Background

The Cambridge neuropsychological test automated battery (CANTAB) is a set of computerized visuospatial tests used to probe cognition in humans. The non-human primate (NHP) version of the battery is a valuable translational research tool to quantify cognitive changes in NHP models of disease by allowing direct comparison with performance data from human patient populations. One limitation is the long training times required for NHPs to reach appropriate levels of task performance, which is prohibitive for high throughput experimental designs.

New method

We report a new training regimen to teach NHPs a subset of CANTAB cognitive tasks using a method of successive approximations (shaping), where rewarded behaviors progressively approximate the goal behavior, and sequential task learning is used to build upon previously learned rules. Using this refined method, we taught 9 adult rhesus macaques to perform three tasks: the self-ordered spatial search (SOSS), delayed match-to-sample (DMTS), and paired associative learning (PAL) tasks.

Results and comparison with existing methods

NHPs learned all three cognitive tasks in approximately 130 training sessions, roughly 200 sessions faster than previously published training times. NHPs were able to perform each task to a stable level of performance (>80 % correct) enabling their use in future cognitive experiments.

Conclusions

Our approach of behavioral shaping reduced the time to train NHPs to performance criteria on SOSS, DMTS, and PAL tasks. This allows efficient use of the NHP-adapted CANTAB to compare cognitive changes in NHP models of neurological disease with those observed in human patient populations.

中文翻译：

---

使用剑桥神经心理学自动测试电池对恒河猴进行行为整形。

背景

剑桥神经心理测验自动电池（CANTAB）是一组计算机化的视觉空间测验，用于探测人类的认知能力。该电池的非人类灵长类动物（NHP）版本是一种有价值的翻译研究工具，可通过与人类患者群体的性能数据直接比较来量化疾病NHP模型中的认知变化。一个局限性是NHP要达到适当水平的任务性能需要很长的训练时间，这对于高通量实验设计是不利的。

新方法

我们报告了一种新的训练方案，该方法使用相继逼近（整形）方法向NHP教授CANTAB认知任务的子集，其中奖励行为逐渐逼近目标行为，而顺序任务学习则用于先前学习的规则。使用这种改进的方法，我们教了9只成年恒河猴来执行三项任务：自排序空间搜索（SOSS），延迟匹配样本（DMTS）和配对联想学习（PAL）任务。

结果和与现有方法的比较

NHP在大约130个培训课程中学习了全部三个认知任务，比以前发布的培训时间快了大约200个课程。NHP能够以稳定的性能水平（正确率> 80％）执行每个任务，从而使其可以在未来的认知实验中使用。

结论

我们的行为塑造方法减少了将NHP训练为SOSS，DMTS和PAL任务的性能标准的时间。这可以有效利用适应NHP的CANTAB，将神经疾病NHP模型中的认知变化与人类患者群体中观察到的认知变化进行比较。