早期ALS检测：Ahmadreza Keihani博士的实时神经技术解决方案

当你稍微做一点——为什么噪声在这里很重要？

When you do a little- and why noise is here important?

我会解释这一点。

I will describe that.

而且大脑向肌肉发送信号与反馈之间存在延迟。

And also there is a delay between the brain when sends the signals to the muscles and a feedback.

肌肉会将这些信息传回大脑。

Muscle sends that to the brain.

因此存在噪声效应和延迟。

So there is a noise effect and delay.

所以相干性无法捕捉这一点。

So coherency cannot capture this.

如今，以简单而有意义的方式使用相干性很有趣，展示了许多良好的结果。

Nowadays, use coherency is interesting in a simple and meaningful way, show many good results.

但这存在一个问题。

But there was a problem with this.

我们思考过如何更好地揭示这种耦合关系以展现因果关系？

And we thought that how we can reveal this coupling in a better way to show the causality?

因为正如我所说，在ALS中，重要的是要弄清楚问题是从大脑到肌肉，还是从肌肉到大脑。

Because here, as I said, in ALS, it's important to know that it comes from brain to the muscle, the problem, or from muscle to the brain.

因此你也需要一种方法来展示这一点。

And so you should have a method also to show this.

这是我硕士期间研究非线性模式时的经验。

That was experience from my master working in nonlinear patterns.

还记得我们在视觉和听觉系统中开发非线性模式吗？

Do you remember we developed nonlinear patterns in the visual and auditory systems?

我们运用非线性动力学和混沌系统的方法向听众解释什么是非线性动力学和混沌。

And we used the approach of nonlinear dynamics and chaos systems to describe what is the nonlinear dynamics and chaos to audience.

你知道，这很难解释，但混沌系统是具有封闭数学形式的系统。

You know, it's hard to explain it, but the chaotic system are systems that have a closed form mathematics.

这很简单。

It's simple.

例如，逻辑回归Xn加一等于R乘以Xn乘以一减Xn。

For example, logistic regression Xn plus one equals to R Xn and one minus Xn.

这是个简单的方程。

It's simple equations.

但问题在于存在一个参数，当你改变它时，系统会到达某些分岔点，之后你无法预知系统会如何运行，或者说无法预测该系统的功能。

But the problem is that there is a parameter that when you change it, you go some points that are bifurcations, and you don't know what system should do after that, or you cannot predict the function of that system.

这就是分岔点，人们都知道蝴蝶效应。

So this is bifurcation points, and people know the butterfly effect.

这种函数存在于我们的身体里，存在于我们的信号中。

This function is present in our body, in our signals.

是的，我们有封闭的公式，简单的公式，但函数行为将是不可预测的，明白吗？

Yes, we have closed formula, simple formula, but the function will be unpredictable, know?

简单来说，什么是分岔？

To make more simple, what is bifurcation?

假设你在大海里，举个例子。

Let's assume you are in ocean, for example.

你有猎人和鱼群。

You have hunters and the fishes.

分叉意味着如果你过度捕猎，鱼类将会灭绝。

Bifurcation means if you do more hunting, fishes will be extinct.

如果猎人数量少，捕鱼量会很高但资源会不足。

If you have low hunter, fishing will be high and the source is not.

这个点就是所谓的分叉点，举个例子。

This is the point that knows as a bifurcation, for example.

这是个临界点。

It's the critical point.

那个点很重要。

That point is important.

因此了解系统基于这些行为的表现并定义这些临界点很重要。

So knowing the system based on these behaviors and defining these points are important.

明白吗？

Okay?

因此，从非线性动力学和混沌理论的视角可以揭示许多线性函数（如相干性）无法展现的模式与现象。

So nonlinear dynamics and chaos point of view can reveal many things, many patterns that you cannot see with the linear functions like coherence.

明白吗？

Okay?

那么，说实话，人们选择更复杂方式的原因是什么？你知道，这让理解变得困难。

So, and what is the reason, honestly, to go to the more complex ways that people, you know, makes hard for them to understand?

你为什么知道？

Why you knew?

我们只是需要证明我们正在做更多无害的事？

We just need to show that we are doing more harmless?

不是。

No.

这是有原因的。

There is a reason.

如果从信号处理的角度看，我们的大脑信号和肌肉信号都是非平稳的。

If you look at with the signal processes point of view, our brain signal, our muscle signal are non stationary.

当你定义傅里叶变换或线性系统时，总是假设系统是平稳的。

And when you define Fourier or linear things, you always assume that systems are stationary.

这就是原因所在。

So this is the reason.

这就是有时使用线性系统会出现的问题，例如。

This is the problem sometimes when using linear systems, for example.

因此我们采用了递归图方法。

So we have used the recurrent plots.

在这种非线性、动态和混沌的视角下，存在一种方法。

There is a method in this perspective, nonlinear, dynamic, and chaos.

从那个角度看，你可以观察到肌肉激活时两个系统之间的耦合现象。

In that point of view, you can see the coupling of the two systems together when the activation in muscle happens.

无论它是否发生在大脑中。

It happens in the brain or not.

两个振荡系统A和B相互耦合振荡。

Two oscillations, A and B, two systems oscillate with each other.

它们相互振荡的时间点是什么，时间差是多少，信号是从大脑传到肌肉还是从肌肉传回大脑？

What are the times that they oscillate with each other, and what is the difference in time, and it's coming from brain to the muscle or from the muscles to the brain?

这就是我们采用这种方法的原因。

So this was the reason we used this approach.

这帮助我们发现了ALS患者与健康对照组之间的差异，包括延迟差异和耦合指数。

And this helped us to find the difference between the ALS and healthy controls, the delay differences and the coupling index.

因此我们将其命名为实时皮质肌肉耦合指数的定义。

That's why we named it defining a real time cortical muscular coupling index.

这就是我们使用的指标之一。

This was the index, one of the indexes that we use.

我们最初也采用了线性方法，但后来也转向了这个方法。

We also started with linear one, but shifted also to this one.

所以这就是这个指标。

So this is the index.

现在，我们开发了一个系统，使用ReTech作为记录设备的一部分，这个双模态系统能提供执行等长收缩和动态收缩任务所需的信息。

Now, we developed a system using the ReTech as a part of recording and a system, bimodal system, that provide you the information to do isometric and dynamic contraction task.

监测你施力时的情况，包括用脚或手施加的力度大小以及进行的收缩类型。

Monitor when you do the force, how much force you use your foot or your hand, and what type of contraction you do.

因此该系统能帮助我们同时记录脑电信号和肌电信号，配合神经反馈系统——即监测施力并让受试者了解应在传感器上施加多大力的系统。

So that system helps us alongside the recording simultaneously brain and muscle signals with the neurofeedback system, the system that monitors the force and subjects see how much force should insert on the sensor.

我们通过图形界面告知受试者需要将传感器施压到某个特定水平，受试者能实时看到这个施力值。

So we say that force this sensor until this level with the GUI and subjects see that force.

当实时观测到脑电信号与肌电信号时，数据会输入我们的算法。随后我们采用端点贝叶斯优化机器学习方法，基于非线性方法监测这种耦合关系。

And in the real time, when you see the brain signals and the muscle signal, it goes to our algorithm, Then we use the endpoint Bayesian optimization machine learning approach, and we monitor that coupling based on the nonlinear approach.

当受试者施力时，你能观察到耦合强度的增减变化，以及大脑向肌肉发送信号的延迟时间。

And you do the force in subjects, you see the coupling increasing or decreasing, the delay between sending function from the brain to the muscles.

延迟时长是多少？

What is the delay?

哪种任务类型能更清晰地区分健康人与患者的差异？

And what type of the task could show the difference between healthy and patients in more clear way.

如果没有其他问题我就讲到这里，虽然可以继续展开，但这次就先告一段落。

I would stop here if you have other questions, but I can go forward, but I will stop here for this time.

好的。

Yes.

非常感谢。

Thank you so much.

您真是一位出色的讲师。

You are such a wonderful lecturer.

我想任何大学都会很乐意请您去授课。

I think any university would be happy to have you to teach courses.

这太棒了。

That's amazing.

我相信学生们在您授课时都非常感激您。

And I'm sure that students so much appreciate you when you teach.

所以我非常确定情况确实如此。

So I I'm absolutely certain that that's the case.

非常感谢。

Thank you so much.

讲解得如此透彻清晰。

So well explained and so clear.

你通过动机和方法的选择，展示了万物是如何真正相互关联的。

And with the motivation and the approach you took and how everything is really connected, connected, you showed.

是的。

Yes.

你早期的研究，可能当时甚至没预料到它会在未来工作中扮演如此重要的角色。

Your earlier studies that you probably at that time even didn't anticipate how much role it can play in your future work.

对吧？

Yes?

确实如此。

Definitely.

毫无疑问。

Definitely.

是的。

Yes.

至于后续发展，在我获得博士学位后，他们再次给予了我尊重。

And for continuation, after my PhD, again, they had the respect.

我对他们所有人都心怀感激。

I appreciate all of them.

他们邀请我担任德黑兰大学神经科学与神经影像学系的助理教授，但我当时年轻，还有机会来美国深造。

They invited me to be assistant professor in neuroscience and neuroimage department of Tehran University, but I was young and I had the opportunity to come to The US to learn more.

在这里，我又从那些经历中学习成长。

So here, again, I learned from that experiences.

而且，我再次产生了将这些成果转化应用于精神病学领域的兴趣。

And again, I was more interested to translating this to the psychiatry.

我来这里向资深科学家们学习，今年也获得了在BCIS学院展示新研究成果的机会。

I came here to learn for more senior scientists, and that was my chance this year to present the new works in the BCIS school.

我们又一次在这里观察大脑。

And again, we look at the brain here again.

你知道的，当你具备数学思维时，总会先着眼于系统整体，其他事物就能成为你的输入素材。

You know, when you are mathematical, always look at the systems and then other things can be input for you.

你要建立一些基础，而不仅仅是某一个特定基础。

You build some foundations, not only a specific foundation.

我们系统地看待问题。

We look at the problem systematically.

明白吗？

Okay?

在伊朗他们教了我们很多数学，但当时我们并没有真正理解。

They teach us many math in Iran, you know, and we we don't feel them in that time.

也许后来，当我们不断深入并遇到实际问题时，才明白当初他们教我们的那些数学是什么。

Maybe after, you know, when we go higher and higher and see the real problem, then we found out what was that mathematics that they teach us?

大学里有很多数学课程，足足100个学分，各种数学内容。

Many units in the university, 100 units, mathematic, everything.

明白吗？

You know?

如今我发现，比如说什么是系统辨识？

And nowadays I found that, for example, what is the system identification?

明白吗？

You know?

大脑是一个系统，你进行许多输入和输出，以及它们之间的关系。

Brain is a system, and you do many inputs, many outputs, and the relation between them.

在这里，我们再次将这种观点转化应用于精神病学领域。

And here, again, we, we translate that point of view here again for psychiatry.

例如，通过TMS或现在使用的TFAs技术——经颅聚焦超声，实现无创的深部脑刺激。

For example, with TMS or currently now with TFAs, transcranial focused ultrasound, to stimulate the deep brain stimulation non invasively.

关于经颅磁刺激（TMS）和大脑皮层，你知道，我们整个职业生涯都在研究这个，虽然我还处于早期职业阶段，但这十年来，比如我们把大脑放在这里，用各种不同模式进行刺激，因为我们将每次刺激都视为输入信号——无论刺激类型如何——这是由系统特性决定的，然后我们观察响应。

For TMS, for cortex, you know, what we have done during the whole career, you know, I'm early career, you know, but for these ten years, for example, we put the brain here and we stimulate with different kind of modalities because we see every stimulation as an input signal no matter what type of the thing because of the system, and we see the response.

但每种刺激模式都有不同的作用。

So but each stimulation modality has a different role.

例如，TFOS能作用到更深层区域。

For example, TFOS doing the more depth.

而TMS，你知道，主要针对皮层区域。

TMS doing the, you know, for cortical one.

在伊朗，我们进行了听觉刺激和视觉刺激的研究。

And in Iran, we have done auditory stimulation, visual stimulation.

我在思考如何将这些整合在一起。

And I'm thinking how we can put all these together.

举个例子，让我们用一个简单的例子来说明它们之间是如何关联的。

For example, let's make a simple example how these are related to each other.

当你发送一个周期性视觉模式并记录脑信号时，你会看到基频及其谐波。

When you send a periodic visual pattern and record the brain signal, you see the fundamental frequency and the harmonics of that.

这被称为稳态视觉诱发电位。

And this is known as steady, steady visual evoked potential.

明白吗？

Okay?

这是经典研究中最早被使用的众所周知的方法。

It's well known in classic one, first one that people use.

我想展示其中的关联性。

I want to show the relation.

在听觉方面，存在一种现象。

In auditory, there is a phenomenon.

同样地，你发送一个周期性的听觉模式。

Again, you send a periodic auditory pattern.

当你记录大脑信号对应这种模式时，会看到一个稳定的、直接的听觉诱发电位。

When you record the brain signal recording that pattern, you see a steady, a straight auditory evoked potential.

在大脑的反应中，你再次看到基频和谐波。

In the response of the brain, again, you see the fundamental frequency and the harmonics.

没错。

Exactly.

人们还进行了感觉干扰实验。

And the people do perturbation, sensory.

他们通过连接手部传感器进行干扰。

They perturbate with connecting sensor to the hand.

明白吗？

Okay?

同样地，当你扰动周期性频率并记录信号时，你又会看到与此相关的稳态感觉反应。

And again, when you perturbate periodic frequency and you record the signal, again, you see a steady state sensory regarding to this.

所以如果你观察它们，它们看起来都一样。

So if you look at them, they all look same.

所以也许存在某种类型的神经元或网络结构。

So maybe there is type of neuron, type of network.

为什么当你改变刺激类型、感觉类型时，从信号频率的角度看，总会观察到相同的模式？

Why they always look if you change the stimulus, type of stimuli, type of sense, from the signal point of view, frequency point of view, why you see the same pattern.

这是我看到的一个简单例子。

This is one simple example that I see.

有时在脑机接口领域，我们会偏离实际，因为炒作，因为你知道的。

Sometimes in BCI, we go in you know, we deviate from the real things because of hype, because of I know.

我们需要这种炒作。

We need the hype.

我们需要这个东西。

We need this thing.

我们需要与时俱进。

We need to be update.

但有时那些花哨和科幻的东西比现实更重要。

But sometimes the fancy things and sci fi things are more weighted to the real one.

例如，哪些部分是有意义的？

For example, what is the meaningful parts?

你知道吗？

You know?

不仅仅是控制虚拟的东西，那些科幻玩意儿。

It's not just only controlling virtual something, you know, sci fi things.

明白吗？

You know?

我认为我们应该更关注每种模式的意义。

I think we should more focus on the meaning of every modality.

标记是什么？

What is the markers?

每当出现新的标记或新的模态，就为我们开辟了一条真正的研究和工作路径。

Every time there is a new marker or new modality, that makes a true a pathway for us to work and research on that area.

我认为现在很难划定界限。

That's I think that now it's hard to make a border.

脑机接口的定义是什么？

What is the definition of brain computer interface?

两天时间，这很难。

Two days, it's hard.

很难明确区分TMS、TFOS、神经技术、信号、大脑、AI这些概念。

It's hard to say that TMS, TFOS, neurotechnology, signals, brain, AI.

那么你如何定义传统的脑机接口、脑机交互、人机交互、人机协作呢？

So how do you define a classical brain computer interface, brain machine interface, human computer interaction, human computer.

如今这很难界定。

It's hard nowadays.

你看，现在所有领域都是相互交融、跨学科的。

You know, everything is mixed, multidisciplinary.

我认为寻找意义和关联很重要。

And I think finding meaning and relation is important.

这就是我学到的。

That's what I learned.

我现在正在思考如何通过TMS、TFOS再次处理那个指标。

And I'm thinking now with TMS, TFOS, how we can go through that index again.

我们之前有动态和静态的手部收缩。

We had the dynamic and static hand contraction.

现在如果用TMS作为输入，皮质肌肉信号会发生什么变化？

And now if you do, for example, TMS as an input, what happens to the corticomuscular signal again?

延迟会有什么影响？

What happens to the delay?

这次你知道了因果关系，知道输入添加后会发生什么。

This time, you know the causality, where you add the input and what happens after that.

然后，再次反馈这个结果。

And again, feedback of that.

我认为这正是我试图从你的视角去理解的观点。

I think this is the point of your perspective that I'm going toward that.

从计算的角度来看，我们应该有...好的，这些就是模态方法。

And from the computational point of view, we should have Okay, this is the modalities.

你实施干预措施。

You do the intervention.

通过监测效果。

From monitoring the effects.

所以我们使用神经影像技术。

So we use neuroimagines.

我们使用信号，这些同样只是模态手段。

We use signals, and these are just, again, modalities.

我们对它们进行的后处理、后分析操作，其实并不那么关键。

What we do on them, what we do post processing, post analysis, that's not so important.

有时我向资深前辈学习，有时则指导学生，有时我们会思考那些大型技术，比如fMRI，因为fMRI效果更好。

Sometimes I learn and I'm learning from seniors, but sometimes I'm working with the students, and sometimes we think about the huge technology, for example, fMRI, because fMRI is better.

不。

No.

有时候，例如，我们用传统的脑电图能揭示出比那更好的信息。

Sometimes, for example, we can reveal with classic EEG better information than that.

你知道吗？

You know?

这是另一件我在做研究时首先应该谨慎对待的事情。

That's another thing I should be, firstly, careful when I'm doing research.

有时候问题可以用更简单的方式解决，难题也能迎刃而解。

Sometimes questions can be solved, hard problems can be solved in simpler way.

有时候问题可以用最高深的数学方法解决，但我们必须谨慎。

Sometimes question can be solved with the highest complex mathematical way, but we should be careful.

首先，我应该谨慎地从那些复杂模式和复杂分析中提炼出意义。

I should be, firstly, careful to make meaning from that complex patterns, complex analysis.

比如当我在这里与临床医生讨论计算时，我们需要更谨慎地理解其含义、进行可视化呈现，并让他们感受到这些对他们意味着什么。

When I talk about computations here, for example, with clinicians, we should be more careful to make a meaning, make a visualization, make a feeling, what that means for them.

例如，我们在这里运用机器学习技术研究一些精神疾病的早期阶段。

For example, we are doing machine learning here for the technique early stage of some psychiatric things.

他们想知道机器学习背后的原理是什么。

They want to know what is the back of the machine learning.

你知道，现在的问题在于。

You know, the problem is that now.

原因是什么？

What is the cause?

是什么原因导致了这一预测结果？

What is the cause that makes this prediction?

这背后的原因。

This back of that.

因此我们正努力为临床医生解读这些信息，同时我也在向他们学习，获取他们的视角，共同实现这一目标。

So we are trying to make meaning for the clinician, and I'm trying to learn from them, get their perspective, and make this happen.

希望如此。

I hope so.

是的，完全正确。

Yes, absolutely.

非常感谢你。

And thank you so much.

你很好地展示了工程学背景如何为你提供解决神经科学问题、临床神经科学问题的通用方法。

You really illustrated so well how your background in engineering provides you the universal tool of how to approach neuroscience questions, clinical neuroscience questions.

基本上，你已经具备了一个完善的框架来处理所有这些复杂问题，并研究这些能带来真正价值（尤其是临床应用价值）的相互作用。

Basically, you have already a very well formed framework to work on all these complex issues and investigate all these interactions that can provide a real value and especially for clinical applications.

你很好地展示了这种背景与可创造的潜在临床解决方案之间的美妙联系。

So very nice connection that you illustrated between that background and applicable clinical possible solutions that you can create.

所以非常感谢，

So thank you so much,

Madhasa。

Madhasa.

谢谢。

Thank you.

是的。

Yes.

太棒了。

That's beautiful.

非常好。

Very nice.

我还是希望我们能回到你现在的工作上。

And I still would like us to get back to what you are doing now.

但在那之前，让我们先完成关于ALS项目本身的讨论。

But before that, let's finish talking about the project itself for ALS.

是的，当你研究心脏肌肉耦合指数时，你已经谈到了项目的启动、背景、问题以及你的解决思路。

Yes, when you were looking at Cardiac Muscular Coupling Index, You already talked about the initiation of the project, the background, the problem, how you approached it.

现在也许你可以总结一下结果。

Now maybe you can summarize the results.

那么你的主要发现是什么？

So what were your main findings?

确实如此。

Exactly.

完全正确。

Exactly.

好的。

Okay.

这是个很好的问题。

That's a great question.

是的。

Yes.

我们的发现是这样的。

What we found is this.

明白吗？

Okay?

我们发现了两个能揭示上运动神经元信息的标记物。

We found two markers that can unveil the upper motor neuron information.

第一，皮质肌肉耦合指数的强度，以及针对ALS患者应采用的适宜测试类型。

One, the strengths of the corticomuscular coupling index and the suitable type of test that you should use in ALS patients.

我们对手部和足部同时采用动态收缩与静态收缩两种测试方式。

We use both dynamic and aesthetic contraction tasks for foot and hand.

我们发现，在动态任务中，皮质肌肉耦合指数能够揭示健康受试者与患者之间的差异。

And we found that for dynamic tasks, the corticomuscular coupling index could reveal the difference between the healthy subjects and patients.

该标记物的强度变化表现为：先降低后升高。

The strengths of this marker, decreasing, increased.

明白吗？

Okay?

另一个关键指标是延迟时间。

The other thing is delay.

在健康对照组中，我们发现执行任务与不执行任务时存在显著延迟差异。

For the healthy controls, we found that when you are doing tasks and not doing tasks, there is a delay and it's significant.

这是应有的现象。

It should be.

对于患者来说，在执行任务时这种延迟不存在，可能是因为他没有执行任务、无法执行任务、肌肉不工作，或者耦合关系不存在。

For patient, again, for a task, this delay is not there because maybe he's not doing the task, he's not able to do the task, or muscles are not working, or the coupling is not there.

而这种延迟确实不存在。

And this is not there.

所以延迟和强度这两个指标。

So the delay and the strength.

我们发现强度变化的原因更多出现在疾病早期阶段，从上层到下层，从大脑到低级皮质。

And we found that the strength, the cause is more for early stages from upper to the lower, from the brain to the lower cortico.

脊髓束负责传递这些信息。

The spinal tract is responsible for sending these messages.

我们通过这种耦合系统展示了这一点，如今通过弥散张量成像（DTI）可以证明皮质脊髓束受到了影响。

And here we showed it with this coupling, this system, and is evidence nowadays with the DTI, diffusion tensor imaging, that the corticospinal tract is affected.

甚至经颅磁刺激（TMS）也显示这条通路受到了影响。

And even TMS also showing that this pathway is affected.

在非侵入性实验室联合脑电图和肌电图中，你也能再次观察到这个现象。

And in non invasive laby joint EEG and EMG, you see that again.

这与那些结果是互补的。

It's complementary of that results.

我很好奇。

And I'm very curious.

与你合作的临床医生提供了什么意见？

What was the input of your clinician with whom you worked?

他认为这些结果有多重要？

How relevant he found those results?

翻译好了吗？

Translated?

它们现在就能用了吗？

Can they be already used?

还是你需要做些额外工作才能将它们应用到临床实践中？

Or you need to do something additional to implement them in clinical practice.

完全正确。

Exactly.

感谢您提出这个问题。

Thank you for asking this question.

是的。

Yes.

目前对于帕金森病、ALS的类似病症，我们可以通过这种方式明确区分ALS与健康对照组的不同。

Now for Parkinson's, for mimics of ALS, we can do that to specify now for now we found ALS is different from healthy control.

所以现在真正的问题是，其他运动神经元疾病、其他皮质肌病的情况如何？

So the real question now is that, so what about other motor neuron disease, other cortical muscular disease?

如果它们与ALS相对于健康对照组都表现相同，那我们就面临真正的挑战了。

If all of them are same as ALS to the healthy control, again, the real challenge comes up for us.

那么我们该如何将其与类似病症及其他皮质肌病区分开来？

So how we can differentiate that with MIMICS and with other corticomuscular disease?

我们需要明确这一点。

We should specify that.

首要目标是确认这个指标、这个系统是有帮助的。

First point was to know that this is helpful, this metric, this system.

现在我们需要明确这一点。

Now we should specify that.

与模仿病症相比，ALS的特异性是什么？

What is specification of only ALS compared to mimics?

这是下一步需要完成的工作。

This is the next step that should be done.

关于这个应用有个有趣的故事，在我博士期间，存在这种差异，我的导师医师在场，最后他说现在无法判断他们测试的结果，因为他们是通过电话询问的。

And funny story about the application of this, in my PhD, there's this difference, the physician, my supervisor was there, and at the end, he said that now I can't tell what was the result of their tests because they are asking me through the call.

你知道，这才是真正的问题所在。

You know, this is the real question.

他们询问医生，那个系统对我们现在的测试结果是什么？

They asking physician, what was the result of that test on that system for us now?

老实说，我们应该做得更多。

And honestly, we should do more.

现在我们已有初步结果，但这正是需求所在。

Now we have preliminary results, but this is the need.

这就是我在定义新项目时学到的。

That's what I learned when I want to define new projects.

必须要有必要性和需求，并定义出真正的问题陈述。

There should be a necessity and need and define real problem statement question.

但人们需要这个。

But people need this.

人们需要知道他们所处的阶段。

People need to know what their stages are.

我们现在正在思考，如果你患有ALS会怎样？

And we are thinking now, what happens if you have ALS?

这个指标在纵向研究中会发生什么变化？

What happens to this index in longitudinal way?

如何告诉我们你的病情是恶化还是好转？

How can say to us your progress is worse or not?

因为如果你进行康复治疗，耦合强度的下降速度会减缓。

Because the strength of coupling goes up lower if you do rehabilitation.

你可以通过这些优势恢复，也可能不行。

You can recover with these strengths or not.

你可以改变延迟，也可能不行。

You can change the delay or not.

我们想再次沿着这条路走下去。

We we want to follow this path again.

完全正确。

Absolutely.

我希望你们能继续下去，因为这是非常有前景的工作。

And I hope that you will because it's really promising work so that you will continue with it.

从任何角度来看，无论是计算、方法论还是患者方面，这个项目最具挑战性的部分是什么？

What was the most challenging part of this project from any perspective, computational, methodological, patient wise?

最具挑战性的部分是什么？你们是如何解决这个挑战的？

What was the most challenging part and how did you solve that challenge?

问得好。

Good question.

ALS是一种罕见病。

ALS is an orphan or rare disease.

明白吗？

Okay?

寻找ALS患者很困难。

Finding ALS patients is hard.

当你找到ALS患者并设计临床试验时，要找到能够完成任务的患者，同时确保你的系统对他们足够友好，使他们能完成任务以便与LC对照组进行公平比较，这非常困难。

When you find the ALS patients, when you define clinical trials, finding the ALS patients that can do the tasks and your system be user friendly for them to complete the task to be a fair comparison to LC control is hard.

所以寻找ALS患者很不容易。

So it's hard to find the ALS patients.

与他们合作也很困难，因为他们无法完成任务。

It's hard to work with them because they cannot complete the task.

他们无法长时间坐着。

They cannot sit for a long time.

你不能给他们连接太多传感器然后说‘请耐心点’。

You cannot attach many sensors to them and say that, Be patient.

他们不受控制。

They are not controlled.

他们并不健康。

They are not healthy.

你不能随心所欲地做任何事，是的，有时你会有很棒的想法，但在实践中要考虑如何应用。

You cannot do whatever you want to do with Yes, sometimes you have great ideas, but in practice, how they can be applied.

你知道吗？

You know?

所以与他们共事很困难，但事实就是如此。

So work with them is hard, but it is what it is.

明白吗？

You know?

这既是从临床角度，也是从应用角度和方法论角度来看的。

And this is from the clinical point of view, from applicational point of view and methodological.

当你做脑电图时，你比我更清楚。

When you do EEG, you know better than me.

我们对人们说，保持冷静，不要咀嚼，不要笑，肌肉不要收缩，不要做任何动作。

We say to people, Be calm, do not jow, do not laugh, muscle contraction, do not something.

在这里，当我们想同时进行脑电图和肌电图时，我们需要他们进行肌肉收缩。

Here, when we wanted to do joint EEG and EMG, we wanted them to do the contraction.

在某种程度上，我们是在增加更多噪声。

In a way, we are adding more noise.

这就是我说的。

That's what I said.

皮质肌肉一致性在这里很有挑战性。

Corticomuscular coherence is challenging here.

因为在常规脑电图中，你会要求保持平静，闭眼或睁眼，保持静止。

Because in routine EEG, you say that stay calm, close or open eye, and stay calm.

并且不要有任何互动或收缩动作，以免在系统上产生噪声。

And do not any interaction, contraction to make noise on your system.

明白吗？

Okay?

所以在这里，你嗯。

So here, you yeah.

有一些软件包和方法可以降噪，但它们并不总是有效。

There are some packages, some methods for reducing noise and some but always they are not working.

你需要针对典型任务和典型问题谨慎调整预处理步骤。

You should be careful about your typical task, typical problem to adjust the preprocessing.

分析方法是原因之一，再次说明CMC有时会失效。

The analysis is the reason, again, CMC is not working sometimes.

文献中的方法未必适用于你的具体问题。

Always the methods that are in literature may not work for your problem.

所以这是另一个方面。

So this is another part.

实时性。

Real time.

当我们讨论实时性这个话题时，老实说，目前所有东西都还不能算真正的实时。

When we talk about the word real time always in the topics, Honestly, I can say it's not real time yet, all the things.

你们有一些延迟和我们的顶点，你知道，我们的主要目标应该是减少这些延迟，实现真正、真正、真正的实时性，明白吗？

You have some delays and our apex, you know, our main goal should be to decrease that and be real, real, real time, you know?

所以实时性并不总像我们说的那样，你说的实时意味着零延迟。

So the real time always is not like when we say real time, you you means no delay.

但总会有一些延迟。

But there are some delays.

你知道吗？

You know?

当你在下一个环节时，展示的是前一个、后一个的数据。

When you are in the next royal, you're showing the before one, latter one.

这是因为需要时间间隔来获取数据、分析数据，然后展示结果。

It's not because there should be a gap to get data and then analyze, then show that.

明白吗？

You know?

比如说，这次就不是实时的。

It's not, for example, this time.

比如这是针对你上一次的试验。

It's for your last trial, for example.

明白吗？

You know?

那我们就该解决这个问题。

Then we should work on that.

比如当你进行机器学习和更多计算时，有了训练部分后，就应该实时转换它。

And when you do, for example, machine learning and computational more, so you have a training part, you should transfer that in real time.

你知道，当你实时使用训练好的模型时，这又是个挑战。

And you know that when you use the trained models in real time, it's challenging again.

这也是个问题。

That's also a problem.

当你进行贝叶斯优化时，说到优化就意味着背后需要更多处理，所以耗时更长。

And when you do the Bayesian optimization, when you talk about optimization, it means on the back of that is more processing, so it's more time.

这些都是方法论上的挑战。

These are all challenging, methodological.

当我想撰写讨论部分时，最棘手的问题是如何让研究结果对临床医生具有实际意义。

And when I wanted to write the discussions, you know, the worst case, the main problem is that how you can define your results to be meaningful for a physician.

比如这个数学混沌或非线性动力学，从生物学角度究竟意味着什么？

I know that what is, for example, this mathematical chaos or nonlinear dynamic, what's the mean of biological points?

所以有时候真的很困难。

So sometimes it's hard.

我甚至很难向全程参与项目的联合导师传达这个信息。

It's hard for me to convey that message even to my co supervisor that is aware of every step of projects.

这很困难。

It's hard.

说实话，有时候这不仅仅是我个人的问题。

And sometimes, honestly, it is not only my problem.

在实际应用时，你往往需要赋予其特定意义，但有时根本无解。

There is no way to you are when you would apply that, you should make a meaning sometime.

这就是我在项目中始终面临的挑战。

Now I this is my challenge always in the project.

你做了复杂的计算，然后我该如何简化它，如何定义这些计算与真实生物学的关系，让神经科学和生物学领域的人能够理解。

You do the fancy computations, and then how I can make it simple, how I can define what is the relation with the real biology to be defined with neuroscience biology brain for them to understand your computation.

你提到了优化，但他们想要的是简单、有意义、用户友好的东西。

You said optimization, but they they wanted something simple, meaningful, user friendly.

有时候，用简单的方法解决实际问题、难题并不总是那么容易。

And sometimes it's not always simple to solve the real question, hard question with simple ones.

所以这里需要权衡取舍。

So there is a trade off.

有时候确实很难。

Sometimes it's hard.

对我来说，在撰写讨论部分时，可能因为我还处于早期阶段，需要更多学习和经验来掌握。

For me, when I write the discussions, maybe because I'm in the earlier stage, I need more learning, more experience to find that.

但我认为这是一个真正的挑战。

But this is a real challenge, I think.

是啊。

Yeah.

非常感谢。

Thank you very much.

是的。

Yes.

这些挑战定义得非常非常清晰。

Very, very well defined challenges.

我认为它们适用于该领域的许多研究。

And I think they're applicable to many studies in the field.

所以感谢你指出这一点。

So thank you for pointing that out.

正如你已经多次提到的CMC相干性与耦合的区别，以及在本研究中使用耦合而非相干性的重要性。

And as you already mentioned quite a few times this difference between CMC coherence and coupling and the importance in this particular study to use coupling instead of coherence.

也许你可以简要总结一下相干性与耦合的主要区别？特别是为那些不熟悉这个领域但确实想更好理解两者差异以及在不同情况下使用方法的人。

Can you maybe briefly just summarize what is the main difference between coherence and coupling for those who are not familiar with the field but definitely want to understand better the difference and the different ways to use it in different situations.

我会先回头定义一些术语和概念，然后再继续。

I will get back to define some jargons and definitions, then come back.

真正的问题是这些。

The real question is this.

什么是线性系统？

What is linear system?

什么是非线性系统？

What is nonlinear system?

什么是线性系统？

What is linear system?

什么是非线性系统？

What is nonlinear system?

在线性系统中，当你以特定频率输入时，输出端应能看到相同的频率。

In linear system, when you do the output with certain frequency, you should see that frequency at the outputs.

你无法产生新的频率。

You cannot make new frequencies.

在非线性系统中，你无法确定。

In nonlinear system, you don't know.

也许是，也许不是。

Maybe yes, maybe no.

明白吗？

Okay?

所以这类似于相关互信息。

So this is like the correlation mutual information.

问题就出在这里。

So questions are from here.

有时你无法用数学和参数方程来拟合这些问题。

Sometimes you cannot fit mathematic and parametric equation on the problems.

有些系统是未定义的。

Some systems are undefined.

它们没有闭合公式。

They have not closed formula.

它们是随机的。

They are stochastic.

它们是非线性的。

They are nonlinear.

而且你无法简化，用线性方程来建模它们。

And you cannot simplify, model them with linear equations.

有些模式是你无法做到的。

There are patterns you cannot.

你可以从线性部分获取信息，如果使用线性模型，你能得到线性的信息。

You can take part of the linearity from Yes, if you use linear models, you can get information, linear information.

但当系统呈现非线性时，它们同时包含线性和非线性信息。

But when systems are nonlinear, they have both linear and nonlinear information.

这部分会发生什么？

What happens to that part?

好的。

Okay.

至于相干性方面。

And for coherence.

相干性在某种程度上是相关性在频率空间中的转换，明白吗？

Coherence is a transformation of the correlation in frequency space in a way, okay?

这意味着你有一个水平轴表示频率，垂直轴表示相干值。

It means that you have a horizontal axis frequency, you have a vertical axis as the coherence value.

这是一种耦合方式。

It's a way of coupling.

是的，这是一种方式。

Yes, it's a way.

但这取决于频率。

But it's based on the frequency.

在哪个频率上你们能达成一致，实现同步，程度如何？

In which frequency you have clearance, have synchrony, how much?

但这是基于频率的。

But it's based on the frequency.

而清晰度背后的主要假设是，你应该为信号定义一个固定的窗口大小。

And what are the main assumption behind the clearance is that you should have defined a stationary window size for your signals.

那么这个关于非线性系统的假设是真实的吗？

So this assumption for the nonlinear systems is real?

是的，我们做了简化处理。

Yeah, we simplify that.

我们假设如此，但实际上，在事件相关电位中，在瞬态事件（如脑电信号）中确实如此。

We assume, yes, but happens in real, in events related potential, in events that are transient, like the brain signals.

所以这是原因之一。

So this is one reason.

CMC的另一个优势是它广为人知。

The other reason for CMC is good, well known.

我并不是说我们不应该使用它。

I'm not saying that we should not use it.

我们应该用它来捕捉线性部分。

We should use that to capture the linear parts.

但我们必须注意它对窗口大小、噪声和延迟的敏感性。

But we should be aware that sensitivity to windows, sensitivity to noise, and the delay.

你会看到频率，那么延迟是多少？

You will see the frequency, and so what is the delay?

这个问题中提到的内容在哪里？

Where is that in that question?

有些研究开始加入相干性，即带有时滞的相干性。

Some studies come and add to the coherence, coherence with time lag.

是的，它们更花哨一些。

Yeah, they are more fancy.

这是一个起点。

This is a starting point.

项目的一部分涉及线性部分。

One part the project was linear parts.

我们从众所周知的指标开始。

We started with the well known metric.

但耦合，CMC可以是皮质肌肉耦合，或传统方式，即皮质肌肉相干性。

But the coupling, CMC could be corticomuscular coupling, or classical way, corticomuscular coherency.

但耦合是一个更通用的术语。

But coupling is a more general word.

当你讨论耦合时，它就像一种连接性，明白吗？

When you talk about coupling, it's like a connectivity, okay?

耦合。

Coupling.

你可以在这个总称下定义其他度量标准。

You can define other metrics under that umbrella, under that word.

现在你可以说相干性也是耦合下的一种方法。

Now you can say that coherence also is a one method under coupling.

非线性动力学的连接性度量是一种方法。

Nonlinear dynamics metrics for connectivity are one method.

互信息，一种方法。

Mutual information, one method.

每种连接性测量都可以归入这个总称下。

Every connectivity measure could be under that umbrella.

所以它不只是一个方法和线性方法。

So it's not just one method and linear method.

在耦合中，你可以定义许多方法。

In coupling, you can define many methods.

我们采用基于联合递归图的非线性动态视角来展示同步性和时间延迟。

And we use the nonlinear dynamic perspective based on the joint recurrent plots to show the synchronization and time delay.

非常精彩的解释。

Very sort of very nice explanation.

我相信我们的许多听众都会很欣赏这一点。

I'm sure many of our listeners will appreciate that.

非常感谢。

Thank you very much.

我的下一个问题还是关于研究发现。

And my next question is about again the findings.

你们是否发现了一些出乎意料的结果？有没有什么是在研究开始时完全没想到会发现的？

Was there anything that you found that maybe was somewhat unexpected, something that you didn't even think of finding when you were starting your research?

是的。

Yes.

这确实是个常见问题，在我的答辩环节和各种场合总被问到。

This was a real question that always in my defense sessions and all the things they asked me.

这也是我经常深入思考的一个问题。

This is one that I challenge a lot also.

当你设计实验任务时，文献评审者往往只关注特定脑区(ROI)。

When you do tasks, when you define tasks, mainly people, literature reviewers, think about specific ROIs.

比如进行运动任务时，人们认为大部分信息应该来自运动皮层。

For example, when you do the motor task, people think that most of the information should come from the motor area.

但我从导师那里学到的是，当我们结合听觉和视觉输入，在进行动态和等长收缩时，运动系统的表现——从结果来看——并不仅源于那个特定脑区。

But I think what I learned from the master, when we combine the auditory and the visual and hear, when you do the dynamic and isometric contractions, What our motor system do, what I see in the results, is not just because of that ROI.

大脑其他区域也参与其中。

The other parts of the brain also engage.

而且我发现某些非运动区域的表现甚至比运动区更有趣，这些区域的参与不容忽视。

And I found some part of it more interesting even than motor, and I cannot rule out the engagement of other parts.

这就是我说的具有挑战性的问题之一。

That's one of the questions that I said that is challenging.

我们如何将其与例如你说的运动任务联系起来。

How we can relate this to the for example, you say motor task.

即便你将其命名为运动任务，你仍想讨论例如信息的时间部分。

Even you named it motor task, then you want to talk about, for example, temporal parts of the information.

当你发现耦合时——因为你有高密度脑电通道，有时某些感觉区域会更活跃，如前额叶。

When you find the coupling because you have the high density EEG channel, sometimes you have some sensory more engaged, prefrontal.

这意味着当我们进行运动时，人类不是单一维度的。

So it means that when we do motor, human is not one dimensional.

你正在调动多种机能。

You are engaging multi things.

而当存在闭环时——上行、下行通路，有时我们认为所有信号都源自大脑。

And when you have a closed loop, ascending, descending, sometimes we think that all things come from starting from the brain.

但我们也发现，在某些任务中，从下至上的反馈才能真正带来改变。

But we also found that sometimes in some tasks, feedbacks from the lower to higher makes the real change.

过去几十年里有个定义，我们总是将脑机接口视为解码过程。

And there was a definition in decades about we always think about BCI as a decoding.

来自大脑的信号，我们就应该解码。

What comes from the brain, we should decode.

而现在有另一种看待问题的方式——大脑的编码。

And now there is another way of looking the problem, encoding of the brain.

比如对于盲人，虽然没有眼睛，但可以在视觉皮层进行编码来绕过眼睛。

For example, for blind people, you don't have eye, but you can do an encoding on your visual cortex, for example, to bypass your eye.

这部分需要更多关注，我觉得很有意思。

This is the part that needs more attention, and I found it interesting.

这就是我说的，有时来自外部的反馈传入大脑会产生影响，并不总是源自大脑内部。

That is the thing that I'm saying that sometimes feedback from the outside going to the brain makes the things, and it's not always comes from the brain.

这是我们在某些任务中发现的，特别是动态任务中反馈非常重要。

This was that part that we found in some tasks, especially for dynamic tasks are important feedbacks.

是的。

Yes.

这是个很棒的观察。

That's a great observation.

而且重点不仅在于解码，还包括用可以输入大脑并用于不同目的的新信息对大脑进行编码。

And also the emphasis on not only decoding but encoding the brain with the new information that can be fed into the brain and used for different purposes.

是的，非常感谢。

Yes, thank you so much.

我的下一个问题是，你们在这个项目中采用了哪些主要的创新或技术创新的方法，可能是之前未曾使用过的？

My next question is what would be the main innovations or maybe technological innovations approaches that you used for this project that maybe haven't been used before?

老实说，如果你仔细看，我们系统的每个部分之前都存在。

To be honest, if you look at that, every part of our system was there before.

在记录方面，我们采用互补视角，然后在系统后端进行计算。

The recording, we use the perspective to how to make it complementary and then use the computation on back of the system.

比如你看RECOVER项目，RECOVER XNG拆卸时他们就没使用肌电图。

If you look at RECOVER, for example, RECOVER XNG takedown, they don't use the EMG.

在某种程度上，我们完成了那个项目，但同时使用了肌电图。

In a way, we have done that project, but using also EMG.

你知道吗？

You know?

比如说，你看看GTEC。

If you look at, for example, GTEC.

GTEC是他们自己的系统。

So GTEC is their system.

而生物反馈系统，我们用的是自己的系统，并且拥有专利。

And biofeedback system, we use our system, and it has a patent.

但从某种程度上说，硬件方面并没有巨大的技术创新。

But in a way, there is not a huge technological innovation of hardware.

没有。

No.

但所有部分对于实现问题创新、解决问题都是必要的。

But all parts are necessary to make innovation in the problem, you know, solve the problem.

这就是所有部分的组合。

That's the combination of all parts.

但并不是，比如说，花哨的或者你知道的？

But no, for example, fancy or you know?

我们使用了现有的系统。

We used a system that was there.

我们利用了它的双模态性来监测传感器，处理不同类型的任务，并进行实时计算将所有部分耦合起来做出预测。

And we used the, I think, the bimodality of that to being able monitor sensor, different kind of tasks, and making a real computation to couple all parts together to make the predictions.

这就是有时候我们谈论的——从一开始这对我来说也很困惑。

That's what's the that's sometimes when we talk about the this was also for me from the start, it was confusion.

当人们谈论新技术、新创新时，他们总是认为需要全新硬件、大型芯片、图形处理，或者类似科幻视觉、机器人技术之类的东西。

When people talk about new technology, new innovation, they always think that the bulk of the new hardware, huge chips, graphy, or, you know, like science, vision, robotics or something.

但我认为创新和突破往往发生在软件部分，主要是在计算层面。

But I think the innovation and the things happens on the soft part of it, you know, in computation mainly.

我认为新系统——当我们谈论系统时，我们想到的是硬件，但它不仅仅是硬件。

I think new systems, when we talk about system, we think hardware, but it's not just hardware.

我认为问题本身、结果以及硬件背后的计算才构成完整的系统，而不仅仅是物理部件。

I think that the questions, the results, and the computation on the back of the hardware is the system, not just physical part.

是的，物理部分有时是提供工具让你提问的关键，但有时并非如此。

Yeah, physical part sometimes is key to provide you tools to ask your question, but sometimes not.

所以我们在这里使用传统的脑电图系统，但也能获取肌电图信号。

So here we use the conventional EEG system, but to be able also acquire EMGs.

我认为这很简单，每个人都可以验证这个问题。

I think, is simple, and everyone can test this question.

每个人都有脑电图设备，只要那台设备能获取肌电信号，他们就能做到。

Everyone has EEG just to be able if that EEG is capable to get the EMG, they can do it.

他们甚至不需要设备就能通过肌肉收缩创建自己的神经反馈。

They can create their own neurofeed by contraction even without that.

假设他们不想用我们的系统进行收缩。

Let's assume they don't want our system to contract.

他们可以做简单的操作，同时采集脑电图和肌电图信号。

They do simple ones and acquire simple EEG and EMG together.

你明白吗？

You know?

但他们应该清楚自己想要用这些联合信息做什么类型的分析。

But they should be aware of what type of analysis they want to do with the joint information.

是的。

Yes.

完全正确。

Absolutely.

而且这个想法本身确实非常具有创新性。

And even the idea of itself, yes, was very innovative.

当我们谈论创新时，首先关键在于理念以及如何实施，正如你所说，利用现有资源，是的。

And when we talk about innovation, it's really about the idea first of all and how we implement, like you said, with what you had, Yes.

你实际上已经成功实现了，而且没有使用非常花哨昂贵的设备和侵入大脑的方法。

And you actually implemented and very successfully without really very fancy and expensive equipment and approaches and going inside the brain.

我认为你最大限度地利用了当时已有的资源产出。

I think you maximize everything, the output of what you had available already at the moment.

当然我们也看到技术和软件以及许多计算手段正在快速发展。

And of course we see that the technology and software and many computational things are developing right now.

是的，我们正经历人工智能的繁荣发展。

Yes, we have this boom in AIs.

众所周知，现在有许多神经科技公司。

All are aware about, we have many neurotech companies.

你今天提到了NG Tech公司，还提到了埃隆·马斯克。

You mentioned today NG Tech, you mentioned Elon Musk.

这些事物中有哪些启发了你的工作？

What is from all of that that inspires you in your work?

你从哪里获得灵感

Where are you getting the inspiration

来从事你正在做的事情？

for what you are doing?

你知道，当我们在伊朗确定论文方向时，需要花好几个月进行文献综述。

You know, when when we define our thesis in Iran, you should have a literature review for many months.

明白吗？

Okay?

那么什么是有启发的呢？

So what was inspiring?

从起点来说，你知道，当他们定义PCI系统时，他们讨论的是基于模态的方法。

From a starting point, you know, when they define the PCI systems, they talk about modality based.

对我来说什么是有启发的？

What was inspiring for me?

例如，他们提到基于P伽马的，还有SSVP，比如运动想象。

For example, they say that P gamma based, and SSVP, for example, motor imagery.

他们用模态来命名PCI系统。

And they name the PCI system with modalities.

明白吗？

Okay?

那么新的方法是什么，如果你发现了新方法？

So what is the new ways, if you find new ways?

所以我们应该始终研究那些模态，运动想象，外源性或内源性的。

So we should work always on that modalities, motor imagery, exogenous or endogenous.

对于内源性信号，你经过处理后就放在那里。

For endogenous, you go through it, you put it there.

那么新方法是什么呢？

So what is the new ways?

这就是让我受启发的地方。

That was my inspiring.

是他们定义的。

They defined it.

他们根据模态对脑机接口进行了解析。

They dissect the BCI based on modalities.

非侵入式的。

Non invasive.

在侵入式之后，非侵入式方面他们提到基于脑电图。

In invasive, then in non invasive, they say that EEG based.

在侵入式方面，他们举例说是基于植入体的。

In invasive, they say that implant based, for example.