P ac icalMEEG | 2025 – Guioma Niso
Sou ce le el analysis II:
Analysing sou ce ime-se ies
Guioma Niso
Aix-Ma seille Uni e si é F ance | 27-31 Oc obe 2025
P ac icalMEEG
1
guioma [email p o ec ed] @Guioma Niso
P ac icalMEEG | 2025 – Guioma Niso
The human b ain
2
(S. Ramón y Cajal)
The human b ain
~86.000 millon o
in e connec ed neu ons
Each wi h ~1000
synap ic connec ions
P ac icalMEEG | 2025 – Guioma Niso 3
Human b ain
Elec oencephalog aphy
EEG
EEG
Magne oencephalog aphy
MEG
MEG
Measu ing human b ain ac i i y non in asi ely?
P ac icalMEEG | 2025 – Guioma Niso
Human b ain
Elec o/Magne o encephalog aphy (EEG/MEG)
Elec ic/Magne ic fields gene a ed by cell assemblies
4
P ac icalMEEG | 2025 – Guioma Niso
Sou ce econs uc ion
5
P ac icalMEEG | 2025 – Guioma Niso
Sou ce econs uc ion
6
Fo wa d
In e se
P ac icalMEEG | 2025 – Guioma Niso
Sou ces
How do we model he
sou ce ac i i y
ma hema ically?
Fo wa d model
Head
How do cu en s flow om
he sou ce owa ds ou
senso s h ough he issues?
Senso s
Wha senso s do we ha e
and hei posi ion wi h
espec o he b ain?
7
P ac icalMEEG | 2025 – Guioma Niso
Plana g adiome e
δB/δx
Axial g adiome e
δB/δy
Magne ome e
Bz
00
8
Fo wa d model: Senso s
(Adap ed om Field ip)
EEG MEG
0
P ac icalMEEG | 2025 – Guioma Niso
Alignmen o MEG da a: channel posi ions, headpoin s
Co egis a ion: MRI + M/EEG space → one coo dina e sys em (fiducials)
9
HEAD POINTS
Alignmen
CHANNELS
ANATOMY
MRI
T1W
dicom/nii
SEGMENTED
F eesu e
FUNCTIONAL
MEG
MEG DATA
Raw da a
Fo wa d model: Senso s
CORREG
MRI-MEG
FIDUCIALS
Ma k fiducials
Senso s
Wha senso s do we ha e
and hei posi ion wi h
espec o he b ain?
P ac icalMEEG | 2025 – Guioma Niso
ll posed p oblem: mo e sou ce poin s han senso s ( housands s hund eds)
→ infini e numbe o solu ions
16
SENSORS SOURCES
~300 ~15.000
In e se model
Cons ain s o
make i sol able
In e se model
Es ima e sou ce b ain ac i i y
om measu ed senso da a
P ac icalMEEG | 2025 – Guioma Niso 17
Ŝ = W mŜ: es ima ed sou ce ac i i y
W: in e se model (imaging ke nel)
m: measu ed senso da a
Spa ial fil e s
Ŝ: independen sou ces
uni gain and minimize a iance
●Beam o me s
Dis ibu ed sou ces
Ŝ: dis ibu ed sou ces
minimize esiduals and noise
●Minimum no m es ima ion
Dipole fi ing
Ŝ: one o e y ew sou ces
cons ain s: limi sou ces
●Single dipole
In e se model
(Wes ne e al. 2022)
P ac icalMEEG | 2025 – Guioma Niso
Cons ained: no mal o co ex Uncons ained: 3 o hogonal dipoles
Sou ce ac i i y
19
1 dipole 3 dipoles (no m) Absolu e aluesAbsolu e alues
P ac icalMEEG | 2025 – Guioma Niso
SOURCES
20
ALL ROIs ATLAS
P ac icalMEEG | 2025 – Guioma Niso
SOURCES
21
ALL ROIs Scou unc ion
Mean: A e age all he signals.
Mean(no m): A e age absolu e
alues o all he signals.
PCA: Fi s mode o he
P incipal Componen Analysis.
Max: Fo each ime poin , ge
he maximum ac oss all he
e ices.
Powe : A e age he squa e o
all he signals.
RMS: Squa e oo o a e age
he squa e o all he signals.
All: Re u ns all he signals.
ATLAS
P ac icalMEEG | 2025 – Guioma Niso
We can’ a e age indi idual sou ce maps → p ojec o a common empla e
P ojec o de aul empla e
22
P ac icalMEEG | 2025 – Guioma Niso
SOURCES
~15000
23
B ain signal analysis
?
SIGNAL
SENSORS
~300
CONNECTIVITY
PREPROCESSING
EVOKED
POTENTIALS
SPECTRAL
ANALYSIS
TIME–
FREQUENCY
WAVEFORM
SHAPE
CROSS-FREQ
COUPLING
p
a
NETWORKS
ACQUISITION
P ac icalMEEG | 2025 – Guioma Niso
Connec i i y
24
S uc u al
physical connec ion
21
Func ional
ela ion be ween signals
12
P ac icalMEEG | 2025 – Guioma Niso
Func ional Connec i i y
25
1
2
3
?
Fo a comp ehensi e e iew on unc ional and effec i e
connec i i y me ics: (Niso e al. 2013)
P ac icalMEEG | 2025 – Guioma Niso
Pea son’s co ela ion (COR)
Linea co ela ion in ime domain be ween
x( ) and y( ) a ze o lag
COR = 1
COR = -1
COR = 0
COR = 0
Cohe ence (COH)
Linea co ela ion be ween x( ) and y( ) as a
unc ion o he equency
26
Classical Me hods
P ac icalMEEG | 2025 – Guioma Niso
Connec i i y
49
Non Di ec ed
Func ional connec i i y
12
Di ec ed
Effec i e connec i i y
12
Symme ical Non symme ical
P ac icalMEEG | 2025 – Guioma Niso 50
S eng h: sum o weigh s
o links connec ed o i
Deg ee: numbe o
links connec ed o i
Clus e ing: Likelihood o
neighbou s also connec ed
Modula i y: Max wi h-in,
Min be ween modules
Walk Sho es pa h
Pa h
Be weenness:
Numbe o all sho es
pa hs in he ne wo k ha
con ain i
Ring La ice Small Wo ld Random Comple e
Complex Ne wo ks
(Rubino & Spo ns, 2010)
P ac icalMEEG | 2025 – Guioma Niso
SOURCES
~15000
51
B ain signal analysis
?
SIGNAL
SENSORS
~300
CONNECTIVITY
PREPROCESSING
EVOKED
POTENTIALS
SPECTRAL
ANALYSIS
TIME–
FREQUENCY
WAVEFORM
SHAPE
CROSS-FREQ
COUPLING
p
a
NETWORKS
ACQUISITION
P ac icalMEEG | 2025 – Guioma Niso
Good scien ific p ac ice (Niso e al. 2022a,b)
54
Online esou ce:
h ps://o eoni.gi hub.io
P ac icalMEEG | 2025 – Guioma Niso
¡Muchas g acias!
52
¡Muchas g acias!
@Guioma Niso
guioma [email p o ec ed]
G upo de Neu oimagen
P ac icalMEEG | 2025 – Guioma Niso
Sou ce le el analysis II:
Analysing sou ce ime-se ies
Guioma Niso
Aix-Ma seille Uni e si é F ance | 27-31 Oc obe 2025
P ac icalMEEG
53
guioma [email p o ec ed] @Guioma Niso
