Tuberculosis Screening Centre Introduces AI "Reading Assistant" to Enhance Accuracy and Efficiency of Image Analysis


Share this article

Tuberculosis Screening Centre Introduces AI "Reading Assistant" to Enhance Accuracy and Efficiency of Image Analysis

Tuberculosis is a common infectious disease, but unlike COVID-19 and influenza, it lacks rapid diagnostic test kits. Doctors often must rely on subjective experience to search for clues in chest X-rays.

Dr Tay Jun Yang (37), Head of the National Centre for Infectious Diseases (NCID) National Tuberculosis Screening Centre , has been interpreting an average of more than 20 chest X-rays daily over the past five years. He typically spends less than a minute examining each image, becoming alert when white shadows appear in the upper portions of the lungs.

"White shadows also appear in chest X-rays of pneumonia and lung cancer patients, and some white areas are the patient's bronchial branches. We often have to rely on experience to decide whether patients need further examination."

To improve the accuracy and efficiency of image reading, the Tuberculosis Screening Centre has enlisted a tuberculosis reading "assistant" for Dr Tay Jun Yang and others – 12 medical staff in total – since September last year, providing them with an additional pair of "eyes" when interpreting chest X-rays.

This assistant works very "diligently," able to circle suspicious areas in chest X-rays within one to two minutes before radiology reports are available, then providing preliminary analysis for attending physicians' reference, with an accuracy rate exceeding 90%.

Dr Tay Jun Yang told reporters that the "assistant" is a specially trained artificial intelligence (AI) image reading software. Since the beginning of last year, it was provided by Korean medical AI company Lunit, through the coordination of Synapxe – the national healthtech agency under the Ministry of Health (MOH). This software is also part of the MOH's IT platform AimSG.

"The new software has the 'experience' of analysing 100,000 chest X-rays from around the world. After introducing it locally, we validated it using 500 local chest X-rays... It has made us more confident in our work."

Over the past six months, the new software has assisted in analysing nearly 3,000 chest X-rays. The centre has also provided software usage training for medical staff, with beneficiaries including not only doctors but also nurses responsible for triage assessment.

Cheng Hong, Advanced Practice Nurse (APN) at NCID, has worked at the centre for over four years, primarily responsible for providing preliminary explanations of conditions to patients before doctors' diagnostic results are available.

She said: "The doctor's diagnosis is paramount, but if the software's preliminary analysis results are obvious and consistent with our judgement, we can reduce patients' anxious waiting time, allowing them to go home with peace of mind earlier or prepare for further examination."

Local scientists identified the cause of secondary infections in tuberculosis patients

Although tuberculosis is curable, patients may still suffer lung infections after treatment. Local clinical research shows that approximately one-third of non-tuberculous infected individuals have a history of tuberculosis.

The Agency for Science, Technology and Research (A*STAR) recently discovered that this correlation is due to the formation of special structures (tuberculous granulomas) in the lungs of tuberculosis patients even after recovery, which provide a "protective umbrella" for bacteria, allowing them to evade immune system clearance and antibiotic treatment.

In an interview with Lianhe Zaobao, Dr Stefan Oehlers, Principal Investigator at A*STAR's Bacterial Pathogenesis Lab , and researcher Huang Shiyong , who were in-charge of this project, highlighted that this study took two and a half years and was funded by MOH. It aims to provide prevention and early intervention for tuberculosis survivors to reduce secondary infections after they complete months of antibiotic treatment for tuberculosis.

Dr Oehlers said: "We are using this discovery to research prevention and early intervention for tuberculosis survivors against reinfection, which also benefits the overall health of an ageing society."

The research paper was published in the British scientific journal Nature Communications in November last year.

Dr Caroline Choong, Clinical Director of the National Tuberculosis Care Centre (NTBCC) and Consultant in Respiratory Medicine at Tan Tock Seng Hospital (TTSH) , pointed out that this research once again emphasises that when considering tuberculosis, one cannot stop at merely "completing the treatment course."

She said: "Completing tuberculosis treatment does not always mean the lungs have returned to normal. Many patients still have scars and damage remaining in their lungs. This research highlights the need for careful follow-up and monitoring after tuberculosis treatment, especially for those whose lungs remain susceptible to infection after having tuberculosis."

 

结核病筛查中心引入AI“读片助手” 提升阅片精准度和效率

结核病是一类常见传染病,但在诊断时却没有像冠病和流感那样的快速检测试剂,医生常须凭借主观经验,从X光胸片里寻找蛛丝马迹。

国家传染病中心结核病筛查中心主任郑竣扬医生(37岁),过去五年来平均每天要判读20多张X光胸片,停留在每张片子上的目光往往不到一分钟,若发现肺部上方出现白色阴影,就会引起他的警觉。

“肺炎和肺癌的胸片里也会出现白色阴影,而且有些白色部分是病患的气管分支,我们时常要靠经验决定是否需要让病患做进一步检查。”

为了提高阅片的精准度和效率,结核病筛查中心从去年9月起,为郑竣扬等12名医护人 员请来了一位结核病读片“助手”,为他们在判读胸片时增加一双“眼睛”。

这个助手工作很“勤快”,能在放射科的胸片报告出来前,用一两分钟时间先将胸片中的可疑部分圈出来,再将初步分析提供给主治医生做参考,准确率超过九成。

郑竣扬告诉记者,“助手”是一个经过专门训练的人工智能(AI)读片软件,从去年初开始,经保健卫生部属下的全国医疗科技机构新联科技(Synapxe)协调,由韩国医疗人工智能公司Lunit提供。这个软件也是健卫部信息科技平台AimSG的一部分。

“新软件拥有分析10万张世界各地胸片的‘经验’,在引入本地后,我们又用500张本地胸片对它进行验证……它让我们工作时更自信了。”

过去半年,新软件已协助分析了将近3000张X光胸片,中心还为医护人员提供软件使用培训,受益者中除了医生,还有负责分诊评估的护士。

国家传染病中心高级执行护士程红,在中心工作了四年多,主要负责在医生的诊断结果出来前,向病患初步解释病情。

她说:“医生的诊断是首要的,但如果软件的初步分析结果显而易见并和我们的判断相一致,我们就可以缩短病患焦急等待的时间,提前让他们安心回家或准备接受进一步检查。”

本地科学家找到结核病患继发感染原因

结核病尽管可治愈,但病患在接受治疗之后,还可能再遭受肺部感染。本地临床研究显示,大约三分之一的非结核病菌感染者有既往结核病史。

新加坡科技研究局最近研究发现,这种关联性是由于结核病患者即使康复后,他们的肺部会形成一种特殊结构(tuberculous granulomas),能为细菌提供“保护伞”,让它们得以躲避免疫系统的清除以及抗生素治疗。

负责这一项目的新科研传染病研究所科学主管厄勒斯(Stefan Oehlers)博士和研究员黄诗咏在接受《联合早报》采访时指出,这项研究耗时两年半,得到卫生部经费支持,旨在为结核病康复者进行预防和早期干预,减少他们在接受了数月的结核病抗生素疗程后,患继发感染。

厄勒斯说:“我们正在利用这项发现,研究预防和早期干预结核病康复者再受感染,这也有利于老龄化社会的整体健康。”

这项研究的相关论文,已于去年11月在英国科学期刊《自然通讯》(Nature Communications)上发表。

国家结核病医疗中心临床主任,陈笃生医院呼吸内科顾问医生钟倩慧指出,新科研的这项研究再次强调,在看待结核病时,不能只停留在“完成疗程”这一层面。

她说:“完成结核治疗并不总是意味着肺部已经恢复正常,许多患者的肺里仍然残留疤痕和损伤。这项研究突出了在结核治疗结束后进行仔细随访和监测的必要性,尤其是针对那些在得了结核病后肺部仍易被感染的人群。”

Source: Lianhe Zaobao © Singapore Press Holdings Limited | Reproduced with permission.


Similar article:

 

Related articles

X

By continuing to use and navigate this website, you consent to the use of cookies in accordance with our Privacy Policy.

Confirm