Application of image registration methods to gammagraphy explorations with meta- iodobenzylguanidine for the treatment of neuroblastoma

Abstract

Neuroblastoma is a childhood cancer with a high mortality. To detect it, a gammagraphy scan is done, injecting prior to it a radioactive isotope, the iodobenzylguanidine (MIBG), to the patient. MIBG reacts with the cancer, showing the tumor/s on the result image of the scan. Sadly, there is not a standard way to take these images (intensity, distance of the projector from the patient, position of the patient, etc. are not defined), plus the fact that the images are very low in quality: contrast is low, noise is very high, borders are not clear... All together makes difficult the work of the medical specialist to identify properly the tumor and even harder when he must compare images of the same patient. We want to make it easier for the specialist to monitor a patient's treatment. If he can differentiate properly if the state of the patient on the last scan's result is better or worse than the previous one, he can decide if the treatment is going well or bad, respectively. In general, he can have a more accurate progress of the patient. In order to do that, we propose an algorithm for image registration adapted to our problem. Image registration consists on aligning two similar images, which would allow us to find easily differences on them. If we can register two scans of the patient, it would be fast for the specialist to compare the tumors in both of the images at the same time and take a decision. We have developed several algorithms and a small interface to support them. With this tool, the specialist is able to register two medical images of a patient, either automatically or manually, selecting points in both images to do a correlation. After trying the tool with 34 gammagraphy images, it gives reasonable results for most of them, being some of them not possible to register even manually due to their heterogeneity.