| Abstract: | Background and purpose — A correct diagnosis is essential for the appropriate treatment of patients with atypical femoral fractures (AFFs). The diagnostic accuracy of radiographs with standard radiology reports is very poor. We derived a diagnostic algorithm that uses deep neural networks to enable clinicians to discriminate AFFs from normal femur fractures (NFFs) on conventional radiographs.Patients and methods — We entered 433 radiographs from 149 patients with complete AFF and 549 radiographs from 224 patients with NFF into a convolutional neural network (CNN) that acts as a core classifier in an automated pathway and a manual intervention pathway (manual improvement of image orientation). We tested several deep neural network structures (i.e., VGG19, InceptionV3, and ResNet) to identify the network with the highest diagnostic accuracy for distinguishing AFF from NFF. We applied a transfer learning technique and used 5-fold cross-validation and class activation mapping to evaluate the diagnostic accuracy.Results — In the automated pathway, ResNet50 had the highest diagnostic accuracy, with a mean of 91% (SD 1.3), as compared with 83% (SD 1.6) for VGG19, and 89% (SD 2.5) for InceptionV3. The corresponding accuracy levels for the intervention pathway were 94% (SD 2.0), 92% (2.7), and 93% (3.7), respectively. With regards to sensitivity and specificity, ResNet outperformed the other networks with a mean AUC (area under the curve) value of 0.94 (SD 0.01) and surpassed the accuracy of clinical diagnostics.Interpretation — Artificial intelligence systems show excellent diagnostic accuracies for the rare fracture type of AFF in an experimental setting.Atypical fractures occur at atypical locations in the femoral bone and show a strong association with bisphosphonate treatment (Odvina et al. 2005, 2010, Shane 2010, Shane et al. 2010, Schilcher et al. 2011, 2015, Starr et al. 2018). In contrast to the metaphyseal area, which is the site for the majority of all fragility fractures, the diaphyseal region is where atypical fractures occur. As is the case for any other insufficiency-type fracture of the diaphysis, atypical fractures show specific radiographic features, such as a transverse or short oblique fracture line in the lateral femoral cortex and focal cortical thickening (Schilcher et al. 2013, Shane et al. 2014). These features differ from those of normal femur fractures (NFFs), which show oblique fracture lines and no signs of focal cortical thickening (Shin et al. 2016b).Early and correct diagnosis of AFF is essential for appropriate management (Bogl et al. 2020a), which minimizes the risk of healing complications (Bogl et al. 2020b). In clinical routine practice, conventional radiographs are used to diagnose complete AFF. However, routine diagnostic accuracy is poor, and < 7% of AFF cases are correctly identified in this way (Harborne et al. 2016).Artificial intelligence (AI), deep learning through convolutional networks, has proven effective in the classification (Russakovsky et al. 2015) and segmentation (Ronneberger et al. 2015) of medical images in general, and for bone fractures in particular (Brett et al. 2009, Olczak et al. 2017, Chung et al. 2018, Kim and MacKinnon 2018, Lindsey et al. 2018, Adams et al. 2019, Urakawa et al. 2019, Kalmet et al. 2020). Given the very specific radiographic pattern of these fractures, AI appears to be a useful tool for finding the needle (AFF) in the haystack (NFF).We evaluated the abilities of different deep neural networks to discriminate complete AFF from NFF on diagnostic plain radiographs in an experimental setting and we assessed the effect of limited user intervention on diagnostic accuracy. |
