Mixed wavelet-watershed method for nodule detection in high resolution scintimammography

The recent developments in gamma cameras have lead to significant improvements of intrinsic spatial resolution in radioisotopic imaging. Especially, using smaller photodetectors (PSPMT, photodiodes) enables to build more compact devices which permits to place the detector very near from the organ under study and thus to enhance the global spatial resolution. However, these improvements need to reduce image pixel size which worsens the signal to noise ratio in every pixel. Moreover, many clinical examinations, such as scintimammography will be all the more useful since small nodules can be detected. In this work, we present a method combining a denoising wavelet filtering method with watershed determination, both leading to hot nodule enhancement. This method was applied to Monte Carlo simulations of a 64 mm high volume containing hot nodules of various sizes with a tumor to background ratio equal to 5. Although nodules were not visible in raw images corresponding to 10 min acquisitions, the described process allowed the detection of 88% of 6 mm nodules and more than 97 % of nodules whose diameter was greater or equal to 8 mm. Results obtained on phantoms without hot spheres permit to assess the pertinence of the proposed method. Furthermore, the whole process runs very fast.