Machine learning in MRI

Machine learning (ML) represents a subfield within artificial intelligence (AI) and involves finding patterns and regularities in a large amount of training data. A special machine learning method that uses multilayer artificial neural networks is called "deep learning" and has developed rapidly in recent years.

Figure 1: Intensity curves of target regions (orange, red, pink) in time-resolved MRI images.

At our institute, different machine learning methods are used to improve different aspects of MRI measurements. The main fields of application are the optimization of MRI acquisition parameters (keyword: "cognitive sensing"), the improvement of image reconstruction by deep neural networks as well as the automated analysis of the acquired images (e.g. segmentation or detection of landmarks). In addition to "deep learning", model-based learning ("physics-informed learning") is also applied in order to reduce the complexity of the learning process by introducing prior knowledge about physical laws.