A novel computerized approach to enhancing lung tumor detection in whole-body PET images

Positron emission tomography (PET) is complimentary to other imaging modalities such as CT and MRI and provides a unique and effective means for detecting tumors in vivo through tissue metabolism measurement. At the majority of clinics, only the attenuation-corrected images are read by the physician for tumor diagnosis; the unconnected images are not examined, losing critically important information for a small portion of patients. We have developed a novel image processing method capable of automatically detecting and ranking tumor candidates in the lungs using the whole-body PET images. The intended utility is to visually prompt tumor candidates, assisting the physician to achieve better diagnosis, especially when the candidates appear to be subtle. The technique takes advantage of different information contents in the emission, corrected and uncorrected images. It processes the images three-dimensionally and the processing consists of segmentation, multi-thresholding with volume criterion, and heuristics-based tumor candidate ranking. This method is fast in computation and display and thus is suitable for real-time applications using high-end PCs. Our preliminary retrospective study involving nine patients has yielded promising results.