Identifying Incidental Findings from Radiology Reports of Trauma Patients

Incidental findings are the findings that are incidentally discovered in the context of radiology diagnostics which can potentially affect the health of an individual. (Thomas, 2016)

Incidental findings can be a life-saver however they can also be harmful. Minor abnormalities sometimes could be an indicator to some serious diseases. A seemingly harmful lesion can lead to lifelong follow-up, further imaging and appointments, unwarranted treatment, and even radical surgery, only for the pathology to turn out to be benign.

Gaurav Trivedi, a Phd student at University of Pittsburgh, developed and evaluated an automated pipeline to identify incidental findings in radiology reports of trauma patients at the sentence and section levels using a variety of feature representations.

“Text + Concept” Features

1. For text, the “Bigram” language model is used in feature representation.
2. For concept, SNOMED-CT (Systematized Nomenclature of Medicine — Clinical Terms) provides the knowledge base.

Embeddings

1. One-hot encoding: is used for categorical variables.
2. Dense vectors

Models

1. Traditional Machine Learning Methods: Naive Bayes, Random Forest, Logistic Regression, SVMs
2. Deep Learning Methods: CNN-based deep learning techniques

Result

The best performance was achieved by using CNNs with Pre-trained embedding at both sentence and section levels. This provides evidence that such a pipeline is likely to be clinically useful to identify incidental findings in radiology reports in trauma patients