Automated Phenotyping in Epilepsy

全球有 6500 万人患有癫痫症，2017 年诊断出 15 万例新癫痫病例 美国每年。然而，癫痫的治疗选择仍然不足，许多患者患有癫痫病 难治性癫痫发作、认知合并症和治疗的副作用。一个专业 开发新疗法的障碍是临床前癫痫研究 通常需要对癫痫发作进行劳动密集型且昂贵的 24/7 视频脑电图监测 由人类观察者对癫痫表型进行主观评分（以广泛使用的拉辛量表为例 行为性癫痫发作）。最近，达塔实验室表明，复杂的动物行为是由 亚秒级时间尺度的定型模块（“音节”）并根据特定规则排列 （“语法”）。使用称为运动排序的方法可以在没有观察者偏差的情况下检测这些音节 (MoSeq)，采用 3D 摄像头视频成像与人工智能 (AI) 辅助视频相结合 分析来表征行为。通过 Soltesz 和 Datta 实验室之间的合作，令人兴奋的数据 获得的结果证明 MoSeq 可以适用于癫痫研究以执行客观的、 在慢性颞叶癫痫小鼠模型中对小鼠进行廉价且自动化的表型分析。在这里我们 建议进一步测试和改进 MoSeq，以解决癫痫领域长期存在的根本性挑战 研究。这包括开发拉辛量表的客观替代方案、对 MoSeq 进行测试 一种自动抗癫痫药物（AED）筛选方法，以及人类观察员的开发 癫痫发作、癫痫发生和认知合并症的独立行为生物标志物。此外，我们 计划大幅扩展 MoSeq 的癫痫相关功能，包括自动跟踪 目前的方法无法实现更精细的身体部位（例如前肢和面部阵挛）。最后， 我们建议将 MoSeq 的分析流程开发为直观、廉价、用户友好的形式 从而可以轻松地与研究界共享。我们预计这些结果将具有潜在的 通过展示自动化、客观、用户化的可行性和力量，对该领域产生变革性影响 对获得性和遗传性癫痫表型进行独立、廉价的分析。

There are 65 million people worldwide with epilepsy and 150,000 new cases of epilepsy are diagnosed in the US annually. However, treatment options for epilepsy remain inadequate, with many patients suffering from treatment-resistant seizures, cognitive comorbidities and the negative side effects of treatment. A major obstacle to progress towards the development of new therapies is the fact that preclinical epilepsy research typically requires labor-intensive and expensive 24/7 video-EEG monitoring of seizures that rests on the subjective scoring of seizure phenotypes by human observers (as exemplified by the widely used Racine scale of behavioral seizures). Recently, the Datta lab showed that complex animal behaviors are structured in stereotyped modules (“syllables”) at sub-second timescales and arranged according to specific rules (“grammar”). These syllables can be detected without observer bias using a method called motion sequencing (MoSeq) that employs video imaging with a 3D camera combined with artificial intelligence (AI)-assisted video analysis to characterize behavior. Through collaboration between the Soltesz and Datta labs, exciting data were obtained that demonstrated that MoSeq can be adapted for epilepsy research to perform objective, inexpensive and automated phenotyping of mice in a mouse model of chronic temporal lobe epilepsy. Here we propose to test and improve MoSeq further to address long-standing, fundamental challenges in epilepsy research. This includes the development of an objective alternative to the Racine scale, testing of MoSeq as an automated anti-epileptic drug (AED) screening method, and the development of human observer- independent behavioral biomarkers for seizures, epileptogenesis, and cognitive comorbidities. In addition, we plan to dramatically extend the epilepsy-related capabilities of MoSeq to include the automated tracking of finer-scale body parts (e.g., forelimb and facial clonus) that are not possible with the current approach. Finally, we propose to develop the analysis pipeline for MoSeq into a form that is intuitive, inexpensive, user-friendly and thus easily sharable with the research community. We anticipate that these results will have a potentially transformative effect on the field by demonstrating the feasibility and power of automated, objective, user- independent, inexpensive analysis of both acquired and genetic epilepsy phenotypes.