Self-Learning Algorithms For Different Imaging Datasets

AI-based evaluation of medical imaging data typically necessitates the development of a unique algorithm for each task. Scientists have now presented a new method for configuring self-learning algorithms for a wide range of imaging datasets – without requiring specialized knowledge or massive computing power.

Researchers have now developed a method that adapts to any imaging dataset dynamically and completely automatically, allowing even people with limited prior expertise to configure self-learning algorithms for specific tasks. The method, known as nnU-Net, can deal with a wide range of imaging data. For example, it can process electron and fluorescence microscopy images in addition to CT and MRI images.

Despite competing against highly specific algorithms developed by experts for specific individual questions, the researchers achieved the best results in 33 out of 53 segmentation tasks in international competitions using nnU-Net.

The team is making nnU-Net available as an open-source tool for free download. nnU-Net can be used right away, trained using imaging datasets, and perform special tasks without requiring special computer science expertise or significant computing power.

AI-based evaluation of medical imaging data has been used primarily in research contexts and is not yet widely used in the routine clinical care of cancer patients. However, medical informatics specialists and physicians see significant potential for self-learning algorithms for highly repetitive tasks frequently required in large-scale clinical studies. nnU-Net can assist in realizing this potential.

Read more

Related Content: OCT Scan To The Next Level