1
00:00:08,174 --> 00:00:09,884
My name is Concha Bielza.

2
00:00:09,884 --> 00:00:12,971
I am professor in the Department
of Artificial Intelligence

3
00:00:12,971 --> 00:00:15,974
at the School of Computer Science of UPM.

4
00:00:15,974 --> 00:00:18,727
I co-lead the Computational Intelligence
Group.

5
00:00:18,727 --> 00:00:22,355
This is a research group
devoted to modelization

6
00:00:22,439 --> 00:00:25,316
from statistical
and machine learning perspectives

7
00:00:25,316 --> 00:00:28,319
and also devoted
to heuristic optimization.

8
00:00:28,403 --> 00:00:32,240
I am also co-director of the Ellis Unit
Madrid,

9
00:00:32,323 --> 00:00:38,413
an important application area
since 2009 is Data-Driven Neuroscience,

10
00:00:38,496 --> 00:00:42,083
where we build models from data
to answer different

11
00:00:42,083 --> 00:00:44,335
research questions.

12
00:00:50,550 --> 00:00:52,093
A major line of research

13
00:00:52,093 --> 00:00:55,430
has been in neuroanatomy
with different contributions.

14
00:00:55,472 --> 00:00:57,015
As the classification

15
00:00:57,015 --> 00:01:01,019
of cortical and gabaergic interneurons
according to their morphology.

16
00:01:01,102 --> 00:01:06,441
This is a hard problem since neuroanatomists
 dont agree on the nomenclature.

17
00:01:06,524 --> 00:01:09,402
Hence,
this supervised classification problem

18
00:01:09,402 --> 00:01:12,405
doesn't have instances with clear labels,
which is very,

19
00:01:12,405 --> 00:01:15,366
very challenging from the machine
learning point of view.

20
00:01:15,450 --> 00:01:19,913
We have tried to identify
similar groups of dendritic spines.

21
00:01:19,996 --> 00:01:21,331
The spatial distribution of

22
00:01:21,331 --> 00:01:25,376
synopsis along
dendritic networks has been also studied

23
00:01:25,460 --> 00:01:27,545
the simulation of dendritic trees

24
00:01:27,545 --> 00:01:31,007
of pyramidal cells, or the wiring economy.

25
00:01:31,007 --> 00:01:36,554
Also for this kind of cells testing
the hypothesis that our arborizations

26
00:01:36,554 --> 00:01:41,518
optimize brain connectivity
in terms of total wiring length.

27
00:01:47,899 --> 00:01:51,861
The second main line of research
is in neurological diseases

28
00:01:51,945 --> 00:01:54,239
where we have tried to search for

29
00:01:54,239 --> 00:01:57,742
genetic biomarkers in Alzheimer disease

30
00:01:57,826 --> 00:02:00,787
or the classification of task related

31
00:02:00,787 --> 00:02:04,207
mental activity
from Magnetoencephalography data,

32
00:02:04,374 --> 00:02:09,379
which is very relevant in brain computer
interface and in Parkinson's disease.

33
00:02:09,379 --> 00:02:14,300
We have identified different subtypes
with motor, and non-motor symptoms

34
00:02:14,384 --> 00:02:17,762
or the prediction
of the dementia development.

35
00:02:17,846 --> 00:02:22,016
And also we have tried to predict
genetic health related quality of life

36
00:02:22,016 --> 00:02:25,019
measure with five dimensions mobility,

37
00:02:25,019 --> 00:02:27,689
self-care, usual activities,

38
00:02:27,689 --> 00:02:32,527
pain, discomfort and anxiety, depression
by using the very,

39
00:02:32,527 --> 00:02:36,406
very specific quality of life
Parkinson Disease Questionnaire.

40
00:02:36,489 --> 00:02:39,367
In epilepsy, we have tried

41
00:02:39,367 --> 00:02:43,997
to predict the outcome of temporal lobe
epilepsy surgery

42
00:02:44,080 --> 00:02:49,085
to see whether the patient
will fully recover from epilepsy or not

43
00:02:49,169 --> 00:02:51,212
by using clinical variables

44
00:02:51,212 --> 00:02:55,758
and also pathological
on neuropsychological evaluations.

45
00:02:55,842 --> 00:02:58,511
We have also implemented an open machine

46
00:02:58,511 --> 00:03:01,848
learning and a statistical web framework

47
00:03:01,931 --> 00:03:06,060
with neuroscience applications,
and this is called NeuroSuites

48
00:03:06,144 --> 00:03:09,314
Some of the works have been published
in top journals

49
00:03:09,314 --> 00:03:13,818
like Nature Reviews Neuroscience
or Nature Neuroscience.

50
00:03:13,902 --> 00:03:16,029
And also always

51
00:03:16,029 --> 00:03:19,282
we have preferred models
that are called Bayesian networks

52
00:03:19,282 --> 00:03:22,285
because they are interpretable,
and this is very important

53
00:03:22,535 --> 00:03:25,246
in building
unexplainable artificial intelligence

54
00:03:25,246 --> 00:03:30,418
where humans can trust
on intelligent systems and avoid biases

55
00:03:30,418 --> 00:03:33,504
and discriminative and opaque outcomes.

56
00:03:40,511 --> 00:03:43,514
Our work in neuroscience has been framed

57
00:03:43,640 --> 00:03:47,727
within two major projects of ten years
duration.

58
00:03:47,727 --> 00:03:51,648
 one starting in 2009, the Cajal

59
00:03:51,648 --> 00:03:55,485
Blue Brain Project,
funded by the Spanish Ministry

60
00:03:55,568 --> 00:03:58,071
on the second from 2013

61
00:03:58,071 --> 00:04:01,532
to this year 2023,
The Human Brain Project.

62
00:04:01,741 --> 00:04:05,119
This is a FET flagship
of the European Commission

63
00:04:05,203 --> 00:04:07,789
where we have been involved
from the very beginning in the ramp

64
00:04:07,789 --> 00:04:11,209
up phase and in the three specific
grant agreements 

65
00:04:11,209 --> 00:04:16,464
In this projetcs, currently we are 
developing a genetic addressing toolbox

66
00:04:16,547 --> 00:04:19,092
with a web interface and a plugin

67
00:04:19,092 --> 00:04:21,970
for the Brain Atlas view of EBRAINS.

68
00:04:21,970 --> 00:04:25,723
EBRAINS is the digital brain research
infrastructure created

69
00:04:25,723 --> 00:04:30,478
by the Human Brain Project in this case,
and massive Bayesian networks are used

70
00:04:30,561 --> 00:04:34,816
for interactively learning gene regulatory
networks from gene expression.

71
00:04:34,816 --> 00:04:37,402
Data extracted from the Atlas.

72
00:04:41,698 --> 00:04:43,074
I think it students have

73
00:04:43,074 --> 00:04:48,288
a great opportunity to gain insights
from the application of machine learning

74
00:04:48,371 --> 00:04:51,958
to all kinds of data
that are available nowadays

75
00:04:52,000 --> 00:04:56,129
generated by different neural technologies
like morphological,

76
00:04:56,212 --> 00:05:00,591
electrophysiological,
molecular clinical data, etc.

77
00:05:00,675 --> 00:05:05,471
I think that we have done our bit
in a student education by offering to book

78
00:05:05,471 --> 00:05:10,727
data driven computational neuroscience,
machine learning and statistical models.

79
00:05:10,810 --> 00:05:13,938
This book was published in 2020
by Cambridge

80
00:05:13,938 --> 00:05:17,275
University
Press and has more than 700 pages.