Researchradiographer
Simon Lysdahlgaard
Imaging Research Initiative Southwest, Department of Radiology and Nuclear Medicine, University Hospital of Southern Denmark, Esbjerg, Denmark
| Project management | ||
| Project status | Open | |
| Data collection dates | ||
| Start | 01.01.2022 | |
| End | 31.12.2023 | |
Radiographers and deep learning algorithms compared to reporting radiographers when rejecting lateral knee x-rays
Short summary
Radiologists and radiographers disagree on image quality, leading to extra lateral knee x-rays, increasing radiation exposure, and risk. The ALARA principle seeks to minimize this. Today's computer vision tech, specifically convolutional neural networks (CNNs), can assist in medical imaging classification. Training a CNN using a radiologist as a reference could aid radiographers in making informed rejections, minimizing unnecessary patient exposure.
Rationale
Ionizing radiation (e.g. x-rays) increase the risk of long-term stochastic effects, such as cancer and cell mutations. Radiographers need to satisfy diagnostic and technical criteria and keep patient exposure As Low As Reasonable Achievable (ALARA). Radiographers assess the images for proper positioning, adequate exposure, patient motion blur and other quality defects that could potentially affect the diagnosis before images are released to radiologists for interpretation. Whaley et al. (1) found that radiographers and radiologists only moderately agreed in their perceptions of image quality evaluation. Radiographers rated the images on technical attributes, whereas radiologists tended to be more accepting and rated images on their diagnostic capability. A study by Mount et al. (2) explored the impact of differential opinions between practitioners on potential rejection rates of lateral knee x-rays. The study found that differences in perceptions of image quality directly influenced rejection rates with radiologists accepting far more images than radiographers. The rationale or background of this scenario lies in the quest to improve the decision-making process in radiography, specifically in determining the quality of lateral knee x-rays. The perception of image quality and the criteria for rejecting an x-ray vary between radiologists and radiographers, leading to unnecessary rejections and thus, unnecessary radiation exposure for patients.
Description of the cohort
Patients referred from general practitioners to knee x-rays for the diagnosis of osteoarthritis will be included. Patients with severe stages of osteoarthritis or differential diagnoses, any metal or prosthesis implant and under the age of 18 will be excluded.
Data and biological material
A dataset will be obtained retrospectively from all available x-ray rooms in the Department of Radiology and Nuclear Medicine at the Hospital of South West Jutland in the period from 2010-2020. Images are extracted from the Picture Archiving and Communication System as the DICOM-format. An extra dataset of 200 lateral knee x-rays will be retrospectively collected during the period of 2022.