Impact of Artificial Intelligence Detecting Fractures in the Emergence Department : a Pragmatic Prospective Study

Impact of Artificial Intelligence Detecting Fractures in the Emergence Department : a Pragmatic Prospective Study

Impact of Artificial Intelligence Detecting Fractures in the Emergence Department : a Pragmatic Prospective Study

Traumatic emergencies are the primary reason for consultation in emergency departments and standard radiography is the primary imaging exam for osteoarticular trauma. However, with the increase in the number of patients admitted to emergency departments and thus the increased workload for emergency room attendants, Interpretation of radiographs in trauma emergencies is made more difficult, resulting in a high risk of misinterpretation. The growing presence of artificial intelligence in the medical field, notably through the involvement of diagnostic software on imageries, makes its use more relevant in the aid of the replay of osteoarticular imageries. A recent meta-analysis of 32 studies evaluating the performance of artificial intelligence in fracture detection found comparable performance between experienced radiologists and AI-based diagnosis. However, these were mainly retrospective studies, and thus more distant from the reality of its use in a care stream such as emergencies. The objective of this study is therefore to prospectively validate the use of artificial intelligence software during its implementation in an emergency department for patients admitted for a suspicion of osteoarticular trauma.

After completing the X-ray requested, the senior emergency physician reads the X-ray (native image only), gives his diagnosis (fracture yes/ no and localization) and reports if he requests the specialist (orthopedist/ radiologist) or not and the reason why he asks for it (urgent management, doubt about a fracture). After processing the radiography with the "Boneview" software, the emergency physician makes a second reading taking into account the analysis of artificial intelligence. It indicates its result and its decision in the same way. Then, he performs the practical management of the patient: specialized call, exit, urgent management. During the systematic rereading of the radiographs made in the emergency department the night before and night by the radiology intern, he makes a first reading of the native images and gives the result (fracture yes/ no and localization). Then, he makes a second reading assisted by "Boneview" and gives the result again. A radiologist specialized in osteoarticular imaging will read the radiographs initially native. He will note the result (fracture yes/ no and localization) then read the radiographs annotated by the software. He will give the result again. For the patient, there will be no additional imaging exams, but the x-ray will be read a secondary time with artificial intelligence assistance. This pet reading causes a change in patient management. Indeed if a discrepancy is noted between the reading of the radiography by the internal radiology or the conclusion of the emergency physician and the reading of the senior radiologist specialized in osteoarticular imaging, the patient will be recalled and reconvoked to the emergency department if necessary

Inclusion Criteria: Major patient Admitted to the emergency department after trauma less than 48 hours Patient with an indication on an x-ray of the limbs or/and pelvis Express patient consent Affiliated patient or social security beneficiary Exclusion Criteria: Polytraumatized patient X-ray of the corso-lumbar spine, skull, cervical spine (all parts of the body not affected by the intended use of the software) Pregnant, lactating or parturient patient