Developing an artificial intelligence diagnostic tool for paediatric distal radius fractures, a proof of concept study.

INTRODUCTION: In the UK, 1 in 50 children sustain a fractured bone yearly, yet studies have shown that 34% of children sustaining an injury do not have a visible fracture on initial radiographs. Wrist fractures are particularly difficult to identify because the growth plate poses diagnostic challenges when interpreting radiographs. METHODS: We developed Convolutional Neural Network (CNN) image recognition software to detect fractures in radiographs from children. A consecutive data set of 5,000 radiographs of the distal radius in children aged less than 19 years from 2014 to 2019 was used to train the CNN. In addition, transfer learning from a VGG16 CNN pretrained on non-radiological images was applied to improve generalisation of the network and the classification of radiographs. Hyperparameter tuning techniques were used to compare the model with the radiology reports that accompanied the original images to determine diagnostic test accuracy. RESULTS: The training set consisted of 2,881 radiographs with a fracture and 1,571 without; 548 radiographs were outliers. With additional augmentation, the final data set consisted of 15,498 images. The data set was randomly split into three subsets: training (70%), validation (10%) and test (20%). After training for 20 epochs, the diagnostic test accuracy was 85%. DISCUSSION: A CNN model is feasible in diagnosing paediatric wrist fractures. We demonstrated that this application could be utilised as a tool for improving diagnostic accuracy. Future work would involve developing automated treatment pathways for diagnosis, reducing unnecessary hospital visits and allowing staff redeployment to other areas.

Incidence of hyperoxia in trauma patients receiving pre-hospital emergency anaesthesia: results of a 5-year retrospective analysis.

BACKGROUND: Previous studies have demonstrated an association between hyperoxia and increased mortality in various patient groups. Critically unwell and injured patients are routinely given high concentration oxygen in the pre-hospital phase of care. We aim to investigate the incidence of hyperoxia in major trauma patients receiving pre-hospital emergency anesthesia (PHEA) in the pre-hospital setting and determine factors that may help guide clinicians with pre-hospital oxygen administration in these patients. METHODS: A retrospective cohort study was performed of all patients who received PHEA by a single helicopter emergency medical service (HEMS) between 1 October 2014 and 1 May 2019 and who were subsequently transferred to one major trauma centre (MTC). Patient and treatment factors were collected from the electronic patient records of the HEMS service and the MTC. Hyperoxia was defined as a PaO2 > 16 kPA on the first arterial blood gas analysis upon arrival in the MTC. RESULTS: On arrival in the MTC, the majority of the patients (90/147, 61.2%) had severe hyperoxia, whereas 30 patients (20.4%) had mild hyperoxia and 26 patients (19.7%) had normoxia. Only 1 patient (0.7%) had hypoxia. The median PaO2 on the first arterial blood gas analysis (ABGA) after HEMS handover was 36.7 [IQR 18.5-52.2] kPa, with a range of 7.0-86.0 kPa. SpO2 pulse oximetry readings before handover were independently associated with the presence of hyperoxia. An SpO2 ≥ 97% was associated with a significantly increased odds of hyperoxia (OR 3.99 [1.58-10.08]), and had a sensitivity of 86.7% [79.1-92.4], a specificity of 37.9% [20.7-57.8], a positive predictive value of 84.5% [70.2-87.9] and a negative predictive value of 42.3% [27.4-58.7] for the presence of hyperoxemia. CONCLUSION: Trauma patients who have undergone PHEA often have profound hyperoxemia upon arrival at hospital. In the pre-hospital setting, where arterial blood gas analysis is not readily available a titrated approach to oxygen therapy should be considered to reduce the incidence of potentially harmful tissue hyperoxia.