In a recent study published in the JAMA Network Open Journal, researchers in Taiwan utilized a deep-learning (DL) algorithm to aid novices in focused assessment with sonography in trauma (FAST).
Study: Use of a Deep-Learning Algorithm to Guide Novices in Performing Focused Assessment With Sonography in Trauma. Image Credit: PanchenkoVladimir/Shutterstock.com
FAST may contribute to quicker access to surgical treatment, shorter stays in the hospital, and lower out-of-pocket expenses. Using prehospital FAST was linked to changes in hospitalization in 22% of patients in a 2005 survey.
The FAST evaluation is regarded as one of the most challenging procedures for image interpretation in ultrasonography, which is extremely operator-dependent.
About the study
In the present study, researchers investigated the relationship between artificial intelligence (AI) instruction and the FAST performance quality of inexperienced ultrasonography operators.
This quality enhancement research was conducted between March 20 and April 20, 2022. A total of 30 operators, including 10 nurse practitioners (NPs), 10 emergency medical technicians (EMTs), and 10 registered nurses, were randomly chosen and assigned to one of two research groups: those who received AI instruction and those who did not.
The team incorporated the DL-based guidance algorithm into an application that collected pictures from ultrasonography devices and offered real-time quality input to allow AI guidance. This DL algorithm provided qualified and non-qualified images with an accuracy of 0.941.
The operators were instructed to perform FAST examination in 10 healthy patients to gain a five-second clip of the normal view in three minutes.
The main outcome was diagnostic accuracy, which was separately evaluated by three highly qualified echocardiographers. A scale from one to five was used to evaluate the diagnostic quality, with greater results indicating higher quality. A quality rating of four or greater was deemed adequate for clinical use. The time needed to complete the examination was the supplementary outcome.
A total of 30 operators produced 300 ultrasonography images in total. The sample group comprised five men and five women, with a body mass index (BMI) between 21.5 and 27.8 and a median age of 46.
The diagnostic quality ratings' intraclass correlation value was 0.97, indicating high dependability. Operators with AI assistance had better median and mean rates of acceptable quality scores than operators without AI help.
Additionally, a better quality score and percentage of appropriate quality were linked to AI instruction.
Longer examination times were linked to AI assistance, mostly seen in the initial stages of procedures. In a subgroup study of operators' employment, NPs and EMTs who received AI instruction had higher-quality diagnostic results.
The study findings demonstrated that the DL algorithm could assist beginners in obtaining accurate diagnostic pictures. The diagnostic quality score and the percentage of approved recordings were considerably higher with AI assistance.
It might initially take longer to finish a test with AI assistance; however, it was anticipated that the learning curve would be less steep for beginners using FAST.
The researchers believe that future study on therapeutic applications is required to completely assess the possible advantages of using AI in treating patients with traumatic injuries.