Knee osteoarthritis (KOA) diagnosis is based on symptoms, assessed through questionnaires such as the WOMAC. However, the inconsistency of pain recording and the discrepancy between joint phenotype and symptoms highlight the need for objective biomarkers in KOA diagnosis. To this end, we study relationships among clinical and molecular data in a cohort of women (n=51) with Kellgren-Lawrence grade 2–3 KOA through Support Vector Machine (SVM) and a regulation network model (RNM). Clinical descriptors (i.e., pain catastrophism (CA); depression (DE); functionality (FU); joint pain (JP); rigidity (RI); sensitization (SE); synovitis (SY)) are used to classify patients. A Youden's test is performed for each classifier to determine optimal binarization thresholds for the descriptors. Thresholds are tested against patient stratification according to baseline WOMAC data from the Osteoarthritis Initiative, and the mean accuracy is 0.97. For our cohort, the data used as SVM inputs are KOA descriptors, synovial fluid (SL) proteomic measurements (n=25), and transcription factors (TF) activation obtained from RNM [2] stimulated with the SL measurements. The relative weights after classification reflect input importance. The performance of each classifier is evaluated through AUC-ROC analysis. The best classifier with clinical data is CA (AUC = 0.9), highly influenced by FU and SE, suggesting that kinesophobia is involved in pain perception. With SL input, leptin strongly influences every classifier, suggesting the importance of low-grade inflammation. When TF are used, the mean AUC is limited to 0.608, which can be related to the pleomorphic behaviour of osteoarthritic chondrocytes. Nevertheless, FU has an AUC of 0.7 with strong importance of FOXO downregulation. Though larger and longitudinal cohorts are needed, this unique combination of SVM and RNM shall help to map objectively KOA descriptors. Acknowledgements: Catalan & Spanish governments 2020FI_b00680; STRATO-PID2021126469ob-C21-2, European Commission (MSCA-TN-ETN-2020-Disc4All-955735, ERC-2021-CoG-O-Health-101044828). ICREA Academia

Abstract. Objectives. Knee alignment affects both the development and surgical treatment of knee osteoarthritis. Automating femorotibial angle (FTA) and hip-knee-ankle angle (HKA) measurement from radiographs could improve reliability and save time. Further, if the gold-standard HKA from full-limb radiographs could be accurately predicted from knee-only radiographs then the need for more expensive equipment and radiation exposure could be reduced. The aim of this research is to assess if deep learning methods can predict FTA and HKA angle from posteroanterior (PA) knee radiographs. Methods. Convolutional neural networks with densely connected final layers were trained to analyse PA knee radiographs from the Osteoarthritis Initiative (OAI) database with corresponding angle measurements. The FTA dataset with 6149 radiographs and HKA dataset with 2351 radiographs were split into training, validation and test datasets in a 70:15:15 ratio. Separate models were learnt for the prediction of FTA and HKA, which were trained using mean squared error as a loss function. Heat maps were used to identify the anatomical features within each image that most contributed to the predicted angles. Results. FTA could be predicted with errors less than 3° for 99.8% of images, and less than 1° for 89.5%. HKA prediction was less accurate than FTA but still high: 95.7% within 3°, and 68.0 % within 1°. Heat maps for both models were generally concentrated on the knee anatomy and could prove a valuable tool for assessing prediction reliability in clinical application. Conclusions. Deep learning techniques could enable fast, reliable and accurate predictions of both FTA and HKA from plain knee radiographs. This could lead to cost savings for healthcare providers and reduced radiation exposure for patients

Bone marrow lesions (BMLs) have been extensively linked to the osteoarthritis (OA) disease pathway in the knee. Semi-quantitative evaluation has been unable to effectively study the spatial and temporal distribution of BMLs and consequently little is understood about their natural history. This study used a novel statistical model to precisely locate the BMLs within the subchondral bone and compare BML distribution with the distribution of denuded cartilage. MR images from individuals (n=88) with radiographic evidence of OA were selected from the Osteoarthritis Initiative. Slice-by-slice, subvoxel delineation of the lesions was performed across the paired images using the criteria laid out by Roemer (2009). A statistical bone model was fitted to each image across the cohort, creating a dense set of anatomically corresponded points which allowed BML depth, position and volume to be calculated. The association between BML and denudation was also measured semi-quantitatively by visually scoring the lesions as either overlapping or adjacent to denuded AC, or not. At baseline 75 subjects had BMLs present in at least one compartment. Of the 188 compartments with BMLs 46% demonstrated change greater than 727mm cubed, the calculated smallest detectable difference. The majority of lesions were found in medial compartments compared to lateral compartments and the patella (Figure 1A). Furthermore, in the baseline images 76.9% of all BMLs either overlapped or were adjacent to denuded bone. The closeness of this relationship in four individuals is shown in Figure 1B. The distribution of lesions follows a clear trend with the majority found in the patellofemoral joint, medial femoro-tibial joint and medial tibial compartment. Moreover the novel method of measurement and display of BMLs demonstrates that there is a striking similarity between the spatial distribution of BMLs and denuded cartilage in subjects with OA. This co-location infers the lesions have a mechanical origin much like the lesions that occur in healthy patients as a direct result of trauma. It is therefore suggested that OA associated BMLs are in fact no different from the BMLs caused by mechanical damage, but occur as a result of localised disruption to the joint mechanics, a common feature of OA