Evaluation of the accuracy of a surface fitting based registration between 201Tl-SPECT and MR images in brain tumor studies
G. Nicolini 1 , I. Butti1 , L. Conte 2 , R. Novario 1, A. Fachinetti 3
1. Servizio di Fisica Sanitaria, Ospedale di
Circolo di Varese, Varese (Italy)
2. Dipartimento di
Scienze Cliniche e Biologiche, Università dell’Insubria,
Varese (Italy)
3. Servizio di
Medicina Nucleare, Ospedale di Circolo di Varese, Varese (Italy)
Manuscript received: May 15, 2000; revised: September 4 and December 12, 2000.
Accepted for publication: December 14, 2000.
Abstract
Correlation between 201 Tl-SPECT and MRI images improves their diagnostic potential in intracranial tumors. The aim of this work was to assess the accuracy of the algorithm for use in clinical procedures. Our investigations have adopted three different paths: a) with a ‘digital phantom’ to look into the intrinsic limits of the algorithm; b) with a geometric phantom to point out the effective registration errors of intracranial lesions; and c) with 11 patients, placing fiduciary markers on the subjects’ scalps to quantify correlation accuracy. The use of the digital phantom has shown a clear difficulty of the algorithm in recognizing the rotation out-of-plane about X or Y-axes. For the patients where there was not a significant tilt between the X-Y SPECT and MRI planes, the match was successful. On average, a mean distance between corresponding markers of 5.54 ± 0.46 mm was measured that, as shown by verification with the geometric phantom, overestimates the local errors inside the brain. In our opinion, surface matching can be used in clinical routine with a visual validation, based on fiducial anatomic features (the scalp), provided that significant rotations about the X or Y-axes are avoided or pre-corrected.
KEYWORDS: Image fusion; brain tumor; SPECT; MRI.