立即打开
未来医生:超级计算机

未来医生:超级计算机

Michal Lev-Ram 2012-12-07
科技领域的两大热门趋势,也就是数据分析和移动设备正在走进医院。未来,大型计算机系统可以处理包括医学文献到病人体征在内的数据,据此判别症状。然后,医生就可以借助平板电脑或者智能手机上的智能软件掌握病情,更及时地提供相应的治疗。

    医学电视剧的粉丝也许知道败血性休克有多么可怕。不过几乎没人知道,败血症这种导致身体抵抗感染时攻击自身的疾病每年导致多达258,000位美国人死亡。尽管这种疾病易于治疗,却难以诊断。病人们起初可能没什么症状,但却会在短短几个小时内休克。

    数据分析和移动设备这两项高速发展的科技能够帮助诊断包括败血症和癌症在内的多种困扰人么的疾病。大型计算机系统正在持续不断地处理数据——从医学文献到病人的体征一应俱全,并以此判别症状。因此,通过平板电脑或者智能手机上的智能软件,医生就能对即将发生的问题有所警觉。这样,救人于旦夕之间的治疗将会更加及时。

    荷兰信息服务公司威科集团(Wolters Kluwer)下属的医疗公司正在测验能够确诊和治疗败血症的技术。公司最近开始寻找试点医院来参与其“败血症死亡率降低计划”。哥伦布市河畔卫理公会医院(Riverside Methodist Hospital)负责治疗质量和病人安全的医疗总监詹姆斯·奥布莱恩说:“如果这个计划取得成功,败血症的死亡率将降低一半。”

    另一个例子发生在去年,IBM公司的超级计算机Watson在娱乐节目《危险边缘》(Jeopardy!)的智力竞赛中击败人类,登上了报纸头条。这台电冰箱大小的机器正在为其第二职业进行相关准备,这份职业也是纽约纪念斯隆-凯特琳癌症中心(Memorial Sloan-Kettering Cancer Center)试验计划的一部分。Watson借助并行处理技术——即同时处理多项任务的技术——每秒钟最多可以处理500GB的数据。例如,内科医师可以输入切片检查的数据,而这台计算机则会据此列出病人的病史、相关的临床研究和医学期刊。接下来还会给出可能的治疗方法和相应的“信心水平”,即治疗成功率。而最终的决定权依然在医生手中。

    这样的系统究竟有多大价值,目前还无法确定。不过根据市场研究公司Insight Research的数据,接下来六年内,美国卫生保健行业将会在信息产业投入690亿美元。英特尔(Intel)和思爱普(SAP)已经开始与加州大学伯克利分校(University of California at Berkeley)的研究人员通力协作,开发具有竞争力的医用超级计算机。

    不过在这个项目最终证实能提高效率、减少开支之前,医院并不会买账。加州大学旧金山分校(University of California at San Francisco)的校长兼肿瘤医生苏珊·戴斯蒙德-海尔曼解释说:“临床医师只相信证据。”

    还有一个问题在于:全天下的医学数据和处理能力都无法教会计算机如何像医生那样对待病人。更不用说一旦用算法取代医生,电视剧将变得多么无趣。

    译者:严匡正

    Fans of television medical dramas are probably aware of the grim condition known as septic shock. But few people know that sepsis, a disease that causes the body to attack itself in an attempt to fight off infection, kills 258,000 Americans each year. Though easy to treat, sepsis is difficult to diagnose; hospital patients can go from asymptomatic to a state of shock in just a few hours.

    Two fast-growing technologies, data analytics and mobile devices, could help solve vexing problems ranging from sepsis to cancer diagnosis. Large computer systems are increasingly crunching data -- everything from medical journals to patients' vitals -- to recognize patterns. That has allowed intelligent software to alert doctors of impending problems via tablet or smartphone, making lifesaving treatment more timely.

    Wolters Kluwer Health, a division of the huge Dutch information-services firm, is currently testing such technology to identify and treat sepsis. The company recently began enrolling hospitals as pilot sites in its "sepsis mortality-reduction program." James O'Brien, medical director of quality and patient safety at Riverside Methodist Hospital in Columbus, says, "If this is successful, it could cut the rate of death from sepsis in half."

    Watson, the IBM (IBM) supercomputer that made headlines by drubbing humans on Jeopardy!last year, is another example. The refrigerator-size machine is prepping for a second career as part of a trial program at New York City's Memorial Sloan-Kettering Cancer Center. Watson relies on parallel processing -- geekspeak for running multiple tasks at once -- to sift through 500 gigabytes of data per second. A physician can enter the results of a biopsy, for example, and Watson pulls relevant bits of a patient's history as well as clinical studies and medical journals. It then lists potential diagnoses and their varying "levels of confidence," or probability. The final call is left up to the doctor.

    It isn't clear how much the market for such systems will eventually be worth. But the U.S. health care industry will spend some $69 billion on IT over the next six years, according to Insight Research Corp. Intel (INTC) and SAP (SAP) are already working with researchers at the University of California at Berkeley to develop a competing breed of medical supercomputer.

    But hospitals won't buy into the idea until it has been proven to boost efficiency and trim costs. "What clinicians respond to is evidence," explains Susan Desmond-Hellmann, an oncologist and the chancellor of the University of California at San Francisco.

    Another problem: All the medical data and processing power in the world can't teach a computer bedside manner. Not to mention that substituting algorithms for doctors would make for really dull TV.

热读文章
热门视频
扫描二维码下载财富APP