MRI. Currently, high-resolution 2D intermediate-weighted fat-suppressed FSE/TSE sequences can be considered as clinical standard for lesion detection and characterization. Descriptive reporting of lesion location, size, depth, and subchondral involvement should be preferred over using classification systems, given the variability among the latter. Cartilage lesions may ultimately progress to osteoarthritis. Chondral damage and other MRI findings such as meniscal abnormalities, and bone marrow lesions are predictors of disease progression and future knee arthroplasty. Machine learning approaches are expected to integrate imaging biomarkers with demographic, clinical, biochemical and biomechanical predictors, advancing individualized OA risk prediction.