中文摘要：

目的针对基本遗传纤维追踪算法易早熟等缺点,提出结合遗传算法（GA）、模拟退火算法（SA）的脑白质纤维可视化方法。方法在遗传进化的初期,依据适应值,按照轮盘赌选择规则选择纤维,在GA接近收敛的中期引入SA作为繁殖算子,并以一定交叉概率和变异概率调整纤维,获得较优纤维路径。结果改进后的算法追踪的纤维能量更小,且更加符合扩散张量场的分布。结论遗传模拟退火纤维追踪算法能够克服传统GA易早熟及易陷入局部最优解的缺点,可实现脑白质纤维的三维可视化。

英文摘要：

Objective To present a new method combining the genetic algorithm （GA） and simulated annealing algorithm （SA） for 3D visualization of fiber bundles based on the weakness of prematurity of genetic white matter fiber tractography. Methods The roulette wheel selection was used for selecting fibers according to fitness in the early stages of GA. Then in the middle stage of GA, SA was introduced for breeding operator of basic GA, and fiber paths between regions of interest were adjusted to find better fibers at certain crossover probability and mutation probability. Results Compared with basic GA traetography, the improved algorithm could find fibers with smaller energy and more in line with the tensor field＇s dis- tribution. Conclusion The genetic simulated annealing algorithm overcomes the weakness such as prematurity and conver- gence to the local optimum solution, therefore can realize 3D visualization of white matter fibers.