Lung Cancer Detection on CT Scan Images A Review on the Analysis Techniques

0. Introduction

1. Review of existing nodule detection methods

1.1 Pre_processing

1.2 Segmentation

• 2D-based segmentation

① Thresholding based methods^{[7][8][11][10][17][26][76][99]}

② Stochastic methods^[22][40]

③ Region based methods^{[1][18][20][28][59][62][63][79][80]}

④ Contour based methods^{[9][44][47][49][77][82][95]}

⑤ Learning based methods^{[4][19][45][58][90]}

• 3D-based approaches

① Thresholding methods^[91][92][96]

② Mathematical morphology^{[30][38][52][53][56][57][69]}

③ Region-based methods

• Region growing^{[19][21][37][54][55]}
• Graph-based methods(Graph-cut)^[97][98]

④ Model-based methods(Deformable models)^{[14][25][24][29][47][48][83][93][258]}

⑤ Dynamic programming^[86][87]

1.3 Nodule extraction and classification

• Fuzzy classification and Neural Networks---[2][4][12][13][19][43][52][61][62][73][86]
• K-Nearest Neighbor---[25][50][66][81][96]
• Support vector machines---[36][46][65][70]
• Linear Discriminant Analysis---[5][33][41][49][67]

2. Conclusion

• Developing new and better techniques of contrast enhancement;
• selecting better criteria for performance evaluation is also needed.