Hands Free Navigation for Abdominal Imaging

Value Proposition

As the amount of volume per day for each radiologist increases, both in terms of total studies and number of images per study, navigating through all the images has become a rote physical task which leads to repetitive stress injury. Having hands free navigation using voice instead would reduce these injuries, facilitate the workflow of radiologists with disabilities, free up hands/attention for other tasks (such as teaching), standardize individual search patterns/checklists (potentially decreasing missed findings due to interruptions) and overall improve job satisfaction. 

Narrative(s)

A busy body imaging service reading a large number of cases including multi-phase studies, often with thousands of images, is experiencing issues with wrist injuries in their radiologists. This seems to be both caused by scrolling through extensive imaging stacks and exacerbated by it, making work painful and difficult for the radiologists. Voice activation features could be utilized in this case to open exams, select display protocols, and scroll through different series. When scrolling through the different series, the software should be able to auto-detect certain regions and display them on voice command, such as the different organs and the four quadrants of the abdomen. There should also be a standard scroll-through order which each reader can customize and which encompasses all the images in the study. This should be interruptible by voice to stop and look at a certain area. It should also be able to be re-engaged by voice to resume at the point in which the standard scrolling routine was last interrupted. These changes would allow radiologists in this group to work for longer before retirement, alleviate pain to improve focus, and prevent injuries from forming in the younger members. 

Workflow Description

1.  A voice activated command of “Open Unread Body CT List” after login enables the reader to navigate to their reading list of interest. 

2. “Open the first study on the list” opens the study for the reader.

3. “Change my display protocol to CT abdomen and pelvis no priors” would pull up the reader's personal display protocol for this.

4. “Scroll through the axial thin section images” would activate the user’s personal scrolling preferences.  Scrolling should begin at a predetermined speed set as a user preference and should be actively adjustable using the voice phrases of, “scroll faster” and “scroll slower.”

5. “Stop scroll” “Resume scroll” would be used by the reader to pause and examine images as needed.  Manual scrolling should be allowed to be activated by mouse as needed, and when finished, the program should be returnable to where it left off in the auto-scrolling protocol with the phrase, “resume auto-scrolling”.

6. “Go to the liver” or “Go the right lower quadrant” voice commands are used to go back and re-scrutinize areas where the user thinks they may have seen something or wants to examine further for reasons specific to the patient. Users should be able to set their preferences for both order of organs/quadrants and how many passes to make through each.

7. “Resume auto-scrolling” would again take the user back to where they left off in their scrolling protocol.

8.  “Scroll through the series 1” and “Scroll through the first phase post-contrast sagittal” voice commands would also be used in a similar manner (voice activation by series number or description).

Inputs

Voice Commands

Outputs 

PACs action/scrolling

The AI algorithm which automatically detects the type of series/sequence has use far beyond this voice activation schema.  Hanging protocols are often based on series/sequence titles which are notoriously inaccurate.  Inaccurate labels waste radiologist time in searching for the correct series, especially in MRI, and can also lead to inaccurate interpretations.

Beyond voice activation for scrolling, voice activation for measuring lesions and putting them in the report would be a huge efficiency and accessibility gain for image interpretation. The ability to show the same region or lesion by voice activation on any prior imaging would be wonderful as well and could likely piggyback on the training algorithms to recognize organs/regions needed for this project. The more elements of the workflow can be automated and voice-activated, the longer radiologists can practice without injury, lengthening the span of their careers.  Hand free instead of constant searching and clicking of menu buttons would also make it easier for radiologists to teach residents and consult with ordering physicians.