研究發(fā)現(xiàn)，谷歌人工智能比醫(yī)生更能發(fā)現(xiàn)乳腺癌

研究發(fā)現(xiàn)，谷歌人工智能比醫(yī)生更能發(fā)現(xiàn)乳腺癌

Google’s artificial intelligence detected breast cancer better than doctors, a study published Wednesday found.

周三發(fā)表的一項(xiàng)研究發(fā)現(xiàn)，谷歌的人工智能檢測(cè)乳腺癌的效果好于醫(yī)生。

The Google AI program that examined mammogram images from thousands of women in the US and the United Kingdom spotted cancer more accurately than six expert radiologists who reviewed 500 random cases, according to the research — which was funded by the tech giant.

這項(xiàng)由這家科技巨頭資助的研究顯示，谷歌人工智能項(xiàng)目檢查了來(lái)自美國(guó)和英國(guó)的數(shù)千名女性的乳房x光照片，比6名隨機(jī)檢查了500個(gè)病例的放射科專家更準(zhǔn)確地發(fā)現(xiàn)了癌癥。

The software’s review of nearly 29,000 mammogram scans reduced the number of false negatives — tests that are wrongly deemed normal even though cancer is present — by 9.4 percent in the US and 2.7 percent in the UK, according to the study, published in the journal Nature.

根據(jù)發(fā)表在《自然》(Nature)雜志上的這項(xiàng)研究，該軟件對(duì)近2.9萬(wàn)例乳房x光片掃描結(jié)果的分析，減少了美國(guó)9.4%和英國(guó)2.7%的假陰性(盡管存在癌癥，但仍被錯(cuò)誤地視為正常的檢查)。

And the number of false positives — tests that are wrongly said to show signs of cancer — fell by 5.7 percent in the US and 1.2 percent in the UK, the research found. The team behind the study included researchers from Google Health, Northwestern Medicine and Imperial College London.

研究發(fā)現(xiàn)，在美國(guó)，誤報(bào)(被錯(cuò)誤地說(shuō)成顯示癌癥跡象的檢測(cè))的數(shù)量下降了5.7%，在英國(guó)下降了1.2%。這項(xiàng)研究背后的團(tuán)隊(duì)包括來(lái)自谷歌健康、西北醫(yī)學(xué)和倫敦帝國(guó)理工學(xué)院的研究人員。

Google Health said the study “sets the stage for future applications where the model could potentially support radiologists performing breast cancer screenings” — though more research and regulatory approval are needed to show how the technology could boost patient care.

谷歌Health表示，這項(xiàng)研究“為未來(lái)的應(yīng)用奠定了基礎(chǔ)，該模型可能會(huì)支持放射科醫(yī)生進(jìn)行乳腺癌篩查”——盡管還需要更多的研究和監(jiān)管部門的批準(zhǔn)，才能證明這項(xiàng)技術(shù)如何能夠促進(jìn)患者護(hù)理。

The research suggests that AI could help improve the accuracy of mammograms, the most common method of screening for the second-leading cause of cancer deaths among women.

這項(xiàng)研究表明，人工智能可能有助于提高乳房x光檢查的準(zhǔn)確性，這是篩查導(dǎo)致女性癌癥死亡的第二大原因的最常見(jiàn)方法。

Mammograms miss roughly one in five breast cancers, according to the American Cancer Society, and inaccurate test results can cause problems for patients. A false positive can lead to anxiety or pricey additional testing, while false negatives can cause women to think they’re cancer-free when they’re not.

根據(jù)美國(guó)癌癥協(xié)會(huì)(American Cancer Society)的數(shù)據(jù)，乳房x光檢查遺漏了大約五分之一的乳腺癌，不準(zhǔn)確的檢測(cè)結(jié)果可能會(huì)給患者帶來(lái)麻煩。假陽(yáng)性可能會(huì)導(dǎo)致焦慮或昂貴的額外檢查，而假陰性可能會(huì)讓女性以為自己沒(méi)有癌癥，其實(shí)不然。

Researchers trained Google’s AI program with mammograms from more than 76,000 women in the UK and more than 15,000 in the US, according to Google Health. Google said the software outperformed doctors despite receiving less information than human experts, who also had access to previous mammograms and patient histories.

根據(jù)谷歌Health的數(shù)據(jù)，研究人員對(duì)谷歌的人工智能程序進(jìn)行了培訓(xùn)，使用了英國(guó)逾7.6萬(wàn)名女性和美國(guó)逾1.5萬(wàn)名女性的乳房x線照片。谷歌表示，盡管獲得的信息比人類專家少，但該軟件的表現(xiàn)優(yōu)于醫(yī)生。

The study is Google’s latest bid to use its technologies for health purposes — even though some of those efforts have sparked privacy worries.

這項(xiàng)研究是谷歌為了健康目的使用其技術(shù)的最新嘗試——盡管其中一些努力已經(jīng)引發(fā)了對(duì)隱私的擔(dān)憂。