Cross-Center Validation of Deep Learning Model for Muskuloskeletal Fracture Detection in Radiographic Imaging: A Feasibility Study

KVAK, Daniel, Robert HRUBÝ, Jakub DANDÁR, Anora ATAKHANOVA, Matěj MISAŘ a Daniel DUFEK. Cross-Center Validation of Deep Learning Model for Muskuloskeletal Fracture Detection in Radiographic Imaging: A Feasibility Study. In European Congress of Radiology 2024, Vienna. 2024.

Základní údaje

• Originální název: Cross-Center Validation of Deep Learning Model for Muskuloskeletal Fracture Detection in Radiographic Imaging: A Feasibility Study
• Autoři: KVAK, Daniel, Robert HRUBÝ, Jakub DANDÁR, Anora ATAKHANOVA, Matěj MISAŘ a Daniel DUFEK.
• Vydání: European Congress of Radiology 2024, Vienna, 2024.

Další údaje

• Typ výsledku: Konferenční abstrakt
• Utajení: není předmětem státního či obchodního tajemství
• Klíčová slova anglicky: Artificial Intelligence, Musculoskeletal bone, CAD, Plain radiographic studies, CAD, Acute
• Příznaky: Mezinárodní význam, Recenzováno

Anotace

Bone fractures, defined as disruptions in bone continuity, arise from diverse etiologies including trauma, stress injuries, and pathological conditions such as osteoporosis [1]. These fractures, irrespective of the patient's age, manifest a spectrum of clinical presentations ranging from mild symptoms like pain and swelling to severe complications including deformity and functional impairment of the impacted region. Clinically, fracture diagnosis incorporates an injury history, physical examination, and symptom evaluation, notably pain, edema, morphological alterations, abnormal mobility, and occasionally crepitus – a palpable or audible friction in the fractured bone. However, symptomatology in certain fracture types, such as closed or stress fractures, might be subtle or non-specific. Diagnostic approaches for fracture detection and confirmation predominantly utilize X-ray imaging as the initial modality, offering detailed bone structure visualization and fracture identification. Supplementary imaging techniques, including computed tomography (CT), magnetic resonance imaging (MRI), or ultrasound, are employed for complex cases or specific fracture types where musculoskeletal (MSK) X-rays are inadequate. Fracture management strategies depend on fracture characteristics like type, location, and severity, ranging from non-invasive treatments like splint immobilization to surgical interventions using internal osteosynthesis with screws, pins, wires, or external fixation devices. Rehabilitation, encompassing physical therapy and exercises, is crucial for restoring function and strength to the affected region. In the realm of clinical decision-making, the integration of deep learning-based decision support software for radiograph interpretation marks a significant advancement. Despite promising developments, existing algorithms for fracture detection face practical limitations, including the inability to analyze all body parts simultaneously or detect multiple fractures in a single X-ray, which are common in clinical practice. Leveraging sophisticated deep learning models, such software enhances fracture detection accuracy on radiographs, alleviates the diagnostic workload of radiologists, and augments patient outcomes through expedited and precise diagnostic processes [2].