1
00:00:02,130 --> 00:00:04,920
So that was my last role as an academic cat herder,

2
00:00:04,920 --> 00:00:09,870
so I spent the last 10 years trying to get clinicians and academics talking to each other inside Oxford.

3
00:00:09,870 --> 00:00:13,440
And now I'm working on trying to get different disciplines working together

4
00:00:13,440 --> 00:00:18,060
across the UK through the National Consortium of Intelligent Medical Imaging.

5
00:00:18,060 --> 00:00:21,210
As Mike says, it's one of the initiatives we're working around.

6
00:00:21,210 --> 00:00:30,360
The ethics of the use of image data and medical imaging and See Me involves 15 hospital trusts across the length and breadth of the UK,

7
00:00:30,360 --> 00:00:35,910
nine industry partners. So we've got a particular interest in the use of patient and health data by companies,

8
00:00:35,910 --> 00:00:43,920
by small and international global players and of spread across cancer, cardiovascular medicine, metabolic health, stroke, you name it.

9
00:00:43,920 --> 00:00:46,080
We're probably pursuing something with it,

10
00:00:46,080 --> 00:00:52,560
and we were one of five centres set up by through the Life Sciences Industry Strategy just over a year ago to support

11
00:00:52,560 --> 00:00:57,930
aoy developments in medical imaging and digital pathology with the view that these were some of the low hanging fruit,

12
00:00:57,930 --> 00:01:04,320
why you could see the use of AI and the use of healthcare data for innovation and to drive patient impact.

13
00:01:04,320 --> 00:01:10,050
So as I said, we were set up with a range of aspirations to support companies and academics and progressing

14
00:01:10,050 --> 00:01:15,390
particular use cases of AI in health care and ensuring that we're engaging the NHS workforce.

15
00:01:15,390 --> 00:01:19,290
And how are we doing that in an ethical way that you're including patient consent,

16
00:01:19,290 --> 00:01:25,080
both for retrospective as well as prospective data collection, which some of you may have seen in the Guardian over the weekend.

17
00:01:25,080 --> 00:01:32,790
The idea of anonymized data and can you really anonymize data is one that's close to conversations about having an ally for us at the moment,

18
00:01:32,790 --> 00:01:35,160
how people feel about the use of commercial entities,

19
00:01:35,160 --> 00:01:40,350
engaging with their health datasets and also trying to ensure that we're bringing the patient and

20
00:01:40,350 --> 00:01:45,980
public perspective in which is why we've been working with Michael McKenzie and some of the and.

21
00:01:45,980 --> 00:01:54,260
So we're supporting a range of different projects, and they're sort of maps out along a pipeline of development from areas of clinical unmet need all

22
00:01:54,260 --> 00:01:59,660
the way through to companies with solutions that are looking to test and deploy into the NHS.

23
00:01:59,660 --> 00:02:05,810
And those present different challenges in sense of whether they're being prospectively or retrospectively acquired,

24
00:02:05,810 --> 00:02:11,000
whether there is informed consent around these studies and then the aspects around deployment of air

25
00:02:11,000 --> 00:02:18,200
developed by large entities such as health care in the US and the deployment of those into the NHS,

26
00:02:18,200 --> 00:02:24,170
and also the questions around how we ensure that everybody in the ecosystem is creating value back out of that.

27
00:02:24,170 --> 00:02:29,450
So not only the commercial value proposition, but also societal, academic,

28
00:02:29,450 --> 00:02:34,040
commercial and other forms of value that we're working with both Mike and the team and

29
00:02:34,040 --> 00:02:39,200
some of our external collaborators around how to capture those different times of value.

30
00:02:39,200 --> 00:02:46,460
So those are all projects we are progressing through, the UK says, and I wanted to give a couple of quick example.

31
00:02:46,460 --> 00:02:52,490
So this is one that involves when I was a start-up that spun out of the University of Oxford just over a year ago.

32
00:02:52,490 --> 00:03:02,450
So this was an academic research group that have used data from cardiac CT images and are progressing a new measure for fat attenuation index,

33
00:03:02,450 --> 00:03:09,410
which is a new marker that picks up of those 50 percent of people that go for cardiac CT because they're at risk of cardiovascular disease,

34
00:03:09,410 --> 00:03:12,770
about half the percent y going you fine of that 50 percent,

35
00:03:12,770 --> 00:03:19,160
about 25 percent of those are actually at risk of this new marker has been picked up through looking at the same images you were acquiring already,

36
00:03:19,160 --> 00:03:25,100
and you can detect new risk disease cohorts that you can then pick up and do something with more proactively.

37
00:03:25,100 --> 00:03:29,570
So the group inside the university developed an algorithm for this,

38
00:03:29,570 --> 00:03:38,090
and we're supporting them in scaling that up and gathering more datasets to develop that into a solution could then be deployed into the NHS.

39
00:03:38,090 --> 00:03:42,740
This is another quite different cohort, so we're working with a charity called him a Premier.

40
00:03:42,740 --> 00:03:51,710
Tice's UK hemochromatosis is an incredibly common genetic hereditary disorder, which is involved with presents as iron overload,

41
00:03:51,710 --> 00:03:58,130
but is chronically misdiagnosed with a lack of about 20 to 25 years before people are accurately diagnosed.

42
00:03:58,130 --> 00:03:59,810
The treatment is giving blood,

43
00:03:59,810 --> 00:04:06,410
so it's not of interest to big pharma companies because they're not going to make a Big Bang for their buck out of trying to cure these patients.

44
00:04:06,410 --> 00:04:11,270
But it is an unmet need where the use of imaging can pick up the disease through an MRI scan,

45
00:04:11,270 --> 00:04:15,660
and you can detect the iron levels in the liver and get them treated much sooner.

46
00:04:15,660 --> 00:04:22,130
So we're supporting a company and progressing how you would take that standard of care MRI images and flag up to

47
00:04:22,130 --> 00:04:28,220
patients and clinicians that they are at risk of having excessive iron and then be treated through giving blood.

48
00:04:28,220 --> 00:04:37,040
And these patients. So Kareena is actually the chair of the charity, and she was, as you can see, symptoms at 20 and over 25 years to be diagnosed.

49
00:04:37,040 --> 00:04:40,400
The charity themselves are very interested in directly donating their image

50
00:04:40,400 --> 00:04:44,120
and other patient datasets for research and for commercial research as well.

51
00:04:44,120 --> 00:04:53,060
So talking with them about what it means to be a patient that wants to donate the image records directly for research and commercial use.

52
00:04:53,060 --> 00:05:00,890
Just the other end of the extreme is E stroke suite. So this is deployed in the NHS currently and is a new method for mobile device pick

53
00:05:00,890 --> 00:05:05,810
up detection of patients more rapidly for those at risk of stroke and intervention.

54
00:05:05,810 --> 00:05:10,400
And this is about putting that absolution into the hands of the clinicians to be able to detect to read

55
00:05:10,400 --> 00:05:17,750
the scans remotely and then get the patients that need to be into the appropriate clinical care pathway.

56
00:05:17,750 --> 00:05:22,100
And we've been talking about deploying that into eight new hospitals into London,

57
00:05:22,100 --> 00:05:29,060
and that would generate over 17 million pounds worth of quality adjusted life year benefit in the next three years if that was deployed.

58
00:05:29,060 --> 00:05:37,880
So very significant potential benefits from an app based solution and just a kind of kind of screen scrolling through these.

59
00:05:37,880 --> 00:05:43,550
And then another one is critical care suite. So this is something which is a never even in the NHS.

60
00:05:43,550 --> 00:05:48,260
Nasogastric tubes are placed in order to feed patients who can't otherwise feed themselves.

61
00:05:48,260 --> 00:05:52,130
This tends to be a risky situation in the middle of the night in an eye and he department,

62
00:05:52,130 --> 00:05:58,280
where you've got a junior doc who's not slept for 36 hours and is trying to get this patient fed as quickly as possible.

63
00:05:58,280 --> 00:06:04,580
If they're incorrectly placed, they can end up not in the stomach, but in the lungs or in the chest cavity and feeding him begin.

64
00:06:04,580 --> 00:06:11,570
And you can end up either with severe adverse events or you can end up with a dead patient in order to avoid those arising.

65
00:06:11,570 --> 00:06:18,020
We're working with G to pick up an automation for this process, not to take away the decision from the clinicians, but to flag those.

66
00:06:18,020 --> 00:06:23,810
But there may be a problem. So trying to help pick up those cases where there might be a risk because otherwise

67
00:06:23,810 --> 00:06:27,650
these are picked up as a never event and no significant financial as well,

68
00:06:27,650 --> 00:06:32,600
as well as societal and individual patient impacts for this going awry.

69
00:06:32,600 --> 00:06:37,640
So those are sort of three or four of the projects that we're working on across the piece.

70
00:06:37,640 --> 00:06:46,210
And all of this is underpinned with what we've been doing and beginning with Michael McKenzie around, what does it mean to be using these?

71
00:06:46,210 --> 00:06:53,830
Ethically, particularly around image data, which has a certain sense of can you ever truly anonymize the data sets we're working with?

72
00:06:53,830 --> 00:06:58,300
Are we in effect in a position that you can't work with historic records within the hospitals?

73
00:06:58,300 --> 00:07:01,840
Because can you truly anonymize the data without consent?

74
00:07:01,840 --> 00:07:07,180
Can you be using retrospective data sets and the very different perspectives patients versus the general

75
00:07:07,180 --> 00:07:13,840
public feel about the use of image and other datasets for both academic research and commercial research?

76
00:07:13,840 --> 00:07:20,860
So we're working with both the Ethics Centre as well as Future Care Capital and some of a company called the behavioural

77
00:07:20,860 --> 00:07:27,550
architects to discuss some of the key drivers and leave us for having some of these conversations to bring the patients with us.

78
00:07:27,550 --> 00:07:33,250
And we're hoping that by doing this will have generated not only solutions that can help care in the future,

79
00:07:33,250 --> 00:07:39,790
but will have also supported the generation of imaging and other data resource for future research projects.

80
00:07:39,790 --> 00:07:47,498
And we're keen to kind of pull more people and more groups into there. So if you'd like to come and join and see me, then please do get in touch.