Year: 2025
Type: other
Abstract: Hip dysplasia is a complex orthopaedic condition characterized by an improperly shaped acetabulum, leading to structural and anatomical alterations in the hip joint. This creates a static problem, that compromises the joint's, resulting in hip instability due to abnormal load transmission and anomalous concentration of forces within the weight-bearing zone. To detect this condition using X-rays the key indicators include: displacement of the femoral head with its coverage percentage, lateral center-edge (LCE) angle of Wiberg, acetabular roof angle of Tönnis, acetabular index (AI), acetabular depth-to-width ratio, acetabular angle of Sharp and the anterior center-edge angle of Lequesne. These measurements are detected manually in radiographic images, which may lead to variability in the interpretations of the results obtained among doctors of the same specialty and experience levels. Therefore, we present the development of a software that, through artificial intelligence, identifies anatomical positions in hip radiographs and, with mathematical relations, facilitates the interpretation and diagnosis of this condition as a complementary tool. Primarily it is focused on evaluating the angle of Wiberg. Future enhancements are intended to include the detection of additional clinical indicators, improving its diagnostic capabilities for this condition. The model is based on a convolutional neural network (CNN) that has demonstrated good accuracy in identifying positions, achieving a high confidence rate in 80% of the tests performed. While initial results are promising, expanding the training dataset would improve the model’s performance and robustness contributing to the advancement of automated diagnosis in traumatology. Furthermore, the application increases accessibility by not restricting users to specific medical formats like DICOM files, thus enabling broader adaptability and facilitating greater integration in different environments.
Authors: Fernando Pocino Martín, Paula Méndez de Inza, Luis Novoa Rodríguez, Alberto de Santos Sierra
Institution: Universidad Francisco de Vitoria
Language: English
Topic: Medical Imaging and Analysis
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4413905404