Oleg S. Pianykh, Presenter: Nothing to Disclose

Wide spread of clinical perfusion applications led to numerous perfusion implementations and software packages, often poorly correlated with each other. Moreover, the choice of perfusion-computing methodology and even the choice of computational parameters can greatly affect the outcomes of any perfusion implementation. This creates an imminent need for an independent perfusion validation method, which can compare different perfusion implementations in the most objective way. We propose the use of predefined digital image sequences, or perfusion phantoms, to validate the correctness of any perfusion-processing implementation.

Any perfusion algorithm accepts temporal DICOM image sequences as input. Therefore one can build an artificial sequence of DICOM images such that: 1. Each pixel intensity is changing in time according to a known function. 2. The functions and their parameters are chosen to produce predefined perfusion values (blood volume, blood flow, and mean transit time) 3. The predefined perfusion values form certain visual patterns on their perfusion maps 4. Artificial noise is added to known image areas to validate algorithm’s tolerance to noise. When this “phantom sequence” is fed into perfusion-processing software, we expect to obtain well-known perfusion maps with visible predefined patterns. Therefore, by observing the actual perfusion maps, calculated by the software, we can compare them to the expected, and objectively judge the quality of this software implementation.  

We constructed a temporal sequence of 30 DICOM images, which was built to produce perfusion maps with predefined check-board patterns. This sequence was tested with two commercial implementation of perfusion analysis. Both implementations demonstrated different deviations from the expected perfusion maps. This comparison enabled us to select the most accurate perfusion software package

Digital perfusion phantoms provide an easy and efficient way to validate any perfusion implementation. They do not require actual data, scanning, or even any knowledge of perfusion principles. They can also be used by perfusion algorithm developers to fine-tune their processing techniques

Our approach is ideal for this purpose: it is simple, fast, and robust. Therefore it can be used in any clinical practice to objectively validate the quality of perfusion analysis.

Pianykh, O, Perfusion Validation with Perfusion Phantoms.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8016199.html