A scalable physician-level deep learning algorithm detects universal trauma on pelvic radiographs

Abstract

Pelvic radiograph (PXR) is essential for detecting proximal femur and pelvis injuries in trauma patients, which is also the key component for trauma survey. None of the currently available algorithms can accurately detect all kinds of trauma-related radiographic findings on PXRs. Here, we show a universal algorithm can detect most types of trauma-related radiographic findings on PXRs. We develop a multiscale deep learning algorithm called PelviXNet trained with 5204 PXRs with weakly supervised point annotation. PelviXNet yields an area under the receiver operating characteristic curve (AUROC) of 0.973 (95% CI, 0.960–0.983) and an area under the precision-recall curve (AUPRC) of 0.963 (95% CI, 0.948–0.974) in the clinical population test set of 1888 PXRs. The accuracy, sensitivity, and specificity at the cutoff value are 0.924 (95% CI, 0.912–0.936), 0.908 (95% CI, 0.885–0.908), and 0.932 (95% CI, 0.919–0.946), respectively. PelviXNet demonstrates comparable performance with radiologists and orthopedics in detecting pelvic and hip fractures.

Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning

“…We exploited the excellent spatial resolution of radiographs to identify hip implants and unsuspected fragility fractures before estimation of BMD. The tool incorporates our previously published PelviXNet 34 to detect hip fracture and newly developed algorithms to detect hip implants. Furthermore, we developed a VCF assessment algorithm based on a Deep Adaptive Graph network (DAG) 35 , which determines anatomical landmarks for standard six-point vertebral morphometry that facilitates VCF detection using the widely accepted semiquantitative Genant visual method 36 , 37 .…”

“…We detect hip fracture and implant (joint prosthesis, screws, plates, or cement) in the quality assessment process and exclude them from the downstream BMD estimation. An existing model, PelviXNet 34 , is used to detect the hip fracture. PelviXNet consists of a DensetNet-121 backbone neural network and a Feature Pyramid Network and was trained on 5204 pelvic radiographs that had been annotated by experienced physicians using an efficient and flexible point-based annotation scheme.…”

Automatic Hip Detection in Anteroposterior Pelvic Radiographs—A Labelless Practical Framework

“…Medical artificial intelligence is progressive in order to change the healthcare system, and various DCNNs have showed that it is feasible to detect lesions from pathologic images [23] and radiography [24]. These algorithms presented outstanding achievement in disease detection or prediction of whose performance is not inferior to that of the physicians [23][24][25].…”

How Can a Deep Learning Algorithm Improve Fracture Detection on X-rays in the Emergency Room?

“…Kitamura [ 20 ] reported a sensitivity of 0.86 for detecting pelvic fractures. Cheng [ 21 ] recently reported a sensitivity of 92% for detecting pelvic fractures. Jones [ 22 ] reported a sensitivity of 95.2% for detecting fractures throughout the peripheral skeleton, a performance close to that of the algorithm we tested.…”