AI-Assisted Lung Nodule Diagnosis System

The lung nodule artificial-intelligence-assisted diagnosis system is based on a deep learning algorithm to achieve automatic segmentation of the range of ground glass nodules and recognition of typical signs (19 (link)–21 (link)). In the whole calculation process, the system automatically divides the range of the ground glass nodules and calculates the number of voxels corresponding to each CT value in the whole SSN. Each CT value and the corresponding number of voxels are stored as a LIST, and the LIST of the whole nodule is stored as a DICTIONARY. The information obtained is used to calculate the required index by the corresponding formula. First, the CT value threshold of −300 HU was used to distinguish the solid component from the ground glass component. The nodule volume, mean density, solid component volume, percentage of solid component, mass, mass of solid components, and other three-dimensional metrics were calculated based on the voxel method and the corresponding formulas, as follows (Figure 2).
Solid mean density = ${\displaystyle {\sum }_{i=0}^{2000}{x}_i{p}_i}$ (Only including x_i ≥ -300HU).
Percentage of solid components= total number of voxel ≥ −300 HU (solid components)/total number of voxels (all tumor)
Mass= [nodule volume×(mean density +1,000)]/1,000.
Mass of solid components= [solid components volume× (solid mean density +1,000)]/1,000.
Then, the CT histograms were constructed based on the number of voxels corresponding to each CT value in the nodule range. Variance, skewness, kurtosis, entropy, and other density histogram-related indicators were automatically calculated by python coding and the corresponding formulae. Meanwhile, the typical signs detected and identified by the system were confirmed by two radiologists as morphological indicators, including lobar signs, spiculation signs, and pleural traction signs.