Deep Learning System Automates Pipeline Embolization Device Size Selection for Intracranial Aneurysms

An AI decision support system (NeurAneuNet) achieved 95.2% agreement with expert consensus for Pipeline Embolization Device sizing in an independent cohort of 21 intracranial aneurysms, and reduced clinician planning time by 44.8%.

• In a dataset of 600 aneurysms (210 PED-treated), the model achieved 91.8% classification accuracy for device size and a diameter prediction error of 0.24±0.10 mm.
• Aneurysm segmentation Dice coefficient was 0.874±0.03 using a dual-path attention U-Net++.
• When assisting clinicians, the system reduced subjective cognitive workload (NASA-TLX score from 33±8 to 21±5).

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.