Machine Learning is becoming a key component in most image processing workflows. By learning from existing data, it allows to create and use complex algorithms without needing to implement hand-crafted rules. The ImFusion Suite has been designed to exploit such algorithms and make them easily usable by any user - even the most advanced ones like Deep Learning.
Creating a training dataset is easy to do with the ImFusion Suite, thanks to its extensive data support that allows to load the most common image formats, as well as existing annotations like binary masks or meshes. With the interactive algorithms for segmentation and registration, the user can directly annotate any image, whether it is in 2D, 3D, or has multiple channels. It is also possible to artificially generate new data by applying random geometric perturbations (both rigid transformations and deformations) as well as intensity transformations.
Our framework supports all kinds of machine learning predictions: both multi-label classification and regression tasks, either at the pixel level or on whole images. Both Neural Networks and Random Forests can be trained directly from our software, with a user-friendly interface. Once learnt, these models can directly be applied to any image loaded or streamed to the ImFusion Suite. They can also be used to enhance or complement our segmentation and registration algorithms, allowing for completely automated workflows.
At ImFusion, we make extensive use of machine learning and have successfully applied it to a number of challenging problems, for instance:
We offer to leverage our great experience with machine learning to get the most value out of your data, and speed up your developments by integrating machine learning into your workflow.
 M. Salehi, R. Prevost, J.-L. Moctezuma, N. Navab, W. Wein. Precise Ultrasound Bone Registration with Learning-based Segmentation and Speed of Sound Calibration.
 R. Prevost, M. Salehi, J. Sprung, R. Bauer, A. Ladikos, W. Wein. Deep Learning for Sensorless 3D Freehand Ultrasound Imaging.