Table 1.

Advantages and disadvantages of the comparison methods.

Methods	Advantages	Disadvantages
3D-FCN	Simple implementation	Incorrect determination of non-nodule tissue outside lung as nodule
MRCNN	Simple implementation, multi-resolution model is good for small-size nodule detection	Incorrect determination of non-nodule tissue outside lung as nodule
3D-UNET	Lung parenchyma regions are first detected, lower over-estimation rate	Near-edge regions are easily lost, confusion of small tissues as nodules
PRN-HSN	Lung parenchyma regions are first detected, lower over-estimation rate	Near-edge regions are easily lost, small-size nodule within weakened, low-resolution region cannot be recognized
DCNN	Simple implementation	Small-size nodules on the edges of parenchyma region are easily omitted
CLAHE-SVM	Nodules within low-contrast regions can be detected with the contrast-enhancement pre-processing	Small-size tissues are easily over-estimated as nodules
Mask-RCNN	Small-size nodules can be detected accurately, nodule detection and segmentation are achieved simultaneously	Heavy computational cost, unstable performance on nodule detection
proposed	Pulmonary nodules can be detected accurately with low over-estimation of non-nodule tissues	Nodules within low-contrast, vague regions cannot be detected