Automatic method for computing radiographic parameters of radial metaphyseal fractures in radiographs for surgical decision support

• DOI: 10.29007/phsh
• Publication Date: 2022-12-13T23:10:12Z
• Authors: Avigail Suna; Amit Davidson; Leo Joskowicz; Yoram Weil
• Keywords: Landmark; Ground truth

Purpose Distal radius fractures (DRF) are common types of with a high incident rate. DRF can be treated either by cast or surgery. To determine the clinical procedure and operative management, standardized guidelines have become increasingly common. As indications controversial, radiographic parameters (RPs) provide objective support for effective decision making. Calculating RPs manually from radiographs is time consuming subject to observer variability clinician experience. Our aim was develop an automatic method accurately reliably computing 10 associated in anteroposterior (AP) lateral fractured hand without cast. Methods The inputs AP outputs RP values composite images showing landmark points axes used computation on radiographs. comprises three main steps: 1) segmentation ulna deep learning radiograph pixel classifier; 2) axis extraction segmentations using geometric model-based methods; 3) landmarks generation images. study tested accuracy step 2. dataset consists 20 pairs Ground truth were performed expert co-author. located annotated two co-authors. computed considered accurate (in range) when its value inside inter intra range manual annotation. overall measurements obtained averaging each RP. Results 92.7%. Radial Length Shift within range; Angle, Ulnar Variance Step all cases except one outlier; Gap has outlier cases. 100%: four Palmer Tilt, Dorsal Shift, Gap, variability. Conclusion Automatic distal hands accurately. Precise consistent measurement may improve making process.