When clinicians (or more specifically radiographers) interpret medical images (such as CT or MRI images) they type up their findings into an electronic reporting form, this is generally in the form of a single simple text box. An example of the type of data that could be entered into the form is as follows: “A cancer on the left lung in {A, B} position is suspected but not confirmed with details {X}, {Y} and {Z}”.
This is useful for when a clinician loads a specific dataset to determine the radiographer’s findings. However this is limited as it does not allow for ease of searching through aggregated data such as requesting to see images that have confirmed cancer, as doing natural language processing on the free-form data is not implemented and even if it was it would not work perfectly (due to the complex grammar used).
Therefore it would be useful to create a system with enables more metadata creation (potentially using multiple fields), having separable specific and critical information, allowing for easier searching of data, and enabling more automated analysis of the data to help with diagnosis and treatment planning.
To create the prototypes, firstly Photoshop [1] was used to create the individual screens and icons needed. InVision [2] was then used to bring each of the system components together and create links between each of them so that it gives the user the feel of actually using the system. Each of the prototypes created can be classified as high-fidelity prototypes.
An attribute of "Confidence Level" was created to record one's confidence in their reported findings. This will serve as a reference for other radiologists looking at the report, as they can see if the completed report is a good basis or if the need to start again depending on the confidence.
Overall, two solutions are provided.
A "Question Tree" was used in the conservative to let any radiologist can edit the report by going back to previous questions to change any answers or create a new one, including in an MDT meeting. Furthermore, it provides the complete transparency within the system, so users of the system will not feel the system is hiding how any decisions are made.
We also designed a "Speaking System" for the out-of-box solution. The premise of this system is that user control is done primarily through speaking. The reporting form will ask the radiologist questions through audio clips and use voice recognition to record their responses. This is specifically beneficial for the creation of the initial report. It minimises time spent typing and allows radiologist to focus on the images they are analysing. This is because they do not have to switch their attention between the image screen and the form screen, apart from to confirm the recorded answers.
An additional suggestion, which has not been included in the final solution, is creating a dashboard for managing all possible behaviors, such as creating a new case, calendar a MDT meeting, etc. Prototypes have also been created for as a recommendation for further development.
