Can AI Spot Lung Cancer? Full Story
This London Hospital Is Putting it to The Test
Chest X-ray: Qure AI
People sometimes shy away from the topic of AI because it can be quite “doom and gloom”. Here at Subscribe to AI, we’re aware of AI's danger and the stories you likely see in the headlines. But we both remain deeply committed to finding and sharing human stories that impact everyday people.
There are scientists, doctors, artists and creative people doing wonderful things with the tools, and we’re here to send you the good, the bad, and the bizarre AI stories.
Today, we’re looking at new research that was announced this week and how it might change the way lung cancer gets detected — saving lives.
In England, about 2,000,000 chest X-rays are requested every year. Chest radiography is the most common type of imaging test, making up roughly 20% of all types of these procedures.
Thankfully, most results return normal, but a small percentage of the patients receive possible life-changing news — early indications of lung cancer.
When this happens, patients are referred for a CT scan of their chest to help confirm or rule out the lung cancer diagnosis.
While the national health system in the UK remains something cherished, its wait times are long, with limited staff numbers and a deeply underfunded system. For many years, doctors and patients have sought ways to speed up the process of lung cancer diagnosis. Early detection is crucial.
This week, a new study by University College London Hospitals was announced. The goal is to reduce the time it takes to diagnose lung cancer, and the researchers hope to do so with artificial intelligence software.
The trial will test out an extraordinary tech called qXR (not the friendliest of sounding names) designed to spot abnormalities on chest X-rays.
It was developed by a company called Qure.ai. The software flags scans that would benefit from an urgent scan review by a reporting radiographer, who can then arrange a same-day CT scan if their X-ray indicates possible lung cancer.
So, at this early stage, the AI software goes, “Hey, this looks a little alarming. Can a human radiographer take a closer look at this one ASAP?”
This makes sense because, at this stage, we still want the humans to make big decisions. All the AI tech does is help prioritise the scans and bring the more urgent-looking ones to the top of the queue.
Just read what one of the doctors co-leading the study had to say:
So, what does it even look like?
The AI system produces a secondary image of each X-ray with an overlay to highlight possible abnormalities. If there’s a problem, the scan will also be highlighted on the reporting radiographer's worklist.
And the stats are pretty impressive.
The tech is already in use in other parts of the world and has been studied by Harvard Medical School and Massachusetts General Hospital. They found in their research that 90% of missed or mislabeled chest x-rays they fed into the software were identified with a critical abnormality by the tech with zero false positives. The study concluded that AI could accurately detect mislabeled and missed findings.
For all the “doom and gloom” in the AI space, the bizarre deepfake celebrity content and even weirder AI adult content — the good AI news can always be found in the cancer world. More than ever, people are optimistic about speeding up cancer detection and perhaps even curing and treating types of cancer with AI.
If, in the year 2023, AI can help spot lung cancer early, with technology that didn’t even exist a decade ago, imagine what it might do in ten years.
💬 For the Group Chat!
Copy and paste it — we won’t tell anyone.