Efficient Extraction of Pathologies from C-Spine Radiology Reports using Multi-Task Learning

TL;DR

This paper demonstrates that a multi-task learning approach with Transformer models can outperform or match multiple specialized models in extracting pathologies from cervical spine radiology reports, showing promise for broader medical report analysis.

Contribution

The study introduces a multi-task learning framework that effectively leverages related tasks to improve pathology extraction from radiology reports, outperforming traditional task-specific models.

Findings

Multi-task model outperforms individual models on cervical spine reports.

Tasks are semantically related, enhancing multi-task learning effectiveness.

Method has potential for application to reports of other body parts.

Abstract

Pretrained Transformer based models finetuned on domain specific corpora have changed the landscape of NLP. Generally, if one has multiple tasks on a given dataset, one may finetune different models or use task specific adapters. In this work, we show that a multi-task model can beat or achieve the performance of multiple BERT-based models finetuned on various tasks and various task specific adapter augmented BERT-based models. We validate our method on our internal radiologist's report dataset on cervical spine. We hypothesize that the tasks are semantically close and related and thus multitask learners are powerful classifiers. Our work opens the scope of using our method to radiologist's reports on various body parts.