Osirix is a popular opensource DICOM viewer for the Mac. It is actively developed by Pixmeo , Switzerland. It has become popular in the last years, since it integrates a user friendly interface for medical imaging visualisation with powerfull plugin system that extends its functionaly, enabling advanced image processing techniques

The Sant Pau hospital uses OsiriX as the preferred tool for TMR sequence viewing. Because of this, it is interested in the development of a plug-in for the extraction and validation of myocardial torsion from their TMR sequences. We have previously built a prototype in Matlab for this purpose. However, communication between Matlab and OsiriX is slow and memory intensive. Therefore we have decided to port existing code into Objective C.

Gabor Filter Bank computation

We built two Gabor filter banks for each tag orientation. Both banks are fine-tuned  for each left ventricle level. This allows for an optimally filtering of  complex images. This is crutial, since  the complex phase   remains constant along the cardiac cycle. This  brightness constancy condition is required by classical optical flow (OF) methods.

Sequence Filtering

Both filtered sequences belonging to a variational formulation schema  are used in a to obtain a dense continuous deformation map. We consider this as Harmonic Phase Flow.

This method has been used to determine reference values of ventricular torsion (VT) in a set of 40 healthy volunteers. The results encourage the use of ventricular torsion as a useful parameter for ventricular function assessment in clinical routine.

OsiriX platform acquaintance

To translate Matlab code into native Objective C, we looked at other existing plug-in implementations, such as Brian Jensen’s NMSegmentation plugin from the Technische Universität München. In order to familiarise with the OsiriX plug-in SDK, we developed an initial version which only used a fixed set of points as the input LV. After this, we evolved these points in time using a set of pre-calculated HPF motion matrices. We consider the ventricular torsion as the difference in rotation at two different Short Axis views of the same sequence. To validate our results, we measured the absolute error between the global mean VT of our prototype and Matlab. Additionally, we ensured that the plot curves followed the expected tendency in both cases

HPF computation

HPF is a regularised spectral optical flow, under a variational framework that uses the maximum response of a set of Gabor filters banks. This flow characterises the ventricular torsion in healthy subjects.  In step 1, we detect the harmonic peak which we use for building the Gabor filter banks (step 2). The combination of the response of banks produces a complex image whose phase is related to tissue motion (module 3). Finally, the we track the amplitude of the maximum response for building the displacement map.