The intracranial aneurysms rupture can cause serious stroke, which relates to decline of daily life ability in the elderly. Although deep leaning is now the most successful solution for the organ detection, it requires myriads of training data, the consistency of image format and the balance of the sample distribution. This research innovatively presents the intracranial aneurysm detection problem as a shape analysis problem rather than a computer vision problem. We detect the intracranial aneurysms in 3D cerebrovascular mesh models after the segmentation of the brain vessel from the medical images, which can break the barriers of the data format and data distribution, serving both in clinical and screening. Also we propose a transferable multimodel ensemble (MMEN) architecture to detect the intracranial aneurysms from cerebrovascular mesh models with limited data. To obtain a well-defined convolution operator, we use the global seamless parameterization converting a 3D cerebrovascular mesh model to a planar flat-torus. In the architecture, we transfer the planar flat-torus presentation abilities of three GoogleNet Inception V3 models, which were pre-trained on the ImageNet database, to characterize the intracranial aneurysms with local and global geometric features such as Gaussian curvature, shape diameter function and WKS, respectively. We jointly utilize all these five models to detect aneurysms with adaptive weights learning based on back propagation. Experimental results on the 121 models show that our proposed method can achieve detection accuracy of 95.1% with 94.7% F1 score and 94.8% sensitivity, which is as good as the state-of-art work but applicable to inhomogeneous image modalities and smaller datasets.