Lung modeling for nodule analysis in HRCT

A computerized model of lung has been elaborated with the aim of simulating lung lesions in CT scans. The secondary lobule was used as structuring lung element. Several lung slices that cover the entire organ are included in the model. Broncho-vascular trees have been modeled, each final leaf pointing to one lobule of the predefined slices. Lobar, segmental and sub-segmental areas are defined from the arterial tree. Information on lobule location regarding scissure and pleura have also been included in the model. The first lesion studied is the pulmonary nodule. A software tool was developed which allows the radiologist to spread nodules on lung model in a defined way until the simulated images mimic the CT. Several nodule shapes, mathematically simulated, are available and systematic and non-systematic nodule distributions have been modeled. Several predefined nodule sizes (mean and dispersion) have been carefully defined. Nodule superimposition is avoided by the distribution algorithm unless explicitly authorized. Manual nodule shifting is available, but restricted to the area defined by the distribution to which the nodule belongs. When a CT has been completely analyzed, the different actions performed to obtain the final result are stored in a file which will be connected to a database in which data concerning the patient and final diagnoses will be stored. This lung model will be used to simulate ground glass and consolidation. The software tool described here is a useful help for radiologist education and training and it can lead to better understanding of the complex information present in HRCT images and to appropriate diagnoses.