Synopsis

Timely and accurate MR protocoling is important to ensure best efficiency and diagnostic value in radiology departments. We propose and validate an artificial intelligence based natural language classifier that can assign MR abdomen/pelvis protocols based on free-text clinical indications. We achieve an overall classification accuracy rate of 93% on a test set consisting of 83 free-text clinical indications.

Background

Timely and accurate MR protocoling is important to ensure best efficiency and diagnostic value in radiology departments. Unfortunately, MR abdomen/pelvis protocol assignments remain variable and subjective for requesting clinicians and even among radiologists. The purpose of this study is to develop and validate an artificial intelligence based natural language classifier that can assign MR abdomen/pelvis protocols based on free-text clinical indications.

Materials & Method

253 free-text clinical indications and final assigned MR abdomen & pelvis protocols were retrospectively retrieved from an in-house radiology communication tool. Each entry was manually confirmed to ensure dissociation from any protected health information. The MR protocols were divided into 14 categories by consensus among authors. The dataset was split into 170 training set and 83 test set. Supervised machine learning was performed via a natural language classifier cloud service that undergoes series of string analysis, hypothesis generation, hypothesis & evidence scoring, and final merging & ranking. Accuracy was validated via the test set.

Results

Three most common indications for MR of abdomen and/or pelvis were evaluations for choledocholithiasis (n=67, 26.5%), appendicitis (n=41, 16.2%), and liver mass (n=24, 9.5%). Most common protocols were MRI/MRCP abdomen with gadoxetate disodium protocol (n=92, 36.4%). Training time for generating the classifier took 12 minutes and 34 seconds. Testing time took <0.1 second. The final classifier had an overall accuracy rate of 93% (77 out of 83). Examples of correct and incorrect classifications are given in Figure 1 and 2, respectively.

Conclusion

We successfully developed and validated an artificial intelligence based, clinical decision support tool to recommend MR protocols for abdomen and/or pelvis, with an accuracy rate of 93%.

Acknowledgements

References