先看片子的那个软件

我们科里最近换了新系统，算法会先自动扫描每一张片子，把可疑的地方标记出来，医生还没打开片子，标记就已经在那儿了。说出来你可能不信，我现在看每一张片子，脑子里第一件事就是先看系统标记了哪里。

Our department just switched to a new system. The algorithm automatically scans every scan first and flags the suspicious spots, so before the doctor even opens the scan, the flag is already there. You might not believe it, but these days the first thing I do with every scan is check where the system flagged.

等一下，你是说，那个软件比你先看片子？这听着有点吓人，它到底看得准不准？

Wait, are you saying that software looks at the scan before you do? That sounds a little unsettling. Is it actually accurate?

大部分时候挺准的。上个星期我碰到一个特别小的结节，说实话那天我值了一整晚的班，眼睛已经很累了，差点就漏过去了，是系统先标记出那个地方，我才注意到，一查，真的是个结节。

Most of the time it's pretty accurate. Last week I ran into a really tiny nodule. Honestly, I'd just worked a full overnight shift and my eyes were exhausted, I nearly missed it. The system flagged that spot first, and that's what made me notice it. I checked, and it really was a nodule.

那挺好的，算是帮了你一个大忙。不过我听着还是有点担心，你天天让算法先标记，自己会不会慢慢就不那么用心看了，眼睛也越来越不敏感？而且算法自己会不会误诊，这个准确率有人跟你说清楚过吗？

That's good, it really did you a big favor. But listening to this still worries me a little. If the algorithm flags things for you every single day, might you gradually stop looking as carefully yourself, and your eye grow less sharp over time? And can the algorithm itself misdiagnose, has anyone actually told you clearly how accurate it is?

你说的这个我也想过。不过每张片子系统标记完，最后还是我自己一张一张看完，才能写诊断报告，它不会替我下结论，只是提醒我别漏掉哪里。

I've thought about that too. But after the system flags a scan, I still go through every single one myself before I write the diagnosis report. It doesn't draw the conclusion for me, it just reminds me not to miss anything.

问题是，等大家都信任那个标记以后，那些系统没标记的地方，你还会像以前一样，一寸一寸仔细看吗？还是心里已经默认，没标记就是没问题？

The problem is, once everyone trusts that flag, will you still go over the spots the system didn't flag inch by inch, the way you used to? Or has some part of you already started assuming that no flag means no problem?

老实说，这正是我自己也在提醒自己的地方。不过它确实帮我省了不少时间，扫描一张片子的可疑区域，它比我快得多，我能把更多精力放在那些真正复杂、拿不准的病人身上。

Honestly, that's exactly the thing I keep reminding myself about. But it really has saved me a lot of time. Scanning a scan for suspicious areas, it's far faster at that than I am, and it lets me put more of my attention on the patients who are genuinely complicated and hard to call.

那我再问你一个更难的问题。要是有一天，那个算法很自信地说这个地方没问题，其实它误诊了，是个真正的结节，你没发现，病人也放心地回家了，这个责任算谁的？是你，还是写算法的公司？

Then let me ask you a harder question. Suppose one day the algorithm confidently says a spot is fine, but it's actually a misdiagnosis, it's a real nodule, and you don't catch it, and the patient goes home relieved. Whose responsibility is that? Yours, or the company that wrote the algorithm?

这个我认真想过。最后签字的人是我，报告上写的是我的名字，所以责任肯定是我的，软件从来不替我做决定，它只是一个提醒的工具。

I've thought hard about this one. I'm the one who signs at the end, my name is what's on the report, so the responsibility is definitely mine. The software never makes the decision for me, it's only a tool that reminds me.

可是你想想，要是全国的放射科都在用同一个算法，它要是有一个共同的盲点，是不是所有用它的医生，都会一起漏掉同一种结节？这就不是一个人偶尔看走眼了，这是整个系统一起误诊。

But think about it, if radiology departments across the country are all using the same algorithm, and it has one shared blind spot, wouldn't every doctor using it miss the same kind of nodule together? That's no longer one person occasionally missing something, that's the whole system misdiagnosing together.

你说得有道理，我确实没有想得那么远。不过到目前为止，它帮我发现的问题，比它让我漏掉的要多得多，这也是真的，我没办法因为担心以后，就不用现在能用的东西。

You have a point, I honestly hadn't thought that far ahead. But so far, the problems it has helped me catch are far more than the ones it has caused me to miss, and that's also true. I can't refuse to use something helpful now just because I'm worried about later.

我不是让你别用，我只是希望，等哪天真出了问题的时候，责任能说得清楚，而不是最后什么都算在你一个医生头上，公司自己什么事都没有。

I'm not telling you to stop using it. I just hope that when something really does go wrong someday, the responsibility can be sorted out clearly, instead of everything landing on you as one individual doctor while the company walks away with nothing.

这一点我同意，这不该只是我一个人的事。说实话，我现在已经离不开它了，看每一张片子的时候，都会想它会不会标记，这确实已经变成我看片子的一部分了。

I agree with that, this shouldn't just be one person's problem. Honestly, I already can't do without it now. Every time I look at a scan, part of me is wondering whether it will flag something, that really has become part of how I read every scan.

那我最后问你一句，要是哪天这个系统被关掉了，或者坏了，你自己一个人从头到尾看完一整天的片子，还有把握吗？

So let me ask you one last thing. If the system were shut down or broke one day, and you had to read a whole day's worth of scans by yourself from start to finish, would you still be confident you could?

我希望有把握，可是说实话，我也不是百分之百确定。也许我该找个没有系统的时候，自己练一练，别让这个本事真的慢慢退步了。

I'd like to think I would be, but honestly, I'm not a hundred percent sure either. Maybe I should find a time without the system and practice on my own, so this skill doesn't actually slowly fade away.

这个想法我很赞成。工具是好东西，可是本事还是得留在自己身上，不然哪天工具靠不住了，麻烦的还是你自己。

I really like that idea. A tool is a good thing, but the skill still needs to stay with you, otherwise the day the tool can't be trusted, it's you who's in trouble.

行，那就这么定了，下个星期我值班的时候，先自己看完再看系统的标记，看看两边差多少，这样我心里也更踏实一点。

Alright, it's settled then. Next week when I'm on shift, I'll look through everything myself first before checking the system's flags, and see how much the two differ. That way I'll feel a bit more solid about it too.

好主意，到时候告诉我结果，我很好奇，人一张一张看和算法扫描，到底差在哪里。

Good idea. Let me know the results when you have them. I'm curious where the difference lies between a human looking at a scan one by one and an algorithm scanning it.