Magnetic Resonance Imaging Sequence Identification Using a Metadata Learning Approach

Abstract

Despite the wide application of the magnetic resonance imaging (MRI) technique, there are no widely used standards on naming and describing MRI sequences. The absence of consistent naming conventions presents a major challenge in automating image processing since most MRI software require a priori knowledge of the type of the MRI sequences to be processed. This issue becomes increasingly critical with the current efforts toward open-sharing of MRI data in the neuroscience community. This manuscript reports an MRI sequence detection method using imaging metadata and a supervised machine learning technique. Three datasets from the Brain Center for Ontario Data Exploration (Brain-CODE) data platform, each involving MRI data from multiple research institutes, are used to build and test our model. The preliminary results show that a random forest model can be trained to accurately identify MRI sequence types, and to recognize MRI scans that do not belong to any of the known sequence types. Therefore the proposed approach can be used to automate processing of MRI data that involves a large number of variations in sequence names, and to help standardize sequence naming in ongoing data collections. This study highlights the potential of the machine learning approaches in helping manage health data.

Automatic magnetic resonance imaging series labelling for large repositories

“…The mined information not only needs to consider the requirements of the automatic analysis but also the demands of potential users that intend to perform a search in the image database. Related works propose other DICOM metadata models [15], [16] but the labels used are not enough to be useful in a real clinical environment. The absence of weightings classes such as DCE or information around the existence of fat suppression supposes a lack in the utility of the AI models as data mining tools.…”

“…The features used in the ML models developed in this study are a combination of those used in related works [15], [16] and others proposed by experienced radiologists. Only features automatically generated by machines are included, which have a high percentage of occurrence in the data used (with less than 15% of missing values in the global of all the series).…”

Automated classification of body MRI sequence type using convolutional neural networks

“…Prior work in this area has mostly focused on classifying the series type in brain MRI studies. [3][4][5][6] Mello et al 5 achieved a classification accuracy of 99.27% with the ResNet-18 architefcture trained on 3D brain MRI volumes. MRI series along with DICOM header information was used by Liang et al 3 to achieve near-perfect classification metrics for brain MRI sequence identification.…”

“…[3][4][5][6] Mello et al 5 achieved a classification accuracy of 99.27% with the ResNet-18 architefcture trained on 3D brain MRI volumes. MRI series along with DICOM header information was used by Liang et al 3 to achieve near-perfect classification metrics for brain MRI sequence identification. However, this approach was still reliant on DICOM header data for proper categorization, and this may not always be readily available.…”

“…3 without relying on potentially inaccurate DICOM header information. Of the 11,160 aggregate classifications across the 5-fold cross validation, 22 DWI volumes were incorrectly classified as T2FS.…”