Prediction of knee osteoarthritis progression using radiological descriptors obtained from bone texture analysis and Siamese neural networks: data from OAI and MOST cohorts

Abstract

Background Trabecular bone texture (TBT) analysis has been identified as an imaging biomarker that provides information on trabecular bone changes due to knee osteoarthritis (KOA). In parallel with the improvement in medical imaging technologies, machine learning methods have received growing interest in the scientific osteoarthritis community to potentially provide clinicians with prognostic data from conventional knee X-ray datasets, in particular from the Osteoarthritis Initiative (OAI) and the Multicenter Osteoarthritis Study (MOST) cohorts. Patients and methods This study included 1888 patients from OAI and 683 patients from MOST cohorts. Radiographs were automatically segmented to determine 16 regions of interest. Patients with an early stage of OA risk, with Kellgren and Lawrence (KL) grade of 1 < KL < 4, were selected. The definition of OA progression was an increase in the OARSI medial joint space narrowing (mJSN) grades over 48 months in OAI and 60 months in MOST. The performance of the TBT-CNN model was evaluated and compared to well-known prediction models using logistic regression. Results The TBT-CNN model was predictive of the JSN progression with an area under the curve (AUC) up to 0.75 in OAI and 0.81 in MOST. The predictive ability of the TBT-CNN model was invariant with respect to the acquisition modality or image quality. The prediction models performed significantly better with estimated KL (KLprob) grades than those provided by radiologists. TBT-based models significantly outperformed KLprob-based models in MOST and provided similar performances in OAI. In addition, the combined model, when trained in one cohort, was able to predict OA progression in the other cohort. Conclusion The proposed combined model provides a good performance in the prediction of mJSN over 4 to 6 years in patients with relevant KOA. Furthermore, the current study presents an important contribution in showing that TBT-based OA prediction models can work with different databases.

Radiographic Biomarkers for Knee Osteoarthritis: A Narrative Review

“…In addition to the usual clinical covariates, such as age, gender and BMI, radiographic-based measurements, such as JSW, joint space area (JSA), OARSI JSN grade and KL grade, are commonly used for the prediction of KOA progression [ 27 , 41 ]. Texture analysis methods can play an important role in the improvement of the clinical-based prediction models as they provide numerical measures of KOA-induced bone changes [ 26 ].…”

“…The performance of baseline TBT-based biomarkers was investigated for the prediction of radiographic KOA progression in terms of increase in mJSN [ 29 , 33 , 41 , 42 , 43 , 44 ].…”

“…The VAR method was also used for the calculation of TBT features of 1888 and 683 X-ray knee images from the OAI and the Multicentre Osteoarthritis Study (MOST) cohorts, respectively [ 41 ]. The proposed prediction model included TBT features associated with CNN-based KL descriptors, as well as traditional clinical covariates and radiological JSN scores.…”

“…Besides the studies of Kraus et al [ 27 , 45 ] and Duncan et al [ 34 ], where one knee per person was included, all other studies did not take into consideration the assumption of knee independence. Several studies have trained and validated their models using impressive number of radiographs (>1000 XRs) [ 28 , 35 , 41 , 46 ], whereas several other studies were based on a too limited number of radiographs, (<250 XRs) [ 29 , 33 , 42 , 44 , 45 , 63 ].…”

“…However, for the prediction of KOA progression, although knees with moderate KOA at baseline were intentionally targeted, there was no specific inclusion criterion that was mostly employed. As reported in Table 3 , the criterion for knee inclusion was knees with KL = 1, 2 or 3 in three studies [ 27 , 29 ], with KL = 2 or 3 in five studies [ 28 , 41 , 42 , 43 , 45 ], or with KL ≥ 1 in three studies [ 44 , 46 ], and with KL ≥ 2 in two studies [ 33 , 35 ]. In order to have fair and valid comparisons of the performance of the proposed KOA prediction models, the KOA community should identify a unique definition of both KOA incidence and progression as well as standardize a specific inclusion criterion.…”