Fig. 10From: Automatic detection on intracranial aneurysm from digital subtraction angiography with cascade convolutional neural networksThe architecture of aneurysm detection stage (ADS). The annotation of this figure was identical to Fig. 9. In addition, ‘\({t}\times\)’ on the top of brace denoted t times repeat of these layers. The dual-input layer concatenated the target PCoA region image and the corresponding PCoA region image, following with an FPN to cover various scales of aneurysms. Then, the anchor boxes output prediction results. The RAGS algorithm followed to suppress false-positive casesBack to article page