In the field of medical imaging diagnostics, the detection of cerebral aneurysms has long been a challenge. However, a recently developed deep learning-based model has successfully provided radiologists with a powerful auxiliary tool. This technology not only enhances the detection rate of cerebral aneurysms but also significantly reduces the time required for image interpretation and post-processing. Researchers indicate that such tools hold tremendous potential in improving clinical workflows and enhancing the diagnosis of cerebral aneurysms.
The timely and accurate diagnosis of cerebral aneurysms is crucial for initiating appropriate management strategies, optimizing patient outcomes, and mitigating the impact of this condition on individuals and the healthcare system. Therefore, the development of efficient diagnostic tools is尤为 important.
Image source: Picture generated by AI, image authorized by service provider Midjourney
Under the leadership of Dr. Jianing Wang from the Radiology Department at Hebei University Affiliated Hospital, researchers trained the model on data from nearly 4,000 patients and tested it on an additional 484 patients. During the analysis, the research team had 10 radiologists interpret each case with and without the model's assistance, along with additional evaluations to review the model's standalone performance.
When radiologists used this tool, the time for interpretation and post-processing was reduced by 37.2% and 90.8%, respectively. For junior radiologists, the model's assistance increased the AUC (Area Under the Curve) from 0.842 to 0.881; for senior radiologists, from 0.853 to 0.895. Sensitivity at both the lesion and patient levels also improved with deep learning assistance, and specificity at the patient level was enhanced.
Considering the complexity of intracranial vasculature, aneurysm detection based on CTA (Computed Tomography Angiography) is a time-consuming and challenging task. Additionally, the increasing demand for CTA examinations may lead to radiologist fatigue, which, coupled with the subjective nature of image interpretation, often affects diagnostic accuracy.
The research team added that their tool provides evidence supporting that deep learning-based models can adapt to different examinations, as their model was accurate across a wide range of tests. This addresses the generalizability issues commonly faced by deep learning tools. Similar models may be particularly beneficial for less experienced readers in environments where timely diagnosis is critical.