Author Affiliations
School of Electronics and Information Engineering, Lanzhou Jiaotong University, Lanzhou, Gansu 730070, Chinashow less
Fig. 1. Framework of medical image registration algorithms
Fig. 2. Schematic of characteristic points of each region
Fig. 3. Feature descriptor calculation area range
Fig. 4. Flow chart of proposed algorithm
Fig. 5. Comparison of simulation results of SIFT detection points. (a) MRI image; (b) detection points of traditional SIFT algorithm; (c) detection points of improved SIFT algorithm
Fig. 6. Effective point matching pairs of original SURF algorithm and improved SURF algorithm. (a)(b) Original SURF algorithm; (c) improved SURF algorithm
Fig. 7. MRI sequence images of T1WI, DWI, and PWI. (a)-(g) T1WI image sequences; (h)-(n) DWI image sequences; (o)-(u) PWI image sequences
Fig. 8. DCE image
Fig. 9. Registration results of T1WI, DWI, and PWI. (a)-(g) Registration results of T1WI; (h)-(n) registration results of DWI; (o)-(u) registration results of PWI
Fig. 10. T1WI sequence image registration rate
Fig. 11. DWI sequence image registration rate
Fig. 12. PWI sequence image registration rate
Image | Size | SIFT | SUFT | Ref. [10] | Proposed algorithm |
---|
T1WI | 558×558 | 0.4037 | 0.3392 | 0.2103 | 0.1313 | DWI | 288×288 | 0.2110 | 0.1982 | 0.1407 | 0.1069 | PWI | 560×560 | 0.5954 | 0.5295 | 0.3004 | 0.1510 |
|
Table 1. Root mean square error
Image | Size | Ref. [10] | Proposed algorithm |
---|
T1WI | 558×558 | 186.535 | 90.514 | DWI | 288×288 | 124.134 | 80.746 | PWI | 560×560 | 182.462 | 94.468 |
|
Table 2. Running timems