AI in Radiology: Are Radiologic Technologists Being Replaced or Redefined?

我问你个问题，你先好好想一想。

Let me ask you something and I want you to really sit with it for a second.

如果你的部门明天告诉你，要上线人工智能系统，你是会兴奋，还是会心里一沉？

If your department told you tomorrow that they were rolling out AI, would you feel excited or would your stomach drop?

因为如果你真有这种感觉，那你不是一个人。

Because if you felt that drop, you're not alone.

我遇到过不少放射技师私下找我，或者在会议中拉住我，问我：‘Chaun，人工智能会不会抢走我们的工作？’

I have had radiologic technologists DM me, pull me aside at conferences, asking things like, Chaun, is AI coming for our jobs?

说实话，我当时真答不上来，没有一个让我自己觉得有把握的答案，更没有一个来自一线影像科室内部、真正了解这项技术怎么落地的视角。

And here's the truth, I didn't have a great answer, I didn't have one that I really felt confident in, Not one that was coming from someone on the inside of how this technology actually gets deployed in real imaging departments.

我得去找个懂行的人问问。

I would have found someone.

不过我得坦白，这正是我们在《放射技师说》这个节目里的做事方式。

Now, full transparency, because that is how we do things here on a couple of Rad Techs.

本集由安科锐放射解决方案冠名赞助。

This episode is brought to you in partnership with Agfa Radiology Solutions.

他们赞助了这次对话，但如果你一直听这个节目，你就知道我从不让人买断我的观点。

They sponsored this conversation, but if you've been listening to this show for any length of time, you know I don't let anybody buy my opinion.

我允许他们带来值得交谈的人，比如乔丹·赫米勒，绝对值得聊一聊。

What I do let them do is bring people to the table worth talking to, and Jordan Hermiller, worth talking to.

那我们直接开始吧。

So let's get into it.

今天的嘉宾，起点和你一样——在诊室里操作透视、为病人摆位，后来他成了管理者、大学讲师，还发表了研究论文。

My guest today started exactly where you are, in the room, running fluoro, positioning patients, Then he became a manager, university instructor, published researcher.

我知道这履历还能继续列下去。

I know the list is going on and on.

但他真的值得你耐心听完，朋友们。

He is worth the wait, you guys.

之后他转到了桌子的另一边，现在负责阿加法放射解决方案公司北美团队，这家公司正在应用AI影像技术，他直接与技术、使用技术的临床人员以及接受培训的团队打交道。

Then he went to the other side of the table and now he's managing the North American team at Agfa Radiology Solutions, a company that uses AI imaging where he works directly with the technology, the clinicians using it and the teams being trained on it.

他当过技师、管理者、教师，如今站在临床实践与真实世界AI落地的交汇点上。

He's been a technologist, a manager, a teacher, and now sits at the intersection of clinical practice and real world AI implementation.

他并没有制造这台机器。

He didn't build the machine.

但他懂得怎么使用它、评估它，还能教你我这样的技师如何思考它。所以他谈你科室里的AI时，真的懂你的处境。

He knows how to use it, evaluate it, teach other technologists like you and I how to think about it, which means when he talks about AI in your department, he knows what it's like to be you.

乔丹·赫米勒，RT(QM)，持有众多认证。

Jordan Hermiller, RT(QM) with so many certifications.

欢迎来到《放射技师对话》节目，乔丹。

Welcome to a couple of Rad Techs, Jordan.

谢谢你，肖恩德拉。

Thank you, Chaundria.

这介绍也太棒了。

And that was an unbelievable introduction.

我从来没被这样介绍过，真的非常感谢。

I've never had an introduction like that before, so I really appreciate it.

很高兴能来

You Happy to be

你们值得欢迎，也配得上这份认可。

are welcome and you are deserving.

我们在RSNA见面时，我亲眼看到了Jordan的实际表现。

We met at RSNA and I got to see Jordan in actions.

我是有切身体会才这么说的。

I'm talking from experience.

在RSNA的时候，我简直被震撼到了。

I was blown away at RSNA.

那是我第一次参加。

Was my first time going.

感谢阿法公司邀请我。

Thank Agfa for inviting me.

我还看到了一家真正依靠技术人员的公司。

But I also got to see a company that uses technologists.

我的意思是，像你们这样的岗位，我以前真没见过这样的情况。

I mean, I don't think I've ever seen that before in the positions that you guys are all in.

大多数人也都经历过我们现在的阶段。

Most people have been where we are.

他们能从技术员的角度来谈。

They can speak from a technologist role.

乔丹，跟我们说说你到底是怎么走到今天的。

Jordan, give me the real version of how you got here.

我知道你为一家了不起的公司、为我们整个行业做了很多出色的事。

You know, I know you're doing some great things for an amazing company and for our profession.

能跟我们讲讲，你是怎么从放射科技术员一步步走到今天的吗？

Tell us how did things go from radiologic technologists to where you are now?

当然了，我很喜欢别人问这个问题。

Sure, I love being asked this question.

让我对放射成像产生兴趣的是，我有几位姨妈是护士，她们经常聊医疗方面的事，而我一直对帮助他人、从事医疗工作感兴趣，她们就建议我考虑放射科。

What got me interested in radiography was they have a couple of phenomenal aunts who are nurses, and they actually were talking about healthcare and I had an interest in healthcare and helping people and they suggested radiology.

我觉得像我们这样的技术员，大多都是这样——既关心医疗、想帮助别人，又对技术本身充满热情。

And I think like a lot of us who are technologists, there is that kind of combination of, we're interested in healthcare, we wanna help people, but we're also interested in technology.

这简直可以说是天作之合。

And it's really kind of a perfect marriage, if you will.

所以，正是这一点让我对放射影像和放射学产生了兴趣。

And so that's what got me interested in radiography and radiology.

我去了俄亥俄州立大学，拿到了放射科学的学士学位。

I went to Ohio State and I graduated with a bachelor's in RadSci.

之后，我在俄亥俄州立大学的一家骨科诊所当了几年技术员。

And from there, I worked as a technologist for several years at Ohio State at an orthopedic clinic.

后来，我对管理产生了兴趣。

And then I had an interest in leadership.

于是我回学校攻读了医疗管理硕士学位。

I went back to school to get my master's in healthcare administration.

那时候，我开始兼着教课，到现在我依然还在兼职教书。

And at that time, I actually started to teach as well, which I still do part time today.

我原以为自己会一辈子在医院工作。

I thought I was gonna work in the hospital the rest of my life.

我回学校读书，本来就是奔着这个目标去的，你知道的。

That was the goal of going back to school even, you know.

但我认识一个人，他认识的人里正好有个在安科公司要走的人。

But I knew someone who knew someone who knew someone was leaving at Agfa.

于是他们就问我：你有没有兴趣考虑一下这样的职位？

And so they say, Hey, would you be maybe interested in this type of role?

一开始我心想，算了，我太容易给自己贴标签了，总觉得自己就该干这一行。

And at first I thought, no, again, it can be so easy to, I say pigeonhole, but put yourself in this is what I want to do.

我知道，你也经常提到这一点。

You know, I know it's a lot of what you talk about as well.

我开始认真琢磨这份工作，越想越觉得，这真是把我的两个兴趣完美结合了——既有领导力的成分，又能发挥创意，推动放射学的发展。

I started really thinking about the job, thinking about the job, and it sounded like really a great blend of my interests of, know, one leadership, but even like being able to be creative with where radiology can go.

比如放射学和X光技术的未来，你其实能影响技术发展，帮技师们，也帮患者。

The future of radiology and x-ray, how you can actually help influence technology to help technologists, to help patients.

我觉得这事儿听起来真不错。

And I thought this actually sounds pretty awesome.

所以我已经在爱克发工作五年了，说实话，我一点都不后悔。

And so I've been with Agfa before, on five years now, and I tell you, I don't regret it at all.

我真的特别喜欢这份工作。

I really love it.

我们团队很棒，在供应商这边工作，能用到很多不同的技能。

We have a great team and you really get to use a lot of different skills working on, I suppose, the vendor side as well.

你这段经历正好完美解释了，为什么你特别适合谈这个话题。

That is a wonderful opener for why you are so qualified to talk about this conversation.

你出身于技术员背景，但我特别欣赏你职业上的转型。

You come from a technologist world, but I love how you pivoted in your career.

不过我想聊聊人工智能。

But I do wanna talk about that AI.

人工智能其实并不新鲜。

AI is not new.

它早就成了大家讨论的焦点，尤其是在放射科圈子里。

It's kind of been in the conversations everywhere, especially radiology conversations.

计算机辅助检测已经存在很多年了。

Computer aided detection has been around for years.

为什么现在在放射学的讨论中，AI好像无处不在呢？

Why does it feel like AI is everywhere in the radiology conversations right now?

到底发生了什么变化？

What's changed?

是啊，这个问题问得好，肖恩德拉。

Yeah, I mean, great question, Chaundria.

正如你提到的，你四个月前刚参加过RSNA会议。

As you mentioned, you were at RSNA four months ago already.

真不敢相信时间过得这么快。

Hard to believe.

在RSNA，你走不到十步就能看到AI的宣传到处都是。

Feels like it was yesterday, but you can't walk more than 10 feet at RSNA without seeing AI plaster everywhere.

而且这不只是RSNA的情况。

And it's not just RSNA.

《放射学刊物》已经连续第四年将人工智能评为放射学领域最重要的新闻事件。

Radiology publications has voted AI the most significant news event in radiology for the fourth year in a row.

正如你提到的，如今放射学的重点仍然放在帮助放射科医生解读影像、发现病变上。

As you mentioned, today, in radiology is still primarily focused on radiologists, helping with their interpretations, helping with pathology detection.

我在阿加法的工作重点，也是我们团队的核心，是在诊疗现场的应用。

My work at Agfa, my focus and our focus is at point of care.

它帮助技术人员提高操作的一致性和效率。

It's helping technologists be more consistent, more efficient.

现在诊疗现场有AI的应用吗？

Does AI exist at point of care today?

有的，而且确实已经落地了。

Yes, it absolutely does.

我觉得我们正站在一个关键的转折点上，未来会怎样，我相信我们待会儿会深入聊。

I also think we're kind of really on the edge of what's possible and where we're going, which I know we're going to talk about.

或许更具体一点，来回应你关于‘为什么是现在’这个问题？

You know, maybe more specifically, to answer your question about why now?

为什么现在突然就爆发了呢？

Why does this seem to be blowing up now?

我觉得主要有两个因素正在交汇。

I think there's really two main factors that are converging.

一个是人力短缺的问题。

Is this workforce shortage.

第二个是技术本身，也就是AI技术的快速发展。

And the second is the technology itself, the AI technology itself, which is expanding.

说到人力短缺，如果你去问任何一位放射科负责人，他们面临的最大挑战是什么，肯定会提到资金压力和人员不足。

Regarding the workforce shortage, if you were to survey any radiology leader and ask them what their primary challenges are, they're going to say financial pressures and the workforce shortage.

根据ASRT的数据，短短几年内，空缺率就从6%飙升到了18%。

ASRT data, they said that vacancy rates tripled in just a few years from six to 18%.

我觉得这也是为什么你所做的工作如此重要。

Again, I think that's one of the reasons why what you're doing is so much.

让更多人对成为放射技师产生兴趣，甚至更加热爱这份职业。

Getting people excited and more excited about being a radiographer, being a technologist.

但人工智能几乎能影响任何一种工作流程。

But AI can impact essentially anything, any type of workflow.

因此它能帮助缓解人力短缺的问题。

And so it can help meet this workforce shortage.

这正是为什么现在突然爆发的原因之一。

That's part of the answer of why it's exploding now.

另一个原因，还是在于技术本身。

The other part, again, is the technology itself.

我会尽量说得概括一些，但我觉得很有必要先搞清楚人工智能到底是什么，它能怎么帮上忙。

I'll try to keep this somewhat high level, but I do think it's important, you know, to start to understand what is AI, how it can help.

还有就是，为什么它最近才突然兴起。

And again, why it's kind of taken off so recently.

大概有三个方面。

There's kind of three things.

首先是所谓的深度学习取得了重大突破。

One is the advancements in what's called deep learning.

明白吗？

Okay?

深度学习其实是机器学习的一种。

Deep learning is actually a a type of machine learning.

为了更清楚一点，机器学习讲的是AI如何通过数据学习，而不是靠人工编写明确的指令。

And so maybe just to explain this a bit more, machine learning explains how AI learns, which is from data rather than being explicitly programmed.

在传统的机器学习里，人还是得告诉AI该找什么、该测什么。

With traditional machine learning, a human still has to tell the AI what to look for and what to measure.

而深度学习直接跳过了这一步。

Deep learning actually skips that step entirely.

你只要给它目标和例子，它自己就能找出需要关注的特征，对吧？

Essentially, you give it the goal and the examples, and it develops its own features to look for, right?

这些特征是我们人类根本想不到要去测量的。

These are things that us humans would never think to measure, right?

这正是它在医学影像领域如此强大的原因。

That's what makes it so powerful in medical imaging.

比如，工程师在编写病理检测程序时，可以设定测量皮质厚度，但没人能手动编写出识别所有人体结构差异中细微骨折的代码。

For example, an engineer, if they're trying to code for a pathology detection, they could code measure cortical thickness, but no human can code, detect every subtle fracture across every variation of human anatomy.

这根本做不到，也不现实。

Not possible, not realistic.

而深度学习能发现那些我们根本想不到要去找的东西。

And deep learning can find these things that we wouldn't even know to look for.

其中一个关键点是深度学习。

The one aspect is deep learning.

另一个是大数据的爆发式增长。

Another is this explosion of big data.

想想我们每天从社交媒体、智能手表，甚至电子病历中产生的数据，健康记录就占了全天生成数据的三成。

When you think about all the data we create every day from social media or smartwatches, even health records make up 30% of the data that's generated every day.

众所周知，AI正是靠这些数据变得越来越高效、越来越精准。

As we know, AI thrives on this data to become more effective, more accurate.

最后一点是计算能力的提升。

And then the last piece is the improvement in computing power.

所以如果有玩家在听，你们知道那些显卡吧？其实最初是为游戏设计的，结果发现特别适合训练这些人工智能模型。

So there's any gamers listening, you know, these GPUs, right, they were actually created initially for gaming, but they turned out to be highly efficient for training these AI models.

从那以后，各大公司就开始专门研发这类芯片了。

And so since then, companies have developed these chips specifically for that purpose.

最近技术进步这么快，是因为算法、数据和算力这三者正好凑到了一起。

The technology has advanced so rapidly recently, because the algorithms, the data, and the computing power have all come together.

而且别忘了，放射科其实天生就适合人工智能。

And then on top of that, you also have to consider radiology was really built for AI.

你想啊，我们这个领域本来就是技术驱动的，这是第一点。

You know, if you think about how technology driven our field is, one.

第二点，我们早就有了数十万张带有标注病变的数字化X光片，简直是完美的训练数据。

Two, there's we already had hundreds of thousands of digitized x rays that have labeled pathologies that are perfect, perfect data training data for these AI algorithms.

现在美国已经有超过1300个获FDA批准的AI医疗设备，但更说明问题的是，其中超过75%都集中在放射科。

And so today, there's over 1,300 FDA approved AI medical devices, right, in The United States, which But is what's telling is over 75% of those are in radiology.

另外，不少设备还获得了CMS的‘附加支付’认定，这一点其实相当关键。

And second, a number of them have what's called an add on payment designation from CMS, which again, actually is pretty important.

这说明这些技术不仅仅是有趣好玩的玩意儿。

So that tells us that these aren't just kind of fun, interesting technologies.

它们实际上已经被证明能为患者带来实实在在的临床价值。

These are actually proven to offer real value in clinical care to patients.

也许总结一下，第一，是人力短缺带来的影响。

Maybe just to kind of sum it up, you know, one, it's the workforce shortage being able to impact that.

第二，最近技术发展得特别快。

Two, the technology has advanced so rapidly, somewhat recently.

最后，放射科的环境特别适合AI发挥优势。

And lastly, radiology was really built where AI can excel in this area.

说到技术人员，因为我们现在聊的是技术人员，他们对AI和放射科最大的误解是什么？

When it comes to technologists, because we're talking to technologists, what do they misunderstand most about AI and radiology?

当然。

Sure.

我觉得这也是个非常好的问题。

I mean, I think that's also a great question.

我觉得，关于放射科的AI，最大的误解之一其实是AI到底是什么。

And I think honestly, one of the biggest misconceptions when it comes to AI in radiology is actually what is AI, right?

你得知道，不同行业对什么是AI的定义差别很大。

You have to consider there's a wide range of definitions that's used by industries to define something as AI.

现在AI这么热门，厂商们正好利用这种模糊性来做营销。

And with all the buzz around AI, vendors, you know, can take advantage of that ambiguity in their marketing, right?

比如，如果你把AI定义为任何模仿人类智能的东西——你去谷歌搜一下‘什么是AI’，

For example, if you define AI as anything that mimics AI, like if you were to Google, what is AI?

其中一个定义就是‘能模仿人类智能的东西’，

One of the definitions is like something that mimics human intelligence, right?

比如人类的决策过程。

Human decision making.

如果按这个标准来定义，那其实很多功能都能算作AI。

Well, if that's how you define it, then you can qualify a lot of features as AI, actually.

没错。

Right.

举个例子，自动曝光控制，也就是AEC，就是一个很好的例子。

I mean, I think one good example is the automatic exposure control, the AEC.

AEC确实是为了模拟人类的决策过程而设计的。

Certainly, AEC is used to help mimic human decision making.

作为技术人员，我们当然要对自己的操作技术负责。

Now, as technologists, yes, we should always be responsible for our techniques.

但要说AEC完全不参与决策，那也太不现实了。

But of course, it'd be unrealistic to say that AEC doesn't take some of that decision making from us.

AEC已经存在几十年了。

AEC has been around for decades.

如果今天才推出这项技术，肯定有人会贴上AI的标签，管它叫智能剂量之类的，但它根本算不上真正的AI。

Undoubtedly, it were released today, someone would slap an AI label on it, call it intelligent dose or something, but it's not true AI.

它不是机器学习。

It's not machine learning.

实际上，我会为放射科的培训项目和医院科室做一场关于AI的讲座，这是一场获得继续教育认证的在线研讨会。

And actually, I give a presentation on AI to radiography programs, to hospital departments as a CE approved AI kind of webinar.

我在做这个演讲时，通常会讲一个有趣的故事，来说明真正的AI和传统基于规则的计算之间的区别。

And what I start with in that presentation is actually kind of a fun story to illustrate this difference between true AI and kind of traditional rule based computing.

我讲的是两个机器人朋友的故事。

I tell a story about two two robot friends that you have.

一个叫诺姆，代表传统计算；另一个叫阿蒂，代表人工智能。

One's name is Norm for normal computing, and one is Artie for artificial intelligence.

诺姆和阿蒂都很擅长遵守规则。

And both Norm and Artie, they're great at following rules.

但不同的是，阿蒂还是个很棒的学习者。

But where they're different is RT is also a great learner.

你可以给他看一些例子，他就能自己从这些例子中推断出各种情况。

So you can show him examples, and he can use those examples to figure out different situations on his own.

这在放射科特别重要，因为每张X光片都不一样，每个病人也不一样。

And that's quite important in radiology when no two x rays are the same, no two patients are the same.

这样的能力让阿蒂真的很有价值。

That the RT can really be quite valuable.

接着，我还进一步讲了个故事，说诺曼其实也特别喜欢看书。

And then I also build on that by kind of telling a story about, you know, Norman already also like to read books.

你先听我说完。

Hang with me here.

就是《威利在哪里》那种书，算不算读书你说了算。

It's Where's Waldo, if that counts as reading.

你可以给诺曼和阿蒂看成千上万页的《威利在哪里》，每一页都明确指给他看威利藏在哪儿。

And so you know, you can show both Norman already thousands of pages, Where's Waldo, point out exactly where Waldo is on each page.

下次他们再看到这些页面时，马上就能告诉你威利在哪儿。

The next time they look at those pages, they can immediately tell you where Waldo is.

这确实挺厉害的。

It's so impressive.

你还会邀请朋友来亲眼看看他有多厉害。

You invite your friends over to see him in action.

但迟早会出一本新版的《威利在哪里》。

But eventually, there will be a new publication of Where's Waldo?

你的朋友带来了这本新书。

Your friends bring over this new book.

他们把书递给诺姆和阿蒂。

They hand it to Norm and Artie.

其实他们是递给诺姆，但可惜他完全懵了。

Well, they hand it to Norm, but sadly, he's lost and confused.

他对这些新页上的瓦尔多一无所知。

He has no idea where Waldo is on these new pages.

也许是因为他只是被编程记住了瓦尔多的位置。

Maybe it's because he was simply programmed on the coordinates of where Waldo is.

但那个程序员怎么可能知道他从未见过的页面上瓦尔多在哪呢？

But how is that programmer going to know where Waldo is on these pages he's never seen before?

接下来他们把书递给阿蒂，阿蒂却依然能认出这些他从没见过的页面上的瓦尔多。

The next they hand it to Artie and Artie can still identify Waldo on these new pages that he's never seen before.

这怎么可能呢？

Well, how's that possible?

因为他实际上是通过学习瓦尔多的长相来识别的。

Because he was actually trained on what Waldo looks like.

那件红白相间的毛衣、眼镜和帽子。

The red and white sweater, the glasses, the hat.

人工智能的工作方式很像我们的朋友RT，它能学习并适应新情况。

Artificial intelligence works a lot like our friend, RT, where he can learn and adapt to new situations.

一旦训练完成，它就会记住所有学到的东西。

And once it's trained, you know, it retains everything it's learned.

它能飞快地识别图像，而且不像我们，它从不需要喝咖啡休息，也永远不会累。

Can read images lightning fast, and it never unlike us, it never needs a coffee break, never gets tired.

另外，在那次演讲中我还会提到，大概进行到一半的时候，我们会玩一个叫‘这是人工智能吗？’的游戏。

And then maybe the other thing I'll mention in that presentation is I also about halfway through, we play a game called Is it AI?

我会展示一些属于我们Smart XR产品系列的功能。

Where I share a number of Agfa features that are part of our Smart XR portfolio.

然后我会做个现场调查，解释完每个功能后，问大家：你觉得这是人工智能吗？

And I ask, can I take a survey, a live survey of, you know, I explain the feature, I say, do you think this is AI or not?

我觉得这其实挺让人惊讶的，因为很多你觉得‘哇，这真厉害’的功能，

And I think it's quite eye opening actually, because a lot of the features that you think, oh, wow, that's quite impressive.

确实很有用，

That's very helpful.

比如那些基于视觉的摄像头，能帮你对准位置什么的，大家就说‘这肯定是AI’，但其实不是。

There's vision based cameras that show you, help you position and so on, and they say, oh, that's AI, but it's not.

而有时候，一些你可能觉得挺普通的功能，比如自动旋转图片这个功能，

And then sometimes some of the more, what you might might consider mundane features, like the feature that automatically rotates the image.

到这时候大家反而说‘这肯定不是AI’，但其实它就是。

By that point, they're like, oh, that's not AI, but it is.

因为你想，要是人工去编程实现这个，得给它喂上几百张甚至上千张图片，让它学会正确的朝向，毕竟每张X光片都不一样，

Because if you think about how you would have to program something like that, you have to feed it hundreds and hundreds of images so it can learn the correct orientation and orient the image correctly because no two x rays are the same.

像自动旋转这种功能，真正靠的还是人工智能。

You actually have to use artificial intelligence for some of those features like auto rotate.

说到底，我觉得最大的误解，其实就是‘什么是AI’这个问题。

I guess, again, to sum up a long answer, I think that one of the biggest misconceptions is just what is AI?

我觉得厂商在一定程度上会故意模糊这些界限，以利于自己。

I think vendors can blur those lines to their advantage to some degree.

公平地说，确实有些功能很难说清到底是规则自动化还是人工智能，因为它们两者都用到了。

And to be fair, there is some features where there is a true blurring of is it AI or not, where it obviously uses rule based automation and AI.

但确实，区分是不是AI很重要，最后我想说一句：是不是AI，这真的很重要。

But yeah, I think it can matter whether it's A or not, and this is the last thing I'll say, it can matter whether it's AI or not.

一旦这种AI被部署到放射科，它就固定下来了，对吧？

Once this AI is released into a radiology department, it's locked, right?

它不会再从越来越多的X光片中继续学习。

It's not continuing to learn from more and more x rays.

这其实有几个重要原因。

And that's actually important for a couple of reasons.

第一，FDA出于安全考虑，要求系统必须保持固定不变。

One, the FDA requires that it be locked for safety purposes, right?

也就是说，你得用一个经过验证的算法，不能让它随时间随意改变。

So you have a validated algorithm that you don't want to change over time.

第二个原因对技术人员来说也很重要，那就是在实际操作中，我们很难分清哪些是熟悉的规则型自动化，哪些才是真正的AI。

The second reason it's important, I think important to technologists, is operationally, it can be quite difficult to discern between what is rule based automation that we're used to versus what is true AI.

因为一旦系统上线投入使用，无论是传统的规则型自动化，还是真正的机器学习AI，它们都接收输入、运行内部逻辑，然后给出一致的输出，对吧？

Because you consider once it's released and you're using it, both a normal computing rule based automation and true machine learning AI, they both accept inputs, apply their internal logic and produce a consistent output, right?

规则型自动化和真正AI之间的核心区别，在于它们能处理和解决的问题类型不同。

The real difference between rule based automation and true AI is the type of problems that they can handle and help solve.

我特别喜欢那个‘寻找瓦尔多’的例子。

I really love the Where's Waldo?

他们用了一个视觉类比，我觉得大多数人一看就懂。

They gave a visual that I think most people can compare.

说到Agfa的放射学解决方案，我认为你们在推动AI应用于技术人员岗位方面是走在前列的。

When it comes to Agfa Radiology Solutions, you guys are, in my opinion, pioneering AI in technologists roles.

为了澄清这种混淆，我觉得现在谁都能随便贴个‘AI’标签，搞得我们真的分不清这到底是AI，还是只是自动化。

To clear up the confusion, I feel like that is anybody can kind of brand stuff, and we do get kind of confused, like, is this AI or is it just automation?

那怎么才能简单明了地判断呢？作为一线技术人员，我该怎么分辨？我知道AAC是这么回事，这我明白。

How can text tell if something is, in the most simplest way, like as a tech on the floor, how can I tell if, because AAC, I mean, I get that, that makes sense?

但如今，每个技术人员，尤其是年轻一代的Z世代和千禧一代，可能根本没经历过AAC时代，也记不清它到底是什么，更分不清它和别的技术有什么区别。

But what's something that nowadays every technologist, especially the new generation, Gen Z, even the millennials, maybe they didn't go back far and kind of can remember the AAC or really doesn't, they can't tell the difference between that and something else.

他们没有从传统技术背景过来。

They didn't come from a field background.

他们接触的全是数字化的东西。

So all they've known is digital.

有时候人们会搞混，比如新入行的Z世代技术人员，该怎么判断什么是真正的AI，才不会感到害怕或困惑呢？

And sometimes people get that confused with, Oh, that's How can a technologist, someone, Gen Zs entering the profession, understand when something is true AI so they're not scared, so they're not confused?

我觉得我之前想表达的是，这其实不容易，因为一旦AI功能上线了，不管是基于规则的系统还是真正的AI，它们都不会再持续学习、适应或变化。

Well, I think as I was kind of trying to allude to, it's not easy because they both, again, the AI, once it's in the system Once you have a feature that's released, it's not continuing to learn and adapt and change.

不管是规则驱动的还是AI驱动的，这些功能都很有用，对吧？

Whether it's rule based or AI, they're very useful features, right, regardless.

如果你真想弄清楚某个功能是不是AI，可以问供应商一些问题，比如它是怎么编程的，又是怎么训练的。

If you wanted to truly know whether something was AI or not, I mean, there's some questions you could ask the vendor in terms of how it was programmed, how it was trained.

比如说，它是用真实的临床数据来训练的吗？

Like, was it trained using actual clinical data, for example?

它是用真实的临床数据验证过的吗？

And was it validated on actual clinical data?

它真的能靠这些临床数据自主应对新情况吗？

Did it actually learn to kind of handle new situations on its own with this clinical data?

这倒是一个办法。

That could be one way.

你当然可以仔细想想这些问题：它到底有多可能被编程实现，它想解决什么样的问题。

Certainly you can critically think about these things, how likely could even be programmed, what kind of problem it's trying to help solve.

我知道不是每个人都有条件这么做，但如果你能跟供应商聊一聊，问出这些关键问题，确实有助于分辨到底是规则驱动的自动化，还是真正的AI。

I know not everyone has access to be able to do this, but you know, if you can talk to vendors and ask them these, these pointed questions that does kind of help parse apart whether it is rule based automation or true AI.

有时候，要可靠地判断还真不容易。

Sometimes it can be quite hard reliably know.

现在我有机会在RSMA上亲自接触并操作Agfa的设备，亲眼看看它的AI和自动化功能。

Now I've got a chance at RSMA to actually touch and feel Agfa's equipment and see the AI and the automation for myself.

我简直惊呆了。

I was blown away.

你们可以具体说明一下，当涉及到你们使用AI时，是如何避免这种混淆的。

You can talk specifically how you guys act for clear of that confusion when it comes to your use of AI.

是的。

Yeah.

我们其实没花太多精力去区分哪些是AI、哪些不是。

Don't know that we spend a lot of time really trying to discern this is AI, this isn't.

我觉得对终端用户和工程师来说，重点是这个功能能帮到你。

I think the goal for the end user and the technologists is this is a feature that can help you.

它是怎么运作的。

This is how it works.

当然，如果有人想深入探究细节，我非常乐意和他们讨论。

Now, certainly if they want to ask questions to dig into the weeds, I love having those conversations.

你懂我的意思吧。

You know what I mean?

但除此之外，我们确实在某些方面使用了AI，比如图像处理、降噪、虚拟网格、图像分割、自动旋转，以及辅助定位。

But beyond that, certainly we are using AI here in certain ways, in image processing, with the noise reduction, in the virtual grid, the segmentation, with auto rotating images, with helping positioning.

问题是，我们该怎么告诉用户他们是否在使用AI呢？

The question is, how do we tell them whether they're using AI or not?

我认为随着AI功能不断推广，厂商其实有责任对这些AI功能保持适度透明，现在可能还不明显，但未来一定会越来越重要。

I think what's also can be important as AI continues to be rolled out though, is vendors I think do have responsibility for some of these AI features where I think not so much today, but I think it's coming without overly disclosing.

用户需要清楚地知道，这个AI到底能做什么？

There has to be a transparency between, you know, what does this AI do?

当然，除了优点，也要说明它的局限在哪里。

Obviously, what are the strengths, but also what are the limitations?

当我们讨论该不该信任AI时，理解AI的本质很重要，这会影响你对这些功能的判断和思考方式。

When we talk about maybe when to trust or not trust AI, I think it can be important to understand what is AI because that can play an impact on how you critically think about some of these features when you're seen.

这个回答我很赞同。

I love that answer.

接下来我们要聊一聊全自动X光，因为大家看科幻片、刷新闻时都会想到这个。

Now we're gonna get to the autonomous x-ray part because we all watch sci fi and we we see in the news.

当技术人员听到‘全自动X光’时，脑子里浮现的画面就是：房间里一个技术人员都没有，机器自己全搞定。

When when technologists hear autonomous x-ray, they picture a room with no technologists, just the machine doing everything.

这真的是它的意思吗？

Is that actually what it means?

人们需要这么紧张吗？

Should people be freaking out?

你可能这么问，是因为你现在在RSNA现场，而我们爱克发在RSNA发布了我们对全自动X光的愿景。

You might be asking because again, you're at RSNA and at RSNA, we, Agfa, announced our vision for autonomous x-ray.

不过你说得对，我特别想强调一点：听起来比实际情况吓人多了。

To your point though, something I definitely wanna stress is sounds scarier than it is.

现在的人，包括我自己，一听到‘AI全自动’这种词，就会想：天啊。

I think when people these days, myself included, hear AI autonomous, it's like, boy.

这对我意味着什么？

What does that mean for me?

这对我未来有什么影响？

What does that mean for my future?

这对我孩子又意味着什么？

What does that mean for my kids?

我们说的自主X光，可不是说把人排除在外。

When we're talking about the autonomous x-ray, you know, we're not talking about autonomy from people.

而是说让人为自己减负。

We're talking about autonomy for people.

我之前提到过，现在医疗人力严重短缺。

And what I mean by that is, you know, I already touched on this workforce shortage.

到2030年，岗位空缺率可能达到百分之二十五到三十。

By 2030, there's gonna be 25 to 30% vacancy rates.

而且随着X光检查量不断增加，这情况只会更糟。

And that's only getting worse by the increase in x-ray volume.

还有个很多人没注意到的趋势，就是医疗服务的分散化。

And another area that people don't consider a lot of times is this decentralization of care.

诊疗正从大医院、大城市，逐步转向农村和门诊诊所。

Care is moving from hospitals, from cities, to rural areas and to outpatient clinics.

这么做就是为了把医疗服务送到患者身边，特别是贴近老龄化人群。

I mean, and the reason for that is to bring care closer to patients and closer to our aging population.

这么做是有充分理由的。

There's a good reason to do that.

但问题是，当你听到30%的空缺率时，并不意味着全国各地都是这个比例。

But the problem is when you hear 30 vacancy rates, that doesn't mean 30% vacancy rates across every part of the country.

在很多农村地区，实际空缺率可能高达50%。

When you're in a lot of these rural areas, that means 50% vacancy rates potentially.

我说的‘以人为本的自主性’，指的是在人员严重短缺的情况下，我们怎么保证患者应得的诊疗质量。

When I say autonomy for people, what I mean is how can we maintain the quality of care that patients need, deserve with this real staffing problem?

像我之前说的，我们现在已经在诊疗现场用AI帮技术人员减轻一些繁琐的工作，让他们能专注更重要的事。

Today, we are using AI, as I mentioned, at point of care to help technologists to remove some of these mundane tasks, to allow them to essentially to do more.

他们不用再手动旋转图像了。

They don't have to rotate images.

也不一定非得手动调整照射野。

They might not need to collimate.

由于AI和图像处理技术，他们也不用花那么多时间调窗宽窗位了。

They don't have to do as much window leveling because of the AI and the image processing.

但我觉得更重要的是，光这样还不够。

But what I think is also important to understand is that is not enough.

未来四年，甚至三年半的时间里，我们可能依然面临30%的人员空缺率。

We're going to have four years, three and a half years, potentially we're going have 30% vacancy rates.

这远远不足以满足患者的就诊需求，更谈不上提供他们应得的高质量服务，也无法保证他们不失去接受放射检查的机会。

That is not enough to meet the patient demand and give them the quality of care they deserve and to ensure that they're not losing access to radiographic exams.

这确实是个大问题。

That's a real problem.

光靠这些零零散散的AI功能，还是远远不够的。

And these AI features here and there alone are not enough.

但就我们讨论的全自动X光系统而言，至少阿加夫的设想是：这种设备专门处理特定任务，比如只做胸片——前后位和侧位胸片，对象是确认能自主行走、能沟通的患者，全程几乎不需要技术人员干预。

But for what we're talking about the autonomous x-ray, again, at least Agfa's vision is this would be a room that handles a very specific scope, a system that would really focus on just chest x rays, like PA and lateral chest x rays on patients that are proven to be ambulatory communicative, it'd be able to handle those types of patients with minimal technologists oversight.

明白吗？

Alrighty?

即便在这种情况下，打个比方，你可能会想到沃尔玛那种有人监督的自助结账，我知道很多人挺反感这个的。

And even there, just to draw a comparison, you might think of like the supervised self checkout at the Walmart, which I know people hate.

我明白。

I understand that.

但我们的目标是扩大技师的工作能力，以便满足患者的需求。

But the idea is to, again, extend the capacity of technologists so we can meet the patient demand.

0%的意思是，他们觉得完全可以取代技师的工作。

0% is they're looking to like, Hey, we could do the job of a technologist.

绝对不行。

Absolutely not.

我们该怎么实际地提供帮助，才能持续提供应有的高质量护理？

It's just, how can we realistically help so we can continue to provide the quality of care that there needs to be?

这项技术会怎么运作？

How would this technology work?

过程中会有安全核查机制吗？

Will there be safety checkpoints along the way?

当然，系统必须先确认：这位是正确的患者。

Of course, the system would have to confirm, Oh, this is the correct patient.

我已经确认过，他们具备完成胸部X光检查所需的体能和认知能力。

I have confirmed that they have the mental and physical capacity to kind of position themselves for a chest x-ray.

房间能自动调整患者位置，当然，现在有些功能我们其实已经做到了。

The room would be able to automatically position, which of course, some of these things obviously we already today, right?

系统还得自动调节准直器，识别患者体型，自动调整曝光参数。

It would need to automatically collimate, know, look at the thickness of the patient, adjust the technique.

我们现有的AEC系统已经能做到这一点了。

Of course, we have AEC.

另外，自动质量控制技术现在也已经在应用和发展中。

And something that is also here and coming is potentially due to automatic quality assurance, right?

但这样一来，技术人员就能同时管理两个房间，或者一边照护患者一边进行成像。

But all the while, it might mean the tech can manage two rooms at once or have a patient, they're imaging.

另一个系统会发出提示，说患者已经准备好了。

And then this other system would say, would have give a some type of notification to say, hey, the patient is ready.

你进去检查一下图像，确认有没有问题，一切正常就行。

You go in, you check the images, do you have any questions and everything looks good.

他们可以继续自己的事情了。

They can go on about their way.

但除此之外，真正重要的是要明白，在所有影像检查方式中——X光、MRI、CT等等——最难自动化的，始终是X光的体位摆放，这一点我们都知道：一方面不容易，另一方面对诊断结果至关重要。

But you know, beyond that, I think what's really important to understand is, you know, of all the imaging modalities, x-ray, MRI, CT, and so on, one of the hardest things to automate is always going to be x-ray positioning, which we know is one) not easy and two) critical for a diagnostic study.

所以，在可预见的未来，AI根本不可能为急诊室里无法外展手臂的腋肩患者摆好体位。

And so there's just no feasible near future where AI is going to be able to position a patient in the ED for a axillary shoulder who can't abduct their arm.

也不可能在ICU里，把便携设备推到一个毫无反应的病人床边，自己完成操作。

There's no realistic future where it's going to take a portable and into the ICU and actually non responsive patient.

也就是说，我们讨论的其实是非常特定的一类患者，才有可能实现这种未来需求。

Mean, so, you know, we're talking about a very, again, a very specific scope of patient to try to realistically meet this future demand.

在这个愿景里，技术员的作用依然是不可或缺的。

And the technologist remains absolutely critical, you know, within that vision.

而且，这反而能让技术员更充分地发挥他们的专业能力。

And really, it even allows them to practice at the top of their license.

就像我刚才说的例子，他们根本不需要去点那个旋转按钮，这样他们就能把精力集中在只有人类才能做好的事情上——比如陪伴患者、表达同理心、解答疑问，还有完成那些更复杂、更困难的体位和检查。

Again, you know, even the example I gave, they don't need to be clicking the rotate button, you know, it allows them to focus on areas where only humans excel, which is being present with patients, showing them empathy, making sure their questions are answered and doing some of these more, know, the more difficult positions and type of exams.

我们其实可以做很多别的事，这些事能让我们的工作更轻松，也能让医疗质量提升不少。

There are so many other things that we could be doing, you know, that would make our job and the quality of care much better.

你真的说出了我们心里的顾虑，也让我们安心了，因为我们终于看清了全自动X光到底是什么样子，也看到了安科如何通过AI，确保我们放射技师始终站在患者护理的核心位置。

And I feel like you really gave us, you aligned our fears, you relaxed those because we got to really see what autonomous x-ray really looks like and what Agfa is doing, Agfa Radiology Solutions is doing to make sure we as technologists, we stay central to our patient care by using AI.

我特别喜欢这一点。

I really love that.

非常感谢你澄清了这些，打消了我们很多疑虑和困惑，毕竟现在信息太多，变化也太快了。

Thank you so much for clearing that up and allaying our fears and confusion that is out there because there's just so much information and things are coming fast.

这又回到了安科始终坚持以人为本的理念。

It goes back to Agfa's keeping humans central.

你提到一点特别好，就是你们的AI和全自动X光系统，根本不是为了取代人类。

You made a very good point in making sure that the AI that you use, the autonomous x-ray, is not there to remove humans.

我们其实有很多地方是不可或缺的。

There are so many things that we are needed there for.

把那些琐碎重复的活儿交给系统，让我们能把精力用在更该专注的地方。

So to take away some of those mundane tasks that we don't need to be doing that could be focused somewhere else.

那我们来聊聊患者护理吧。

So let's talk about patient care.

你既然提到了，那我们就深入说说。

Since you touched on it, we're going talk about it.

每个厂商的演示材料里，我都能看到讲流程效率、吞吐量这些，确实很重要。

Because every vendor deck, you know, I've seen talks about the workflow efficiency, the throughput, and those things matter.

但人工智能和自动化到底怎么真正改善了诊室里患者的实际体验呢？

But how does AI and autonomy actually improve what happens to the patient in the room?

你刚才稍微提了一点，比如那些紧张的病人、儿童患者，还有第一次来影像科的病人。

You talked a little bit about it, but the scared patient, the pediatric patient, the patient who's never been in an imaging suite before.

这正是以人为本的体现。

This is keeping humans central.

我觉得我之前已经提到过一点，就是AI自动准直功能。

I think I already touched on a little bit is AI collimation.

已经有研究数据表明，它能把每次曝光的拍摄时间从45秒缩短到31秒，这意义不小。

There's been published studies that show that it can reduce acquisition time from forty five to thirty one seconds per exposure, which is certainly meaningful.

它还能把技术人员调整准直器的次数减少一半。

It also can cut the amount of times that a tech interacts with the collimation in half.

这样一来，他们就有更多时间专注于患者和真正重要的事情。

So it's again, saving them time to be able to focus on the patient and things that matter.

单台设备一年就能节省二十个小时。

It can save twenty hours over a year on a single modality.

这项研究发布时，我们只在远端四肢上用了这个功能。

And that study was actually published when we only had it for distal extremities.

现在我们已经应用到每一次曝光了，对吧？

Today we have it for every single exposure, right?

那节省的时间还会更多。

That number would be even more.

这又是另一个例子，比如技术人员去做床边X光检查时。

So again, that's just another example of technologists goes to do a portable x-ray exam.

他们再也不用操心探测器的朝向了。

They don't have to worry about the orientation of the detector anymore.

他们只需要把设备放在身后，至少能多出几秒钟专心照顾病人。

They can just put it behind them and it just gives them a few more seconds at least to focus on the patient.

我认为这正是Agfa愿景的核心：让技术人员能把精力集中在真正重要的事上，而不是被琐事牵绊。

I think that's a big part of Agfa's vision is it's trying to let them focus on, again, where humans can excel in what matters and not on the mundane things.

我还想提一点，职业倦怠确实是真实存在的问题。

Another point I'd make too is burnout's a real thing.

我每天都会和技师们交流。

I mean, you you talk to texts every day.

我经常和很多技师聊天。

I talk to a lot of texts.

说实话，职业倦怠是个严重的问题。

Mean, burnout is is a real problem.

如果每天要拍的X光片特别多的话，情况就更明显了。

And if there's a lot of x rays to take, essentially.

现在，我举几个例子，说明哪些AI技术已经投入使用，哪些还在发展中。

Now, maybe to give a couple of examples of AI that is here today, but maybe more on the horizon.