Medimq: A Medical Image Diagnostic Learning and Assessment Tool

Cedric Dumas, PhD, Timothy R Coles, PhD, David Fielding, MD, Caroline G Cao, PhD. Commonwealth of Scientific and Industrial Research Organisation, Wright State University.

New endoscopic technologies in the hospital operating room (OR) have dramatically improved the technical performance of surgical procedures. Where open surgery and traditional literature provides teaching from an external view, endoscopes provide a completely different internal perspective. This vision shift challenges the endoscopists’ anatomical and diagnostic knowledge, as high resolution imagery of internal tissue can be provided. In bronchoscopy and colonoscopy for example, high definition images can be captured using short focal distance camera in narrow band imaging (NBI), ultrasound or auto fluorescence techniques.

Training practitioners to identify the previously invisible features of pathologies now observable using new imaging techniques, or visible in higher resolution images, is key to utilising the enhanced diagnosis potential of this technology.
Although it has been shown that frequent assessment can be used to enhance and reinforce learning, current literature-based image recognition learning is passive with no frequent testing available.

As the early training phase (understanding and remembering) can be done with atlas images and videos, the next learning phases (analysing and applying) have to be done with an interactive tool before working on patients. These training activities need to be followed by rehearsal sessions to improve skill and confidence.

We describe here a Medical Images Quiz (, an online interactive image-based training tool that can provide immediate feedback on pathology recognition. The tool allows trainees to directly mark up medical images (see Figure 1) to show where biopsy/procedure sites should be. The sensitivity and the specificity of the trainees’ answers can then be compared to the experts’.

Methods and Procedures
Two training sessions were performed with 40 thoracic doctors, all novices in NBI diagnostic tasks. The trainees were evaluated on their ability to find normal and abnormal vessels in NBI images, and their ability to recognize anatomy (mucosal or submucosal vessels).

Trainees improved in Specificity from 0.67 to 0.90 during the training, but decreased again after eight weeks in a retention test, showing the importance of follow-up training.

With computer-based training systems, medical educators have access to a new tool to train their students, but also to perform a formal evaluation based on their pre-set assessment criteria. This tool can be used before (pre test), and during the training sessions, on the same or different sets of images. After the training, it can be used as a rehearsal or retention test. The Medimq provides a self-guided training program in medical image analysis by providing assessment and online interactive training in addition to traditional methods.


Figure 1: Preparing the test, the expert (left image) marks the right answers in the whole image, by drawing and describing areas with its own taxonomy (centre image). During the test, the trainee (right image) marks biopsy sites interactively and describes them with the same taxonomy.

View Poster

« Return to SAGES 2014 abstract archive

Reset A Lost Password