When every second counts: Using AI to assess brain injuries could save lives
Currently, doctors are swamped by the number of brain injuries and scans they have to see each day, but AI is seen to be a possible solution.
SINGAPORE: It was a Wednesday like any other. Jessie Goh was walking home via the overhead bridge and was stepping on to the pavement when she was hit by an electric scooter and thrown forward, injuring her head.
She was rushed to Changi General Hospital’s Accident and Emergency (A&E) department and had the first of several CT scans of her brain to ascertain her injuries.
Two hours later, her husband, who wanted only to be known as Mr Lee, was given the initial medical assessment and told that Jessie needed another scan done to further investigate the extent of her brain injuries. The 45-year-old was admitted to the high dependency ward that Wednesday night. She was found to have a blood clot at the front of her brain and a hairline fracture at the back of her skull.
A few days later on Saturday, Mr Lee found her unusually lethargic. “I feed her lunch, and she can eat until she falls asleep,” he recounted, but set aside his concerns that day.
Jessie’s condition had not improved when Mr Lee visited her the next day and this time Mr Lee requested that his wife be examined again. A CT scan was done at noon and the doctor gave the all clear. But another CT scan done in the evening had a very different outcome.
“An emergency operation was needed,” the husband recounted. “If not, she might have gone.”
TIME, MANPOWER CRUNCH
According to the National Neuroscience Institute (NNI), it sees more than 1,000 patients with brain injury each year. And Jessie's experience of needing several days and CT scans to accurately diagnose the severity of a traumatic brain injury (TBI) is not uncommon.
NNI medical director Associate Professor Ng Wai Hoe shared during a media briefing earlier this month that after a CT scan is taken, a radiologist specialising in brain scans, also known as a neuroradiologist, would analyse it to make a diagnosis.
Assoc Prof Ng shared that NNI is fortunate it has a team of neuroradiologists on call throughout the day, but this is not the case for all hospitals. Even so, the team would receive about 70 to 100 CT scans over a 24-hour period and, significantly, about 30 of these would come in between midnight and 4am.
“The response time may be slower (during this time period) … due to limited manpower, leading to a time crunch,” the NNI medical director shared.
Professor Tchoyoson Lim, a senior consultant at NNI specialising in neuroradiology, shared at the same briefing that the challenge is to find the “haystack patient” - the patient who needs urgent intervention from the symptoms shown on the scan - amid the pile doctors have to go through each day.
Prof Lim also shared that among his students, even the best trainees would flag “false alarms" in the scans they read because of the complexity of the brain.
DIGITAL BRAIN TO HELP HEAL HUMAN ONES
These challenges are also why NNI signed a three-year memorandum of understanding (MOU) with local medical technology start-up Iota Medtech, which specialises in artificial intelligence (AI) and robotics.
NNI will work with Iota Medtech to provide its database of brain scans to train the latter’s machine learning algorithms. The hope is for the start-up’s solution to be the first pair of “eyes” to look at brain scans as they come in, and alert doctors of the most pressing cases.
“With rising numbers in Singapore affected by neurological conditions, we must continue to improve and innovate care,” Assoc Prof Ng said. “A main concern for head injuries, currently the leading cause of disability and death for adults under 40 in Singapore, is time reduction in the acute treatment stage.
“Employing the use of AI technology can save critical time during diagnostics and transform care outcomes for patients,” the medical director added.
Dr Justin Ker, senior resident at NNI’s Department of Neurosurgery, who is leading the research into the use of AI, said that work done in the laboratory has proven to be “viable” and the next stage is to engage in clinical validation. This means having both the AI and human doctors analyse the same scans, and monitor the results of predictions from both camps, Dr Ker explained, adding the AI system’s predictions will only be used for research at this point.
CEO of Iota Medtech Benjamin Hong shared that the commercialisation phase is “still quite some time away”.
However, he is confident that this collaboration would bear fruit in terms of solutions to “improve on current methods through the use of cutting-edge AI”.
Such tie-ups would go a long way to address the local healthcare sector's relatively slow pace of embracing AI, particularly in the area of diagnostics.
According to Philips’ recently released Future Health Index 2019 report, Singapore’s healthcare professionals are using AI more to improve the accuracy and efficiency of administrative tasks such as staffing and patient scheduling (37 per cent) than for diagnosis (28 per cent) and flagging patient anomalies (26 per cent).
“By primarily using AI for administrative tasks, like scheduling appointments, Singapore’s healthcare professionals risk missing out on the enormous benefits it can bring to patient outcomes,” said Ms Caroline Clarke, CEO of Philips ASEAN Pacific, in the press release.
“Technology will never replace the ‘human touch’, but AI can save time and improve diagnosis accuracy thereby having a huge potential for saving peoples’ lives.”
For Jessie, who has since physically recovered from her brain injury after many months of rehabilitation, the use of AI to aid in the analysing of brain scans can’t come soon enough.
It is “really upsetting”, not just for the patient but for surrounding loved ones, not to have a clear picture of the medical conditions and challenges they have to face. So if the use of AI will cut down the waiting time to achieve such outcomes, she said she is all for it.
As for whether there is a fear that a machine is looking at her brain, Jessie was practical in her response: “When we go for CT or MRI scan, already use computer.
“Whatever works, as long as you don’t keep people waiting,” she reiterated.