I'm a NYC-based radiologist hoping to improve patient understanding/engaging with radiology reports using natural language processing.
Patient-portals that allow patient's access to their medical information is on a steep rise. And, the #2 most common thing patient's look up and want access to is their radiology reports (#1 is laboratory data). The problem is that current surveys reveal that patient's are greatly unsatisfied by radiology reports due to their complex medical jardon, polysyllabic compound words, obscure anatomical references, etc. - they want to understand but are discouraged by the language. I think we can change that.
Specifics of use for NLP
We can use NLP to convert/translate complex radiology reports into much more understandable yet completely accurate/informative patient-centered reports. We can rewrite them in lay language and link certain words to additional information and anatomical maps. We can use NLP to fully engage the patient and involve them in their care by helping them understand the radiology report.
One group at UPenn recently demonstrated proof of concept:
I think there's incredible potential for this technique to change how patient's engage with their medical data.
1) Input: Input a standard/complex radiology report.
2) Run the report through our NLP algorithms
3) Output: A completely translated report written in lay terms (6th or 8th-grade reading level) with hyperlinks to additional information and anatomical maps.
To start, I would like to process, for example 1000 brain MRIs or 1000 abdomen/pelvis CTs, etc. (of which I can provide) to determine the frequency of the most commonly used words and phrases for each given study, and from there we can make a "dictionary" of those words in layman terms, which could than be substituted in to improve patient understanding and engagement...
I'd like to start with one exam type (i.e. Brain MRI) to prove proof of concept.
If you have any interest, please feel free to contact me.