Dr Jack Kreindler only has to look at the piles of toys around his home to see the effects of his dedication to healthcare – and the technology that assists it.
In 2011 his wife Marjorie was sent to see a specialist after the couple struggled to conceive.
When scans and blood tests said everything was normal, they were advised to undergo IVF, advice that would later reveal to have been a bad idea if they had taken it.
In a twist of fate, a specialist on a hunch decided to perform an exploratory operation revealing there was cancer throughout her womb.
Radical surgery looked to be their only option, but that would have ended their hopes of having a biological child of their own.
It prompted Kreindler, a medical technology entrepreneur, to enlist the help of mathematician friends to calculate the risk of the cancer spreading if the operation was not performed immediately and attempting an emergency round of egg harvesting to give them a chance of having a family later.
“I’d been taking maths lessons from a friend for several years because I felt I gave it up too early at school,” Kreindler says.
“Our friend who we asked to help gave us the confidence to go ahead against the best medical advice. It was the right choice for us.”
With the calculations buying them some time, the couple opted for a course of egg harvesting and Kreindler harnessed technology again, using his smartphone to research the success rates of fertility clinics.
“We looked through databases and asked online what people’s experiences were with different clinics, including people who lived 6,000 miles away,” he says.
“The internet gave us the ability to search through masses of information that wouldn’t have been possible 10 years ago, and that helped us choose the right clinic based on the data we could access and analyse online.”
On Christmas Eve 2011, Marjorie underwent the procedure from which they harvested six eggs to give them a chance at becoming parents to a child of their own.
After the operation to remove the womb they also discovered an ovary containing a tumour that was also removed, and the surgeons suggested a much more radical operation followed by chemotherapy as they believed the cancer had spread.
“The operation had been planned but in the meantime we had the benefit of having another set of specialists in histology look at the tumours from both ovary and womb,” Kreindler says.
“When the second set of specialists looked under the microscope they found they were two separate primary tumours, stage one diseases, so the cancer hadn’t spread at all, which was pretty amazing.
“The other operation and treatment would have been massively life-changing – and unnecessary.”
Daughter Sienna was born in 2014 through surrogacy, but the future could have been vastly different for them, which got Kreindler thinking and has shaped much of his work since.
“We may have missed the cancer altogether and Marjorie could not be alive today,” he says.
“We were the lucky ones, but what has driven us is the fact that we shouldn’t have been lucky, we should have been average.”
When doctors are faced with unusual cases they often have to recall their own experiences or research while making a diagnosis, he says, yet there could be a wealth of information available to them if the real world data from individual patients was recorded and used for clinical decision support.
“We may have been a special case, but why weren’t any of the learnings and nuances captured for other special cases like Marjorie, so that we could catch more of these things in the same way?” he asks.
“There may have been tens of thousands of similar cases that map to similar outcomes like Marjorie’s and it got me wondering why we don’t have the support of artificial intelligence as standard to process all our medical imaging, which would then be able to say there is a 95 per cent chance of a particular scan being clear or cancerous, for example.
“In our case the scan showed no clear sign of tumours at this early stage even to the trained expert eye. For most they would have been given the all clear.
“Conversely countless others fall victim to painful, dangerous and costly medical procedures because imaging looks suspicious but turns out not to be.
“I decided it was essential that, as specialists and doctors, we move from relying on our own memories and have the support of machine intelligence to help us make better decisions.”
His first project exploring the application of machine learning to medical imaging began in the US in 2014 and is already a world leader in the early detection of lung cancer, using public databases of lung scans mapped to diagnoses and outcomes to come up with probabilities that can prevent patients from undergoing needless surgery and save anxiety, lives and money.
They discovered that if a CT scan shows up a lump in the lung, a patient would end up having a needle put into that lung 97 per cent of the time – yet this procedure is only required in 25 per cent of cases.
“The cost, the pain caused and the deaths caused by the procedure far outweigh the benefits,” Kreindler says, adding that within four weeks of using the system the 97 per cent of patients who underwent biopsies would only have been 30 per cent if clinicians were supported by the tool.
“Every time someone has a scan you should record all the details about who it is, why have they had the scan and then you follow up that person however many years later and their actual outcomes.
“Was cancer spotted immediately or was it something subtle that turned out to be cancer years later?
“You take that information and put it into an analytics engine to learn what kind of shapes or patterns lead to different outcomes.
“Sometimes the combinations of extraordinarily subtle patterns are far too subtle or complex for human eyes to see.
“Then the next time someone comes along who has a pattern that looks like one that led to cancer you can confidently say they need the further investigation and treatment.
“Thousands of cases like Marje could have been missed since her scan, but the learning from it could have saved their fertility or lives potentially.
“Equally, if something looks suspicious you may be able to say, for example, that only one in 10,000 per people with a scan pattern like that develop cancer over five years, so rescan in two years perhaps, instead of doing more harm than good.”
The project is currently being looked at by NHS decision-makers who are due to debate whether the UK should undertake a national lung cancer screening programme.
“In the US approximately $12bn a year is wasted as a result of unnecessary investigations after a scan has been done,” he says.
“In the UK we estimate that it’s proportional to the smaller population, so we’re hoping that it could save maybe as much as £2bn a year.
“But there are also the victims of medical imaging technology – known as VOMITs – who suffer because of unnecessary procedures, logically speaking that doesn’t need to happen.”
Through his London-based Centre for Health and Human Performance (CHHP) and his organisation in the US, he is encouraging start-ups to build tests that can give early markers of changes before they manifest themselves as cancer.