The current study demonstrated the feasibility of using an AI algorithm based on NLP and ML techniques to convert unstructured free-form text from the findings section of radiology reports into separate subheadings with high accuracy (92–96%). The results of the study also highlighted a well-known problem in radiology reporting: problematic statements, some of which pose difficulties to a structuring scheme by necessity (e.g., complicated cases) and some by error (e.g., dictation errors). While not designed for this purpose, the prediction probability feature of the algorithm may have an application in identifying such statements.
Clear, concise, accurate, and reproducible communication is a universally accepted requirement in the clinical practice of medical imaging , and both expert opinion and formal studies have shown that “structured reporting” in its various forms can improve communication between radiologists and referring providers [2, 3, 7, 13], albeit at the perceived cost of lower productivity [10, 13]. The current study demonstrated that new AI applications might be feasible in combining advantages of both free-form reporting (namely, increased radiologist productivity) and structured reporting (namely, improved communication with providers). In addition to immediate clinical benefits, increased structuring could have utility for data mining applications or when layering additional feature extraction on the report, and the algorithm could be retrospectively applied to legacy reports if indicated. For example, identification of the presence of pulmonary embolism via NLP and ML should be made easier if an algorithm only has to search a single section of a report (Pulmonary arteries) rather than the entire text.
Several different computer algorithms have been applied to radiology reporting in the past [11, 12, 14,15,16]. Studies have used a variety of different methods, i.e., combinations of ML, active learning, and NLP. These studies have had varying goals, most commonly identification and highlighting specific findings, such as critical or abnormal findings [11, 14, 16] and presence of cancer , with wide ranges in diagnostic accuracies (range 82–99%) for the given task. Studies that are directly comparable to ours, i.e., conversion of free-text reports to semi-structured reports, are sparse, but early feasibility studies have been promising .
Nearly 7% of the statements from the 400 test reports proved problematic for the manual observers to label. This highlights intrinsic problems in radiology reporting in general as well as some of the arguments against rigid report structuring. The majority of the problematic statements would not be inappropriate in a free-text report per se, i.e., two sections combined into one statement or findings that could reasonably go into one of several headings, but rather became inconclusive when forced into one particular section. This could be reasonably addressed with either a change in dictation culture or more sophisticated rules layered onto the ML algorithm to address these specific situations. However, there were also a minority (0.5%) of statements that were nonsensical such that their meaning could not be determined enough for manual labeling, presumably from dictation errors or typos. Of these 20, only 3 (15%) received a prediction probability of 1 (meaning that the algorithm believed labeling was correct). Of all of the problematic statements, 66% had a prediction probability of less than 1, compared to only 18% of the nonproblematic statements. We believe that with modification, this feature of the algorithm could be used to identify statements within a report that should be reexamined for clarity. Of course, the algorithm was not designed with this feature in mind and the study was not designed to test such a feature, but we believe it warrants further examination.
This study is not without limitations, some of which highlight intrinsic challenges of ML technology. While over 4,000 individual statements were used for both training and testing, many times more will be necessary to optimize the potential of current ML algorithms, particularly in the accurate labeling of rare findings or uncommon statements. Both the training and testing reports were generated by approximately 40 separate radiologists, but all from the same department, and therefore both heterogeneity in individual reporting style and specific institutional nuances are incorporated into the algorithm. While it is likely a positive feature to have variability when training the algorithm, this did decrease labeling accuracy when applying our “strict” criteria. Conversely, institutional colloquialism could limit the generalizability of algorithms trained in a single radiology department. Our statement segmentation was based on individual sentences, which sometimes lacked the nuance available and necessary to convey information with prose-style dictations. It would potentially be more useful to look at word groupings rather than individual sentences when segmenting reports. Along the same lines, simply rearranging the order of statements from a prose-based report into different subsections might decrease the readability of the overall report. We would expect dictation styles to change with the implementation of this sort of algorithm (less contiguous prose and more individual statements). We only trained and tested our algorithm in English. Applying a “language translation algorithm” to non-English reports to convert to English or using multilingual word embedding might allow other languages to be tested and used. Finally, we must acknowledge that the definition of “structured reporting” varies. We acknowledge that this algorithm performed only the most basic of structuring—applying section labels. Other schemata also incorporate lexicons (e.g., Breast Imaging Reporting and Data System, BI-RADS) in addition to section headings ; applying these more advanced techniques should be examined in future applications.
In conclusion, we demonstrated that an AI algorithm has high accuracy in converting free-text radiology findings into structured reports. This could improve communication between radiologists and referring clinicians without loss of productivity and provide more structured data for research/data mining applications. In addition, the prediction probability feature of the algorithm warrants further exploration as a potential marker of ambiguous statements.