Systems and methods for 3D image distification
摘要
說明書
At block 816, the 3D image(s) obtained at block 814 would be standardized and normalized to a format similar to the 2D image(s). For example, as for the 2D images, if multiple groups of 3D image(s) were obtained that had differing resolutions, then the 3D images could be resized and normalized as described above. In some embodiments, resizing and normalizing the images could require normalizing to the same resolution as the 2D images, e.g., to 640×480, if normalizing to resolution of the 2D images as described above. In other embodiments, the resizing and normalization could be involve using different resolutions or formats than for the 2D images, such as a similar resolution (720×576) or a different format that represented a similar resolution or format, and that still allowed for comparison or interoperability with the 2D images. For example, the 3D images could be Distified, as disclosed herein, which would provide standardization and normalization for the 3D images, for example, by using the output feature vectors in a 3D prediction model for use and comparison in the ensemble model with the 2D images, as describe herein.
In addition to providing interoperability for comparison purposes between the 2D and 3D images, standardization (e.g., via normalizing, downsizing, or Distification, etc.), as described in the foregoing embodiments, can also be useful in improving the performance of the ensemble predictive model because processing the smaller resolution images would require less computer resources and fewer processing cycles than compared with executing the ensemble model on the raw images alone. Other standardization and normalization procedures may additionally be used, such that the 2D images and 3D images become comparable or otherwise interoperable for purposes of use in the disclosed ensemble predictive model.
