Join our mailing list

Automated diagnosis and classification of metacarpal and phalangeal fractures using a convolutional neural network: a retrospective data analysis study

Michael Axenhus, Anna Wallin, Jonas Havela, Sara Severin, Ablikim Karahan, Max Gordon, Martin Magnéli

Acta Orthopaedica, 96, 2025, 13-18.

In the field of orthopedic diagnostics, the integration of artificial intelligence (AI) is an emerging topic, aiming to enhance accuracy and efficiency in fracture detection. The recent study by Axenhus et al., titled "Automated diagnosis and classification of metacarpal and phalangeal fractures using a convolutional neural network: a retrospective data analysis study," marks a significant advancement in this domain.

Hand fractures, particularly those involving the metacarpal and phalangeal bones, are prevalent injuries encountered in emergency departments. Despite their frequency, diagnostic errors remain a concern, potentially leading to complications and delayed treatments.

Axenhus and colleagues addressed this challenge by developing a convolutional neural network (CNN) model designed to diagnose and classify these fractures using plain radiographs. Their model achieved a mean weighted area under the curve (AUC) of 0.84, with sensitivity and specificity rates of 86% and 76%, respectively. Notably, the model excelled in identifying transverse metacarpal fractures (AUC = 0.91) and tuft phalangeal fractures (AUC = 0.97), demonstrating 100% sensitivity in both categories. These findings underscore the potential of AI-driven tools to augment clinical decision-making in orthopedic practice. Accurately identifying and classifying hand fractures, such models can serve as valuable adjuncts to clinicians, potentially reducing diagnostic errors and improving patient outcomes, but also reducing work-load and the possibility to shift attention to other tasks.

However, the integration of AI into clinical workflows necessitates careful consideration. While the performance metrics are promising, the model's efficacy in diverse clinical settings and its adaptability to varying imaging protocols require further validation. Moreover, the ethical implications of AI in healthcare, including issues of transparency and accountability, must be addressed to foster trust among practitioners and patients alike.

In conclusion, the study by Axenhus et al. represents a step toward the use of AI in orthopedic diagnostics. As technology continues to evolve, collaborative efforts between clinicians and data scientists will be essential to harness the full potential of AI, ultimately enhancing the quality of patient care in orthopedics

Co-editor

Taco Gosens

Vascularized fibular grafting following tumor resection demonstrates acceptable long-term outcomes in Denmark: a national retrospective cohort study

Christian Lind Nielsen, Daniel Thor Halberg Dybdal, Peter Vester-Glowinski, Lisa Lyngsie Hjalgrim, Pernille Edslev Wendtland, Birgitte Jul Kiil, Michael Melchior Bendtsen, Michael Mørk Petersen, Thomas Baad-Hansen

Acta Orthopaedica, 96, 2025, 87–93.

Bone sarcomas, such as osteosarcoma and Ewing sarcoma, are rare primary malignant bone tumors requiring wide or marginal resection for survival. Vascularized fibular grafting (VFG) represents an often used biological reconstruction in orthopedic oncology. Its benefits include size, and shape, and when used as a vascular graft blood supply promoting bony healing via hypertrophy. VFG is typically applied for diaphyseal reconstruction alone or in combination with allograft and for intraarticular reconstruction in the upper extremity. Despite its advantages, there remains a relatively high rate of postoperative complications including delayed union and graft fracture, which may require revision surgery.

Nielsen et al. present a retrospective evaluation of a national Danish cohort comprising 27 patients undergoing biologic reconstruction with VFG. The indications were 13 cases of Ewing sarcoma, 12 cases of osteosarcoma, and 2 cases of giant cell tumor. The median age at surgery was 16 years with median follow-up of 82 months. Osteosynthesis was achieved through internal fixation in all cases. The overall rate of graft union was 67%, with a median time to union of 13 months. 20 patients experienced complications necessitating surgical intervention. 2 patients in the cohort underwent above-the-knee amputations during the follow-up period representing a limb salvage rate of 93%. A transient postoperative peroneal nerve palsy occurred in 7 patients, with all cases resolving spontaneously. The 5-year overall survival rate was 81%. Patients with upper extremity tumors were more likely to attain graft union and were less likely to undergo multiple revisions than patients with lower extremity tumors.

Biologic reconstruction using VFG remains a useful option to reconstruct diaphyseal bony defects in adolescents undergoing surgery for malignant bony tumors even with the development of custom-made endoprostheses. The risk of revision surgery remains relatively high and this risk is smaller in patients undergoing upper extremity (humerus) reconstruction.

Co-editor

Ilkka Helenius