A pixel-based machine learning algorithm named massive training artificial
immune recognition system (MTAIRS) is developed to detect and enhance nodules
from computed tomography (CT) scans. This method employs the concept of pixel
machine training where all the pixels in the region of interest (ROI) are
involved in the massive data training. In the ordinary medical imaging analysis,
calculated features from the segmented ROI are used as sample of learning in a
machine learning algorithm, however this will cause the loss of important
information in CT images. Due to the high computation cost in pixel machine
training, artificial immune recognition system is chosen to optimize the pixel
training time of MTAIRS algorithm. The teaching class is formed by the Gaussian
distribution function and every single pixel in teaching image represent a class
of each instance of training ROI. Two versions of the MTAIRS algorithms are
developed to compare their performance in the classification of medical images.
The main advantage of MTAIRS is to reduce the redundant training data while
retaining the most relevant features in the processing of large amount of data
training. MTAIRS have generated promising visualization for nodules enhancement
and accuracy in detection of nodule in CT scans.