Background:
Digital models are an increasingly important teaching and learning tools in medical education. Many earlier 3D (three-dimensional) models have been constructed by merging data from different subjects into one generic approximation. These approaches result in averaged representations that are limited in accuracy and ability to reproduce individual variation. Alternatively, techniques used to illustrate medically accurate single patient data have been typically challenging for the novice viewers to appreciate. We present a rationale and one possible workflow for producing anatomically and physiologically accurate digital models of organs. As an example from our recent work, we describe a visualisation pipeline for a spatially calibrated 4D (3D +time) model of the heart.

Methods:
A 55 year old male, with a kidney function impairment, underwent a free breathing cardiac magnetic resonance (MR) at 3.0 Tesla. A developmental 3D radial sampling acquisition protocol, with intravenous ferumoxytol (Rienso®, Tradeka Pharma) contrast, was used for data acquisition. The data was reconstructed offline, using a three stage workflow defined by Deng et al; including a respiratory signal extraction, sorting data into 16 phases of cardiac cycle and image reconstruction on a per-phase basis. The volumetric dataset was then used to evaluate two study aims. Firstly, the feasibility of producing a 4D model with a commonly available medical visualisation software packages was explored. Secondly, a simplified visualisation pipeline was defined for producing 4D visualisations of organs for medical education.

Results:
The use of radial acquisition scheme (sampling a signal along spikes around the k-space centre, instead of parallel rows as in a traditional plane based MR) allowed accelerated image acquisition, higher spatial resolution and introducing time as a new parameter (i.e. 4D). We found that OsiriX provided the best compatibility with file format requirements of common digital education software packages. As an open source application, its adaptability was essential for novel 4D models. A visualisation pipeline was optimised for OsiriX Lite (Pixmeo SARL, version 8.5.1) and can be used to illustrate medically accurate digital 4D models of organ systems. Our spatially calibrated model of the heart is optimised for illustrating intrinsic cardiac contraction. It provides a good contrast between the blood pool and other tissues, while further optimisation is needed for definition of the epicardial contours and small vessels.

Conclusions:
Introducing time as the fourth parameter to the medically accurate digital models of organs is useful for educational applications. These models allow users to appreciate real anatomy and physiology of internal organs in vivo, while reducing boundaries of traditional teaching settings and encouraging the use of effective educational strategies. In our specific example, visualising the intrinsic cardiac contraction has been challenging with the earlier image processing approaches. For a novice viewer, the myocardial contraction may have gotten masked by the bulk respiratory movement in the chest cavity, due to the pulsatile blood flow in the vasculature and because of ‘through plane’ motion in the traditional acquisitions (plane based +time). While the proposed workflow and visualisation pipeline produce a novel educational 4D model of the heart, the same approach can be easily expanded also to other organ systems.