Recommender-based bone tumour classification with radiographs-a link to the past.

OBJECTIVES: To develop an algorithm to link undiagnosed patients to previous patient histories based on radiographs, and simultaneous classification of multiple bone tumours to enable early and specific diagnosis. MATERIALS AND METHODS: For this retrospective study, data from 2000 to 2021 were curated from our database by two orthopaedic surgeons, a radiologist and a data scientist. Patients with complete clinical and pre-therapy radiographic data were eligible. To ensure feasibility, the ten most frequent primary tumour entities, confirmed histologically or by tumour board decision, were included. We implemented a ResNet and transformer model to establish baseline results. Our method extracts image features using deep learning and then clusters the k most similar images to the target image using a hash-based nearest-neighbour recommender approach that performs simultaneous classification by majority voting. The results were evaluated with precision-at-k, accuracy, precision and recall. Discrete parameters were described by incidence and percentage ratios. For continuous parameters, based on a normality test, respective statistical measures were calculated. RESULTS: Included were data from 809 patients (1792 radiographs; mean age 33.73 ± 18.65, range 3-89 years; 443 men), with Osteochondroma (28.31%) and Ewing sarcoma (1.11%) as the most and least common entities, respectively. The dataset was split into training (80%) and test subsets (20%). For k = 3, our model achieved the highest mean accuracy, precision and recall (92.86%, 92.86% and 34.08%), significantly outperforming state-of-the-art models (54.10%, 55.57%, 19.85% and 62.80%, 61.33%, 23.05%). CONCLUSION: Our novel approach surpasses current models in tumour classification and links to past patient data, leveraging expert insights. CLINICAL RELEVANCE STATEMENT: The proposed algorithm could serve as a vital support tool for clinicians and general practitioners with limited experience in bone tumour classification by identifying similar cases and classifying bone tumour entities. KEY POINTS: • Addressed accurate bone tumour classification using radiographic features. • Model achieved 92.86%, 92.86% and 34.08% mean accuracy, precision and recall, respectively, significantly surpassing state-of-the-art models. • Enhanced diagnosis by integrating prior expert patient assessments.

Deep Learning-Based Detection of Bone Tumors around the Knee in X-rays of Children.

Even though tumors in children are rare, they cause the second most deaths under the age of 18 years. More often than in other age groups, underage patients suffer from malignancies of the bones, and these mostly occur in the area around the knee. One problem in the treatment is the early detection of bone tumors, especially on X-rays. The rarity and non-specific clinical symptoms further prolong the time to diagnosis. Nevertheless, an early diagnosis is crucial and can facilitate the treatment and therefore improve the prognosis of affected children. A new approach to evaluating X-ray images using artificial intelligence may facilitate the detection of suspicious lesions and, hence, accelerate the referral to a specialized center. We implemented a Vision Transformer model for image classification of healthy and pathological X-rays. To tackle the limited amount of data, we used a pretrained model and implemented extensive data augmentation. Discrete parameters were described by incidence and percentage ratio and continuous parameters by median, standard deviation and variance. For the evaluation of the model accuracy, sensitivity and specificity were computed. The two-entity classification of the healthy control group and the pathological group resulted in a cross-validated accuracy of 89.1%, a sensitivity of 82.2% and a specificity of 93.2% for test groups. Grad-CAMs were created to ensure the plausibility of the predictions. The proposed approach, using state-of-the-art deep learning methodology to detect bone tumors on knee X-rays of children has achieved very good results. With further improvement of the algorithm, enlargement of the dataset and removal of potential biases, this could become a useful additional tool, especially to support general practitioners for early, accurate and specific diagnosis of bone lesions in young patients.

SAM-X: sorting algorithm for musculoskeletal x-ray radiography.

OBJECTIVE: To develop a two-phased deep learning sorting algorithm for post-X-ray image acquisition in order to facilitate large musculoskeletal image datasets according to their anatomical entity. METHODS: In total, 42,608 unstructured and pseudonymized radiographs were retrieved from the PACS of a musculoskeletal tumor center. In phase 1, imaging data were sorted into 1000 clusters by a self-supervised model. A human-in-the-loop radiologist assigned weak, semantic labels to all clusters and clusters with the same label were merged. Three hundred thirty-two non-musculoskeletal clusters were discarded. In phase 2, the initial model was modified by "injecting" the identified labels into the self-supervised model to train a classifier. To provide statistical significance, data split and cross-validation were applied. The hold-out test set consisted of 50% external data. To gain insight into the model's predictions, Grad-CAMs were calculated. RESULTS: The self-supervised clustering resulted in a high normalized mutual information of 0.930. The expert radiologist identified 28 musculoskeletal clusters. The modified model achieved a classification accuracy of 96.2% and 96.6% for validation and hold-out test data for predicting the top class, respectively. When considering the top two predicted class labels, an accuracy of 99.7% and 99.6% was accomplished. Grad-CAMs as well as final cluster results underlined the robustness of the proposed method by showing that it focused on similar image regions a human would have considered for categorizing images. CONCLUSION: For efficient dataset building, we propose an accurate deep learning sorting algorithm for classifying radiographs according to their anatomical entity in the assessment of musculoskeletal diseases. KEY POINTS: • Classification of large radiograph datasets according to their anatomical entity. • Paramount importance of structuring vast amounts of retrospective data for modern deep learning applications. • Optimization of the radiological workflow and increase in efficiency as well as decrease of time-consuming tasks for radiologists through deep learning.

Two-Phase Deep Learning Algorithm for Detection and Differentiation of Ewing Sarcoma and Acute Osteomyelitis in Paediatric Radiographs.

BACKGROUND/AIM: Ewing sarcoma is a highly malignant tumour predominantly found in children. The radiological signs of this malignancy can be mistaken for acute osteomyelitis. These entities require profoundly different treatments and result in completely different prognoses. The purpose of this study was to develop an artificial intelligence algorithm, which can determine imaging features in a common radiograph to distinguish osteomyelitis from Ewing sarcoma. MATERIALS AND METHODS: A total of 182 radiographs from our Sarcoma Centre (118 healthy, 44 Ewing, 20 osteomyelitis) from 58 different paediatric (≤18 years) patients were collected. All localisations were taken into consideration. Cases of acute, acute on chronic osteomyelitis and intraosseous Ewing sarcoma were included. Chronic osteomyelitis, extra-skeletal Ewing sarcoma, malignant small cell tumour and soft tissue-based primitive neuroectodermal tumours were excluded. The algorithm development was split into two phases and two different classifiers were built and combined with a Transfer Learning approach to cope with the very limited amount of data. In phase 1, pathological findings were differentiated from healthy findings. In phase 2, osteomyelitis was distinguished from Ewing sarcoma. Data augmentation and median frequency balancing were implemented. A data split of 70%, 15%, 15% for training, validation and hold-out testing was applied, respectively. RESULTS: The algorithm achieved an accuracy of 94.4% on validation and 90.6% on test data in phase 1. In phase 2, an accuracy of 90.3% on validation and 86.7% on test data was achieved. Grad-CAM results revealed regions, which were significant for the algorithms decision making. CONCLUSION: Our AI algorithm can become a valuable support for any physician involved in treating musculoskeletal lesions to support the diagnostic process of detection and differentiation of osteomyelitis from Ewing sarcoma. Through a Transfer Learning approach, the algorithm was able to cope with very limited data. However, a systematic and structured data acquisition is necessary to further develop the algorithm and increase results to clinical relevance.

In Vitro Cartilage Regeneration with a Three-Dimensional Polyglycolic Acid (PGA) Implant in a Bovine Cartilage Punch Model.

Resorbable polyglycolic acid (PGA) chondrocyte grafts are clinically established for human articular cartilage defects. Long-term implant performance was addressed in a standardized in vitro model. PGA implants (+/- bovine chondrocytes) were placed inside cartilage rings punched out of bovine femoral trochleas (outer Ø 6 mm; inner defect Ø 2 mm) and cultured for 84 days (12 weeks). Cartilage/PGA hybrids were subsequently analyzed by histology (hematoxylin/eosin; safranin O), immunohistochemistry (aggrecan, collagens 1 and 2), protein assays, quantitative real-time polymerase chain reactions, and implant push-out force measurements. Cartilage/PGA hybrids remained vital with intact matrix until 12 weeks, limited loss of proteoglycans from "host" cartilage or cartilage-PGA interface, and progressively diminishing release of proteoglycans into the supernatant. By contrast, the collagen 2 content in cartilage and cartilage-PGA interface remained approximately constant during culture (with only little collagen 1). Both implants (+/- cells) displayed implant colonization and progressively increased aggrecan and collagen 2 mRNA, but significantly decreased push-out forces over time. Cell-loaded PGA showed significantly accelerated cell colonization and significantly extended deposition of aggrecan. Augmented chondrogenic differentiation in PGA and cartilage/PGA-interface for up to 84 days suggests initial cartilage regeneration. Due to the PGA resorbability, however, the model exhibits limitations in assessing the "lateral implant bonding".

Thickness of the Stifle Joint Articular Cartilage in Different Large Animal Models of Cartilage Repair and Regeneration.

OBJECTIVE: Regulatory guidelines for preclinical cartilage repair studies suggest large animal models (e.g., sheep, goat, [mini]-pig, or horse) to obtain results representative for humans. However, information about the 3-dimensional thickness of articular cartilage at different implantation sites in these models is limited. DESIGN: To identify the most suitable site for experimental surgery, cartilage thickness at the medial femoral condyle (MFC), lateral femoral condyle (LFC), and trochlea in ovine, caprine, and porcine cadaver stifle joints was systematically measured using hematoxylin-eosin staining of 6 µm paraffin sections and software-based image analysis. RESULTS: Regarding all ventral-dorsal regions of the MFC, goat showed the thickest articular cartilage (maximal mean thickness: 1299 µm), followed by sheep (1096 µm) and mini-pig (604 µm), with the highest values in the most ventral and dorsal regions. Also for the LFC, the most ventral regions showed the thickest cartilage in goat (maximal mean thickness: 1118 µm), followed by sheep (678 µm) and mini-pig (607 µm). Except for the mini-pig, however, the cartilage thickness on the LFC was consistently lower than that on the MFC. The 3 species also differed along the transversal measuring points on the MFC and LFC. In contrast, there were no consistent differences for the regional cartilage thickness of the trochlea among goat and sheep (≥780 µm) and mini-pig (≤500 µm). CONCLUSIONS: Based on their cartilage thickness, experimental defects on goat and sheep MFC may be viable options for preclinical cartilage repair studies, in addition to well-established horse models.

In Vitro Analysis of Cartilage Regeneration Using a Collagen Type I Hydrogel (CaReS) in the Bovine Cartilage Punch Model.

OBJECTIVE: Limitations of matrix-assisted autologous chondrocyte implantation to regenerate functional hyaline cartilage demand a better understanding of the underlying cellular/molecular processes. Thus, the regenerative capacity of a clinically approved hydrogel collagen type I implant was tested in a standardized bovine cartilage punch model. METHODS: Cartilage rings (outer diameter 6 mm; inner defect diameter 2 mm) were prepared from the bovine trochlear groove. Collagen implants (± bovine chondrocytes) were placed inside the cartilage rings and cultured up to 12 weeks. Cartilage-implant constructs were analyzed by histology (hematoxylin/eosin; safranin O), immunohistology (aggrecan, collagens 1 and 2), and for protein content, RNA expression, and implant push-out force. RESULTS: Cartilage-implant constructs revealed vital morphology, preserved matrix integrity throughout culture, progressive, but slight proteoglycan loss from the "host" cartilage or its surface and decreasing proteoglycan release into the culture supernatant. In contrast, collagen 2 and 1 content of cartilage and cartilage-implant interface was approximately constant over time. Cell-free and cell-loaded implants showed (1) cell migration onto/into the implant, (2) progressive deposition of aggrecan and constant levels of collagens 1 and 2, (3) progressively increased mRNA levels for aggrecan and collagen 2, and (4) significantly augmented push-out forces over time. Cell-loaded implants displayed a significantly earlier and more long-lasting deposition of aggrecan, as well as tendentially higher push-out forces. CONCLUSION: Preserved tissue integrity and progressively increasing cartilage differentiation and push-out forces for up to 12 weeks of cultivation suggest initial cartilage regeneration and lateral bonding of the implant in this in vitro model for cartilage replacement materials.

Biomarkers for tissue engineering of the tendon-bone interface.

The tendon-bone interface (enthesis) is a highly sophisticated biomaterial junction that allows stress transfer between mechanically dissimilar materials. The enthesis encounters very high mechanical demands and the regenerative capacity is very low resulting in high rupture recurrence rates after surgery. Tissue engineering offers the potential to recover the functional integrity of entheses. However, recent enthesis tissue engineering approaches have been limited by the lack of knowledge about the cells present at this interface. Here we investigated the cellular differentiation of enthesis cells and compared the cellular pattern of enthesis cells to tendon and cartilage cells in a next generation sequencing transcriptome study. We integrated the transcriptome data with proteome data of a previous study to identify biomarkers of enthesis cell differentiation. Transcriptomics detected 34468 transcripts in total in enthesis, tendon, and cartilage. Transcriptome comparisons revealed 3980 differentially regulated candidates for enthesis and tendon, 395 for enthesis and cartilage, and 946 for cartilage and tendon. An asymmetric distribution of enriched genes was observed in enthesis and cartilage transcriptome comparison suggesting that enthesis cells are more chondrocyte-like than tenocyte-like. Integrative analysis of transcriptome and proteome data identified ten enthesis biomarkers and six tendon biomarkers. The observed gene expression characteristics and differentiation markers shed light into the nature of the cells present at the enthesis. The presented markers will foster enthesis tissue engineering approaches by setting a bench-mark for differentiation of seeded cells towards a physiologically relevant phenotype.

Biodegradable poly (lactic acid-co-glycolic acid) scaffolds as carriers for genetically-modified fibroblasts.

Recent advances in gene delivery into cells allow improved therapeutic effects in gene therapy trials. To increase the bioavailability of applied cells, it is of great interest that transfected cells remain at the application site and systemic spread is minimized. In this study, we tested clinically used biodegradable poly(lactic acid-co-glycolic acid) (PLGA) scaffolds (Vicryl & Ethisorb) as transient carriers for genetically modified cells. To this aim, we used human fibroblasts and examined attachment and proliferation of untransfected cells on the scaffolds in vitro, as well as the mechanical properties of the scaffolds at four time points (1, 3, 6 and 9 days) of cultivation. Furthermore, the adherence of cells transfected with green fluorescent protein (GFP) and vascular endothelial growth factor (VEGF165) and also VEGF165 protein secretion were investigated. Our results show that human fibroblasts adhere on both types of PLGA scaffolds. However, proliferation and transgene expression capacity were higher on Ethisorb scaffolds most probably due to a different architecture of the scaffold. Additionally, cultivation of the cells on the scaffolds did not alter their biomechanical properties. The results of this investigation could be potentially exploited in therapeutic regiments with areal delivery of transiently transfected cells and may open the way for a variety of applications of cell-based gene therapy, tissue engineering and regenerative medicine.

Efficient decellularization for tissue engineering of the tendon-bone interface with preservation of biomechanics.

Interfaces between tendon/ligament and bone ("entheses") are highly specialized tissues that allow for stress transfer between mechanically dissimilar materials. Entheses show very low regenerative capacity resulting in high incidences of failure after surgical repair. Tissue engineering is a promising approach to recover functionality of entheses. Here, we established a protocol to decellularize porcine entheses as scaffolds for enthesis tissue engineering. Chemical detergents as well as physical treatments were investigated with regard to their efficiency to decellularize 2 mm thick porcine Achilles tendon entheses. A two-phase approach was employed: study 1 investigated the effect of various concentrations of sodium dodecyl sulfate (SDS) and t-octylphenoxypolyethoxy-ethanol (Triton X-100) as decellularization agents. The most efficient combination of SDS and Triton was then carried forward into study 2, where different physical methods, including freeze-thaw cycles, ultrasound, perfusion, and hydrostatic washing were used to enhance the decellularization effect. Cell counts, DNA quantification, and histology showed that washing with 0.5% SDS + 1% Triton X-100 for 72 h at room temperature could remove ~ 98% cells from the interface. Further investigation of physical methods proved that washing under 200 mmHg hydrostatic pressure shortened the detergent exposing time from 72 h to 48 h. Biomechanical tensile testing showed that the biomechanical features of treated samples were preserved. Washing under 200 mmHg hydrostatic pressure with 0.5% SDS + 1% Triton X-100 for 48 h efficiently decellularized entheses with preservation of matrix structure and biomechanical features. This protocol can be used to efficiently decellularize entheses as scaffolds for tissue engineering.

Biomechanical testing to evaluate the cut-through resistance of intramedullary nails for the proximal humerus.

Intramedullary nailing for stabilization of proximal humeral fractures is well-established. Complications as part of a cut-through, such as backing out of locking screws, loss of reduction, and perforation of the screws into the glenoid, are equally well-known. The test bench presented in this study enables testing of the cut-through behavior of multiple intramedullary implants on a simulated osteoporotic three-part fracture configuration with three different loading circumstances (A, B and C). In situation A, the glenohumeral dynamic force with progressive loadings entered at an angle of 15° to the humeral shaft. In situation B the force entered at an angle of 35° and in situation C the angle measured 55°. Three different types of nails were tested: the Targon PH with the optimal proximal screw length (T) and with all four proximal screws shortened (Tshort), the Synthes MultiLoc PHN with (S5) and without (S4) the additional calcar screw and, lastly, the PolyAxNail PH, a polyaxial intramedullary nail, in a neutral screw configuration (PAN) and a version with diametrically opposed crossed first and fourth locking screws (PAN10). Significant differences in the three cases were found with the evaluation of the failure load, which represents the cut-through resistance. Case A: Tshort (245.4 ± 18.7 N) - S4 (346.8 ± 18.0 N) (adjusted p = 0.002); Tshort (245.4 ± 18.7 N) - S5 (368.5 ± 12.0 N) (adjusted p = < 0.001); Tshort (245.4 ± 18.7 N) - T (323.5 ± 38.2 N) (p = 0.004); Case B: no significant differences between the study groups (adjusted significance). Case C: PAN (412.5 ± 16.0 N) - S5 (471.5 ± 21.5 N) (adjusted p = 0.007); T (414.0 ± 33.5 N) - S5 (471.5 ± 21.5 N) (adjusted p = 0.008). The optimal screw length has a strong influence on the failure load. Choosing proximal screws that are too short, produces a negative impact on the cut-through resistance. The additional calcar screw of the MultiLoc PHN and the polyaxiality of the PolyAxNail showed a positive effect with regard to the failure load reached.

Primary extrahepatic alveolar echinococcosis of the lumbar spine and the psoas muscle.

Alveolar echinococcosis (AE) of human being caused by Echinococcus multilocularis is a rare but important zoonosis especially in tempered zones of middle Europe and Northern America with endemic character in many countries. Due to the long incubation period, various clinical manifestations, critical prognosis, and outcome AE presents a serious and severe disease. The primary focus of infection is usually the liver. Although secondary affection of visceral organs is possible extrahepatic AE is highly uncommon. Moreover, the involvement of bone and muscle presents with an even lower incidence. In the literature numerous cases on hepatic AE have been reported. However, extrahepatic AE involving bones and/or muscles was described very rarely. We report a case of an 80-year-old man with primary extrahepatic alveolar Echinococcosis of the lumbar spine and the psoas muscle. The etiology, diagnosis, differential diagnoses, treatment options and outcome of this rare disease are discussed in context with the current literature.