All  right.  Hello,  everyone. Thank  you  so  much  for  being  here this  morning  for  the  Georgia  Tech  neuro  Seminar  Series. My  name  is  named  Buzz  seal. I'm  a  neuroscience  PhD  candidate  and  Dr.  Michael gorgeous  lab  over  at  Emory  and  I'll be  introducing  our  speaker  today. It's  my  honor  to  introduce  Dr.  Tilly, old  Bergman  and  Associate Professor  of  neural  stimulation, the  Johannesburg  Johannes  Gutenberg University  Medical  Center  and  minds  Germany. At  the  University,  Dr.  Bergman  is the  deputy  head  of  the  neuroimaging  center  and principal  investigator  of  the  neuron  stimulation  group  at the  Leibnitz  Institute  for  resilience  research, which  focuses  on  studying the  neurobiological  correlates  of emotional  memory  processing  for  resilience. Dr.  Bergman  is  a  biological  psychologists and  cognitive  neuroscientists  by training  interested  in  the  function of  neuronal  oscillations  and  cognition, in  particular  for  the  organization  of human  information  processing  and  the  gating  of synaptic  plasticity  in  the  waking,  sleeping  human  brain. His  methodology,  methodological  focus is  on  those  simultaneous  combination  of non-invasive  transcranial  brain  stimulation  techniques with  neuroimaging  and electrophysiology  and  the  development  of novel  approaches  such  as state  dependent  brain  stimulation. Dr.  Bergman  is  also  the  recipient of  the  prestigious  John  Roth  will  award  by the  brain  box  initiative  to  acknowledge  and  reward excellence  and  non-invasive  brain  simulation  research. Today,  Dr.  Bergman  has  traveled from  Germany  to  present  his  talk  titled combining  transcranial brain  stimulation  with  neuroimaging for  state  dependent  stimulation and  causal  network  interrogation. A  quick  note  for  questions. If  we  could  save  all  questions  for  the  end  of  the  talk, there'll  be  a  brief  15-minute  Q&A  session during  which  we'll  walk  around and  pass  around  the  microphones. If  you  have  a  question.  Alright. With  that,  please  join  me  in welcoming  Dr.  Tilburg  till  old  Bourbon. Thank  you  very  much. Adjustment.  Thanks,  Michael  for  having  me  here. Thank  all  of  you  for  coming  to  this  meeting and  listen  to  the  book  I'm  going  to  present. I'm  a  psychologist  by  training, but  I'm  working  systems  neuroscience  for  17  years  or  so, mainly  with  brain  stimulation  and in  combination  with  imaging,  EEG, FY,  everything  I  can  combine. And  I'm  going  to  talk  about  these  different  transcranial, so  noninvasive  brain  stimulation  techniques. I'm  going  to  give  you  a  bit  of  an  overview  in the  beginning  for  those  of  you not  familiar  with  the  techniques. And  then  I'm  mainly  going  to  talk  about  TMS. So  cranial  magnetic  stimulation as  combination  with  fMRI  and  EEG. And  if  I  have  time  left  in  the  end, I  want  to  talk  a  little  bit  about transcranial  magnetic  stimulation. That's  basically  the  new  kid  on  the  block. So  noninvasive  brain  stimulation  techniques. If  I'm  talking  too  fast or  you  can't  hear  me,  just  let  me  know. The  presenter  stopped  actually  working. I'm  just  you  started. So  just  that  was  it. Okay.  Thanks.  Good. So  transcranial  magnetic  stimulation,  in  a  nutshell, what  we  do  is  we  have  a, a  cup  of  cold  basically  in  a  plastic  housing  Double  coil. Let's  head  over  the  scalp. And  for  the  fraction  of  a  millisecond  is  strong, current  is  running  through  that  coil producing  a  magnetic  field, which  into  the  brain,  where  it induces  an  electric  field  in  the  brain  tissue, which  is  then  exciting. Axons  causing a  full  depolarization  and  action  potentials. Eventually  electric  stimulation  technique. And  the  magnetic  field  is  just  a  Trojan  horse  to bring  it  painlessly  through  the  skull  into  the  brain. If  you  stimulate  the  motor  cortex,  then  synaptically, we're  going  to  excite  this  large pyramidal  cells  in  layer  five, which  project  down  to  the  spinal  cord and  then  onto  the  muscle. We  can  measure  surface  EMG  electrodes, the  so-called  motor  evoked  potential, that  MAP,  that's  the  workhorse  of  the  TMS  community. Because  it's,  the  amplitude  nicely  shows  it's proportional  to  the  number  of cortical  output  neurons  that  you  excite  above  threshold. So  there  are  at  least  three  sine  x is  already  in  network  stimulation  effect. Outside  of  the  motor  cortex, we  can  quantify  the  brain  responds  by  e.g. transcranial  by  TMS  evoked  potentials, TPS,  which  can  be derived  by  multi-channel  EEG  recordings  at  the  same  time. I'm  not  going  to  talk  about  those  a  lot  today. Just  to  show  you  how  TMS  works  in  principle. Then  we  have  another  tick,  tick. I'm  not  going  to  talk  about  it  all. Just  for  the  sake  of  completeness. It  has  electric  stimulation. T  is  we  use recurrence  injected  by two  or  more  electrodes  attached  to  the  scalp. We  talked  about,  if  you  milli  amp  and a  few  volt  only  and  distributed  widely  across  the  cortex. Direct  current  stimulation  here some  early  recordings  animals  from the  '60s  that  show the  principal  nicely  depending  on  the  polarity, it  can  increase  a  spontaneous  firing or  you  can  decrease  the  spontaneous  firing, but  it's  not  triggering  action  potentials. It's  not  like  timess  which  evoke  an  action  potential. It's  more  changing  the  threshold  for spontaneous  activity  to  increase  or  decrease. If  you  do  this  with  an  alternating  current, it  is  that  you  can  entrain  oscillations. And  it  works  nicely  and  in  vitro  and  also  in  animals. Because  they  can  reach high  enough  stimulation  intensities, wet,  it  can  entrain  firing,  e.g. in  humans,  it's  a  bit  more  difficult because  we're  somewhat  limited  with  the  intensities. So  we  are  like  a  borderline  efficacy. Sometimes  you  get  very  nice  effect. Sometimes  it's  just  not  strong  enough  to produce  any  treatment  effect. And  then  the  third  technique, which  just  emerged  in  the  last  few  years. Well,  it's  also  much older  at  first  attempts  have  been  in  the  30s, but  no  re-emerging  as a  technique  for  neuromodulation  in  humans. Ultrasonic  simulation  with  ultrasound  transducer, which  focuses  the  kinetic  energy  of  ultrasound  waves. Humans  250-1650  khz. And  it  focuses  it superficially  are  too  deep  in  the  brain. And  depending  on  the  type of  transducer  and  the  frequency, you  can  have  different  sizes  of these  slightly  cigar-shaped  foci,  e.g. with  500  kw,  about  three millimeter  width,  many  meter  length. This  is  much,  much  more focused  than  what  we  can  do  with  TMS. Letting  them  T  is  an  attempt to  focus  it  under  the  thalamus  in  humans. And  depending  on  how you  position  your  transducers  around  the  cortex, it  can  be  really  millimeter  precision, or  we're  talking  about  this  cigar-shaped,  elongated  foci. A  lot  of  the  energy  is absorbed  by  the  scale  is  scattered  by  this  guy, it's  reflected,  so  it's  the  skull  is  our  enemy  here. Tms  can  overcome  the  scale  easily. But  first,  I  just  don't  really has  a  problem  go  through  it. But  it's,  it's  possible  that  this  lower  frequencies and  one  just  has  to  do  proper  simulations. I  come  back  to  that  later.  Hopefully  you  have  time. And  the  idea  is  that  stopped  working in  that  most  likely it  works  where  the  activation  of mechanosensitive  ion  channels  when  also worked  at  directly  mechanically  on  the  membrane. The  mechanisms  discussed,  but  most  likely  it's working  activating  the  kind  of  sensitive  ion  channels witnessed  an  influx  in cascade  that  eventually  also  cause  action  potentials. So  its  parts.  Potentially  superficial  technique, but  it's  not  an  immediate  synchronization takes  a  few  hundred  milliseconds  possibly. These  are  the  three  main  techniques. Now  for  non-invasive  transcranial  stimulation  in  humans. You  can  do  a  lot  of  different  things  with  it. So  in  principle, online  offline  approaches  only  means  that  you stimulate  and  you  look into  the  immediate  response  of  the  brain  to the  stimulation  and  its  behavior  consequences possibly  flavors. So  you  can  use  it  to  quantify some  properties  of  a  network  excitability, connectivity,  etc. Estimating  strong  enough  to  get  an  output  in MAP  or  TPU,  both  response. Mainly  talking  about  that  today. You  can  interfere  with  ongoing  task related  brain  activity  and disturb  a  brain  region  relevant for  certain  cognitive  performance. And  then  you  see  a  change in  reaction  time  and  error  rates,  etc. You  can  just  gently  modulate  the  timing  or the  level  of  activity  by  glutamate  stimulation, e.g.  routines  for  TACS. And  then  there  are  the  offline  effects were  prolonged  protocol, repetitive  stimulation  and  uses  ATP  or  XD  like synaptic  plasticity  and  thereby  After Effects  it  outlast  the  stimulation  by  minutes  to  hours. Again,  you  can  image  those  later  on. Eeg  with  FMI,  with  behavior,  etc. And  we  try  to  improve  the  targeting  of  these  methods. The  selection  of  parameters  in  terms  of  where, when  and  how  to  stimulate. Also  based  on  imaging  information running  functional  localizers  with  fMRI  or  EEG  MEG, doing  using  the  individual  anatomy  and  if  it modelling  to  know  exactly  where  we  want  to  stimulate. E.g.  evoked  potentials  or the  online  tracking  of ongoing  EEG  activity  to  target  specific  events. I'm  going  to  talk  about  that  today. To  increase  the  temporal  precision. And  to  Pittsburgh,  we  could  also  individualize our  stimulation  by  taking information  from  the  participant  about, let's  say  the  frequency  of  oscillations  to  then  adjust our  stimulation  protocols  accordingly. Non-invasive  ventilation  is  used  a  lot  in cognitive  neuroscience  to  answer  questions about  which  brain  regions  are processes  are  causal  for  specific  functions. Often,  you  will  written  papers, simplified  statements  like  we  saw  in ventilation  technique  to  changing  your  activity and  region  x  that  changes  cognition  and  behavior. So  this  region  is  causal  for  that  function. But  of  course,  there  is  oversimplified. And  behind  it  is  the  implicit  assumption that  your  presentation  technique  produces  an  E-field. If  dog  with  teammates  or  t is  in  the  brain  and  the  target  region, which  then  changes  local  neural  activity, which  then  maybe  also  cause  remote  network  effects. This  is  always  the  network  simulation,  right? Um,  which  then  changes information  processing,  you're  processing  in  the  brain. So  basically  for  cognitive  function, which  eventually  can  be  measured by  a  task  with  certain  behavioral  measurements. If  this  chain  of  causation  with  all  these cause  effect  pairs  and  only  if  all  these  red  arrows here  true  and  realized  actually  in  the  correct  way, then  you  can  make  the  claim  the  same  time. Have  some  off-target  activation  stimulating  region that  you  didn't  mean  to  stimulate, co-activated  because  it's  not as  focused  as  you  would  like  to  have  it. Then  you're  stimulating  networks  didn't  mean  to simulate  effect  on  other  cognitive  functions are  the  same  you're  measuring,  right? Nucleate  conference. Perfect  costume,  nation  confronts  the  TMS, makes  a  click  when  the  coil  discharges and  you  feel  it  on  the  scalp, activate  some  Scott  muscles. T  S  corps  is  a  tingling  sensation  on Skype  because  phosphenes  with  TACS,  etc. All  these  things  activate your  brain  and  thereby  have  effects on  brain  functioning  and  on  your  causal  inference. And  then  different  task  demands, brain  states  which  I'm  going  to  talk  about exactly  which  have  an  effect  and  all  of  those. So  it's  tricky  if  you  want  to  do extra  conclusion  about  causality  here. Especially  given  that  all  of these  variables,  most  of  them  are  latent. They're  not  directly  observable  where the  E  field  we  can  simulate, it's  very  difficult  to  measure  in humans  and  healthy  living  humans. The  new  activity  list  with  neuroimaging. And  this  is  also  where  I'm  going  to  talk  about  that particularly  today.  The  cochlea. Again,  we  can  model  and  we  can  have good  psychophysics  to  measure  the  response. Focused  ultrasound  with  the  same  problems. Just  smaller,  smaller  focus, but  it  can  also  cause simulate  regions  would  not  like  to  stimulate. And  also  there  are  some  auditory  confounds can  come  back  to  at  the  very  first  time. So  the  imaging  is  important because  it  allows  us  to  create  proof  of newer  target  engagement  that the  brain  region  we  meant  to  simulate  actually what  activated  and  to see  the  network  that  ultimate  activated. And  maybe  there's  some  brain  regions that  explain  or  behavioral  effects  even better  than  the  nodes  of the  network  that  we  started  stimulating. So  we  have  all  these  different  imaging  techniques,  EEG, MEG,  FMI,  if  NewRez,  pet,  etc. Slightly  different  spatial  and  temporal  resolution  here. So  the  spatial  and  temporal  resolution, but  they  all  have  the  similar  level. And  we  can  combine different  techniques  like  EEG  and fMRI  to  get  the  best  of  both  worlds. The  temporal  resolution  of  Egypt, good  spatial  resolution  of  fMRI. And  luckily,  we  can  also  combine  them  with the  coarser  grain  manipulation  techniques. Techniques.  And  many  of  them  we  can  combine  online. So  this  is  technically  demanding. It's  creating  a  lot  of  artifacts  in the  imaging  modality  and  I  could  talk  just  days. What  is  artifacts?  So  if  you  have  questions,  feel  free. I'm  just  going  to  skip  over the  technical  difficulties  in  this  talk. Otherwise  I  would  get  stuck  pretty  early. I'm  going  to  start  with  TMS,  fMRI. So  that's  all  set  up  in  minds. We  use  MR.  receive  calls and  seven  channel  surface  because  two  of  them, so  14  channels  can  flexibly  position  them. And  our  team  S  coy  here  actually  on  top. So  we're  stimulating  through  this  atrazine  coil, which  gives  us  good  signal  to  noise  ratio at  the  cortex  level, slightly  worse  deep  in  the  brain, but  there's  always  positive  to  pick. We  have  online  Stereotactic, frameless,  Stereotactic  new  navigation. I'm  using  this  heterozygous  today, even  with  the  subject  is  in  the  center  of  the  ball. So  we  can  continue  tracking  the  TMS  coil  relative  to the  head  where  the  subject is  in  the  MRI  and  is  being  scanned. Questions  like  why  shocked  with  ID. Main  reason  is  we're  able  to  show  proof  of target  engagement  both  at the  cortical  level  where  we  stimulate, but  also  at  the  network  level  in  deeply  in  regions that  we  might  stimulate  indirectly, maybe  the  main  purpose  of  the  study. We  can  also  map  effective  connectivity is  stimulating  one  reach  and  getting  a  response in  another  shows  you  that  there's an  extra  connections  not  functioning  cognitive, effective  connectivity  Reggie,  It's  directed towards  we  can  change parameters  like  the  intensity  or the  coil  orientation  and  see  which  works  best. For  subject.  In  principle, you  can  investigate  state  dependency as  me  so  many  times  that the  network  activation  particularly  depends a  lot  on  what  the  subject  is doing  at  the  time  of  stimulation. The  activations  wrote  it  into  different  networks, depending  whether you're  processing  a  stimulus  at  the  time  or  not, in  which  state  you  are. And  helps  us  to  interpret  our  data. Especially  because  we  can  see the  conference  we  are  trading. I  mean,  you  see  auditory  cortex  activated, which  EMS  click  in  the  EEG  also  get  auditor  confronts, but  it's  more  difficult  to  tell  from  the  technique. We  can  see  activation  due  to  expect, expectation  of  upcoming  stimuli,  sensation,  etc. And  when  we  find  some  behavioral  factor  for  assimilation, we  might  find  a  brain  region in  which  the  activation  of the  Timaeus  induced  activation. Cool.  It's  much  better  with  the  behavioral  effects than  our  original  target. So  we  can  actually  pinpoint  the  We're, the  estimated  effect  in  the  brain happens  That's  behave  with relevant  examples  here  from  our  ongoing  work. When  focusing  on  the  motor  cortex. Food  has  been  already  like  in  the  early  2000s. Well,  there's  a  nice  activation  and  at  the  TMS  coil  we stimulate  but  only  at  super,  super  special  intensities. So  could  we  just  enough  energy  to  get  it  bold  response? At  high  intensities  at  the  motor  cortex, you  get  a  muscle  twitch,  right? And  then  you  get  feedback, which  activates  the  sensory  motor  cortex. There  has  been  work  to  try to  somehow  indirectly  figure  that  out. Okay,  let's  solve  that  by  blocking  that  route. So  we  use  an  ischemic  nerve  block, taking  a  blood  pressure  cuff, inflating  it  to  the  systolic  blood  pressure  and keeping  it  like  this  for  about  50  min. 50  after  about  25  min, the  block  is  complete  and is  not  able  to  produce  any  piece  anymore.  They  disappear. In  red. Sap's  measure  with  EEG  disappear. There's  no  information  going  in  any  direction. Then  we  have  this  15, 20  min  time  to  make  measurements. So  in  this  study, so  we  do  this  thing  in  the  EMR  if it  wouldn't  have  been  already  a  comfortable  enough. And  just  to  show  that our  outer  core  is  whichever  special  are  sensitive. If  you  do  finger  tapping  with  nice  activation, sensorimotor  cortex  and  motor  network.  So  that  works. Then  in  one  session, we  want  to  show  that  the  ischemic nef  blocks  actually  works. So  we  do  electric  stimulation of  the  median  nerve  at  the  wrist  at an  intensity  that  gives  you  all  the  muscle twitches  and  activation  of  the  sensorimotor  cortex. And  then  we  block  it  and  redo  it. And  what  you  see  is,  well, first  we  get  sensory  motor  activation due  to  this  electric  stimulation. During  the  ischemic  enough, it's  gone,  and  that's  the  difference. Okay,  So  the  block  works. Now  we  do  this,  Artemis. We  stimulate  the  motor  cortex. Auditory  cortex  activation  from the  Mesozoic  is  even  louder  in the  EMR  because  of  the  magnetic  fields  they're  acting. Then  those  stimulated  120%  of the  resting  motor  special  so  clearly  above  threshold. So  you  won't  get  MEPs and  twitches  if  it  wouldn't  be  blocked, it's  almost  completely  gone. This  is  a  little  bit  of  activation in  the  postcentral  gyrus. But  most  of  the  differences  in  most  of the  motor  cortex  activation  is  removed. Which  means  actually  other  studies  reporting superficial  defects  in  the  motor  cortex and  other  domestic  oil  basically  show  we  often  feedback. So  it's  good  to  check.  And  be careful  when  you  look  at  the  motor  cortex  because of  its  output  feature, the  MEP  mustard,  which  is  also  the  confound. Now  moving  into  different  regions. So  imagine  we  will  be  interested  in stimulating  the  subgenual  anterior  cingulate  cortex,  e.g. is  a  target  for  the  treatment of  depression  or some  other  regions  that  we're  interested  in. The  vmPFC  for. Fear  extinction,  memory  consolidation. Then  we  can  assume  that  there  is a  polysynaptic  link  from  some  corticosteroids. And  that's  been  a  lot  of  work  from the  community  interested  in treatments  for  depression  and  microphone  publication. These  guys  have  worked  out  that  if  you  look  at resting  state  functional  connectivity with  FMRI  and  you  take  this, Jason  sees  the  target  region is  correlated  anti-correlated  networks. And  you  can  personalize the  treatment  to  basically  by  selecting  the  target coordinate  is  the  coordinate  with  the  boxes  with a  maximally  anti-correlated  Fracture  connectivity with  this  deep  brain  region  in  the  SAC. Problem  is  actually, the  closer  you  are  to  this  optimal  point, the  better  the  treatment  response. And  this  has  not  been  integrated  in  accelerated  protocols like  the  St.  different protocol  that  has  been  developed  at  Stanford. So  people  are  using  this  kind  of  individualization. Now.  The  thing  is, it's  a  single  coordinate.  Rotations  but  unclear. And  we  simulated  the  E  field induced  by  the  Timaeus  in  standard  orientation. When  she  did  it  for  a  lot  of different  points  and  orientations. And  each  term  evaluated  the  field. And  you  see  that  with  the  standard  orientation  will produce  a  field  that's  not  so nicely  overlapping  with  your  target. Or  targets  would  be  the  anti-correlated  cluster  effect you're  stimulating  the  positively  correlated. So  we  then  basically optimized  orientation  and  position for  the  induced  electric  field, which  is  the  effective  part  of the  simulation  to  maximally overlap  with  a  negatively correlated  once  while  trying  to  maximally spare  the  positive  collisions. And  this  results  in  very  different  connotations. So  the  approach  is  basically  we  calculate  for auditory  targets  the  vmPFC  or  function  Connectivity  Map individualized  get the  negative  functional  connectivity target  to  get  the  positive  one. But  because  we  actually  wanted  in  the  study  to  see whether  the  two  have  opposite or  different  effects  in  different  regions. Because  that's  not  empirically  shown  yet. It's  an  assumption  to our  optimization  position  that you  may  score  when  a  participant. Then  actually  also  select  control  targets. One  for  each  cluster  on, for  the  positive,  one  for  the  negative, which  are  maximally  uncorrelated. So  it's  been  as control  condition  because  we  just  moved  a  little  bit  to a  region  where  the  E  field is  mainly  stimulating  boxes  that  have  no  connection, no  freshman  connection  with  our  target. But  we  create  a  similar  noise  and  sensation. This  ultra-wide  needs  two  different  ones. Depending  whether  you're  on  the  site, you  might  get  more  dramas  that  is  more unpleasant  compared  to  the  medial  sides. So  we  have  to  mesh control  sites  and  correlating nicely  matched  as  possible,  right? It's  never  perfect.  Then  we  determined the  stimulus  intensity  also based  on  the  effect  modelling. And  then  we  actually  stimulate  the  subjects  at that  rotation  and  position. We  do  get  a  positive  portraits  bonds  and  of  the  TMS  coil. And  this  time  there's  no  real  friend  feedback. It's  the  DLPFC,  not  the  motor  cortex. And  in  this  case  you're  seeing  a  subject,  pilot  subject. The  vmPFC  bit  Materia, we  get  a  negative  corresponds. This  is  actually  already  the  contrast. So  the  negative  F  to  target with  respect  of  control  chart. So  this  removes  all  the  auditory  confronts or  the  somatosensory  conference, all  the  unpleasantness, maybe  the  anxiety  related  to  being  in  the  MR. So  all  those  effects  are  matched  and  it's only  by  moving  the  color  a little  bit  to  a  different  connectivity  target. So  we  are  currently  running  the  study  with  14, 16  subjects,  but  I  don't  have  the  data  yet, so  I  help  them  sooner. They  able  to  show  that.  But  I  can  show  you already  is  that  the  matching works  for  the  first  16  subject  to  so with  the  loudness  ratings, two  different  clusters,  negative  to  positive, and  then  the  actual  target  and  the  control. They're  not  significantly  different. Or  to  the  unpleasantness  rating  is  similar. And  indignation  intensity, even  though  which  could  result in  different  signal  intensities. Because  we  choose  for the  control  tower  in  intensity  that  causes similar  unpleasantness,  similar  loudness. But  it  might  not  reside  in the  same  physical  intensity,  but  that's  secondary. Okay,  so  hoping  that  the  main  sample  auto  works out  a  method  by  which we  can  indirectly  targeted  brain  structures. Now  actually  taking  the  time, switching  gears  and  going  to  EGI, which  has  the  advantage  of  being  fast, has  a  high  temporal  resolution. And  I'm  mainly  going  to  present  the  work. We  did  not  use  it  as  an  output  but as  an  input  to  decide  when  to  stimulate. The  brain  has  been  treated  for  a  long  time, is  static  black  box  which  is stimulated  some  point  randomized. And  we  get  the  output,  let's  say  MAP  or  whatever, which  has  a  lot  of  variability. If  we  treat  the  brain  is the  dynamic  system  that  it  actually  is  with  a  lot  of fluctuating  brain  states  would sort  our  triads  according  to  the  brainstem. At  the  time  of  stimulation, we  might  get  very  different  outcomes. And  I'm  interested  in  neural  oscillations. I  think  they  also  define  brain  states, global  brain  states,  but  also  local  corticosteroids. They're  often  generated  code  because  the  limbic  loop, depending  on  those  different  structures involved  than  the  control  of  neuromodulators, particularly  noradrenalin,  dopamine,  and  so  forth, which  are  relevant  for changing  actually  the  brainstem with  the  dominating  frequency, the  close  your  eyes  or  you  get  relaxed. Alpha  oscillations,  you  get  drowsy. It  disappears  again,  TDA  kicks  in  and then  flight  sleep  and  sleep  spindles. Deep  sleep,  it's  largest  synchronized  cortical  activity is  slow  oscillations. Typically  not  coma,  but  there  will  be  more  synchronized. And  interestingly,  but  not  in  the  scope  of  this  talk. With  increasing  synchronization,  you  also have  a  loss  of  consciousness  happening. Neural  oscillations  are  involved  in many  different  cognitive  and  motor  functions. Almost  anyone  function. And  the  dysfunctional  in many  neurological  and  psychiatric  disease  that  the  case, why  are  they  important  for  everything? Because  even  if  they  are the  major  building  blocks  for  computation  in  the  brain. And  just  to  briefly  go  over  some  common  principles. So  if  this  rhythmic  input,  output  gain  modulation, so  no  matter  how  an  oscillation  is  generated, more  excitable  in  a  less  excitable  state. Typically. In  the  more  excitable,  stable  it's  depolarized and  excited  states  comes  in. Then  the  students  will  most  likely  also transmit  information  to  fire  an  action  potential. While  the  downstate  once. Hyper  polarized  or  actively  inhibited, maybe  income  action  potentials will  not  result  in  action  potentials. So  it's  a  discretization of  information  processing  and contain,  this  has  consequences. So  if  you  have  neural populations  that  are  aligned  in  phase, they  can  talk  to  each  other  because  they  do two  circles  overlap  with  the  other  phase. They  cannot  communicate  effectively because  the  population  is not  listening  when  the  other  one  is  talking. So  that's  the  idea  of  communications  to  coherence. You  have  hierarchical  cluster  frequency, phase  amplitude  coupling, meaning  that  in the  more  excitable  state  of  slow  oscillation, amplitude  of  a  faster  oscillation is  increased  and  the  more excited  we  said  that  the  amplitude  of  oscillation. So  e.g.  delta  theta  gamma  coupling  coupling. So  it's  nesting  of  fast  and  faster  oscillations, which  always  increasing  amplitude  in  the  excited  state of  the  slower  oscillation. Or  more  sophisticated,  yes, like  getting  bypass  inhibition, which  proposed  for  alpha  oscillations, increasing  amplitude,  there  should be  increasing  inhibition  happening. Which  is  like  rhythmically suppressing  bottom-up  gamma  oscillations which  are  transporting  a  sensory  information. Upstream. Face  coding  phase  procession  in  the  hippocampus. This  red  navigation  experiments Bohr  postulated  also  alpha, gamma  phase  precision  when  scanning  with  the  scenes. And  something  that  we'll  come  back  to. Phase  amplitude  coupling  for  phase  dependent  plasticity. So  if  you  haven't,  let  say, slow  acylation  during  sleep  in Tupelo  was  upstate  in  a  sleep  spindle arrives  exactly  the  most  exciting  and  most  diploid  state. So  that  the  thalamic  cortical  input into  the  apical  dendrites  of this  cortical  neurons  when  the  maximally  excitable  than if  calcium  influx  and  can  produce  ATP  like  plasticity. If  they  would  arrive  to  a  different  time, different  phase  where  they  are  not  excitable, might  not  get  ATP  or  even  ADD  depression. So  we  use  a  technique  called  Nick, real-time  EG  triggered  team is  to  investigate  some  of  these  questions. Here  it's  about  pass  inhibition, which  has  been  shown  that many  hypothesized  for  the  visual  system. We  investigated  this  for the  alpha  of  the  sensory  motor  cortex. And  the  question  was,  is  it  actually  pulsed facilitation  inhibition  but  with  increasing  amplitude? If  increasing  inhibition,  but  there's  always  a  phase zero  inhibition  which  the  same as  if  there  will  be  no  oscillation  present. But  as  it  passed,  facilitation  would  increase the  amplitude  of  asymmetric  increase  in  facilitation. But  there's  always  a  phase  angle with  the  facilitation  is  as low  as  if  there  will  be  no  oscillation, or  is  it  even  symmetric? Boring.  So  what  we  did  is  we targeted  the  lower  20%  of  the  integral  and  we  Alpha. The  oscillation  D  Synchronized, not  present,  but  observable. Or  the  upper  20  per  cent  with a  full-blown  ice  oscillation. Analyze  data  in  real-time  and  they're  stimulated. And  random  phases  if  there's  no  oscillation. But  if  then  the  peak fallen  flank  trough  horizon  flank  randomized. And  we  measured  excitability  with motor  evoked  potentials  and  motor  cortex. And  we  used  a  program  called  short interval  intracortical  inhibition  is  ICI, which  is  second  conditioning  parts, but  low  intensity,  which  is inhibiting  the  second  day  repeated  the  second  part. Let's  assume  that  it's  gaba,  a  receptor  mediated, wanted  to  see  whether  there's  any  GABAergic  effects happening  or  if  its  parts  inhibition. So  that  the  targeting  work  out  well. We  do  see  the  alpha  peak  only  for the  high  power  trials  is  frequency  representation. It's  local  at  the  left  sensory  motor  cortex, but  we  want  to  target  based  on  our  EG  montage. It's  not  some  occipital  alpha  bleeding  in  its  local  one. And  most  importantly,  well, the  average,  we  don't  see any  oscillation  in  the  random  condition, but  we  do  see  targeting  to  peak  and  trough, rising  flank,  where  we target  those  specific  phase  angles. So  there's  processing  has  to  happen fast  to  be  able  to  target  them. And  what  we  found  is  that  there's actually  not  a  single  phase  angle. Well,  we  do  see  an  inhibition  relative  to our  randomized  low  amplitude  condition, but  there  traveled  rising  flank, but  we  have  a  facilitation. So  each  time, like  once  the  cycle  there's  a  facilitation  from  baseline. So  we  can  investigate basic  neurophysiological  questions  with  this  technique. There's  no  change  in  SSI. So  it  looks  like  path  facilitation  in  the  motor  cortex. We  measured  with  MEPs, but  be  aware  that  our  EG  montage is  very  sensitive  for  action  in  the  postcentral, the  somatosensory,  we  offer  waste  generated. So  it's  maybe  that  somatosensory  cortex. We  do  have  extra  parts  inhibition, which  results  in  this  inhibition once  per  cycle  of  the  motor  cortex. And  we're  running  status at  the  moment  to  figure  that  out. My  colleagues  from  tubing  have  showed this  basic  difference  between  pics  in  traffic first  exploded  it  by  repetitively targeting  either  the  less  excitable  peak  or  the  trough. In  one  session  with  triplets  of  100, 200  timess  again  and again  and  again  and  again  in  the  trough. And  if  you  hit  the  mark  set  again  and  again, then  afterwards  you  have  an  increase, increase  in  the  excitability. But  that's  not  the  case  if you  have  the  same  amount  of  stimuli, same  intensity,  but  targeting a  different  state  of  the  oscillation. This  phase  dependent  plasticity, which  could  potentially  be used  also  for  different  oscillations in  different  parts  of  the  cortex. And  which  makes  sure  that  we  actually do  have  this  fancy  pen  plasticity  mechanism  in  the  brain. And  we  can  mimic  it  with this  phase  dependent  plasticity  protocols. This  is  the  first  study  of  this  kind, which  you  did  as  a  PhD  student  actually when  Michael  was  visiting  in  2009, we  figured  out  in here  where  did  this  with  help  of  Sidner, basically  recording  this  data  person  in 2012  while  ago  that  were  targeted. This  legislation.  It's  low, so  it's  an  easy  target. So  the  methods  available, but  then  it  was  possible  in  wet  lacing. Upstairs  and  downstairs,  half  of  them  were  stimulated. Half  of  them  will  just  set the  market  and  did  not  stimulate. So  we  did  the  student  sleep  and this  fairly  comfortable  upright  position  had fixated  new  navigated  noise,  Martin  on  the  ears. And  these  are  the  domestic  evoked  potentials  were  in  red, upstate  targeted  in  blue  down  to  target  it. And  you  see  that  the  average  evoked  potential is  one  of  the  source  lesion that  we  targeted  plus  something  that  rides  on  top  of  it. So  we  have  to  subtract  the unstimulated  twice  and  then  we'll get  you  the  different  scaling. The  Timaeus  evoked  potential. Alright? Which  looks  very  different  than  the same  stimulation  during  wakefulness, is  basically  again,  it's legislation  stimulate  the  brain  during  sleep. It  reacts  with  the  oscillation it  can  produce  in  that  state. Anyway.  You  want  to measure  the  multiple  potentials again  and  bulging  slip  compared  to  wakefulness, there's  general  production  and excitability  which  could  be  at  the  spinal  level. But  it  was  also  systematically different  between  the  up  and  down  stage. So  during  the  hyper  polarizing  phase, it  was  smaller  than doing  the  deployment  phase  was  larger  for  all  subjects. Evoked  potentials  also  showed an  effect  at  the  simulation  site  locally  was  larger. It,  you  respond  to  me  stimulated  during  the  upstate. And  also  when  spreading  flex those  nations  do  over  the  entire  cortex  was  still  larger. When  was  it  upstate?  At  the  time  of  stimulation? This  is  not  an  all  or  nothing  phenomenon, but  we  could  actually  do  signatory  correlations. So  looking  at  the  slow  oscillation  amplitude at  the  time  of  stimulation, which  predicted  the  MEP  amplitude, but  only  from  channels  at the  stimulation  side  of  the  TPS. What  kind  of  signal  analysis the  signal  to  noise  ratio  matter. But  you  can  pin  them  and  then  you  see  the  same  thing. At  when  there's  a  larger  ongoing  oscillation, then  the  TP  is  also  larger. Now,  ten  years  later, we  move  Trump's  slow  oscillations  to  spirit. Son.  Now  former  PhD  student  MOLAP, managed  to  detect  spindles  in  real  time. Now  this  term  is  referring  to,  but  in  the  bed, again,  you  navigated  targeting splendid  free  periods  during  sleep. Different  phase  angles  of  the  spindle. And  then  after  the  spinner. And  again,  targeting  worked  out  as planned  with  stimulating  spindles  and  not some  artifacts  are  adjacent  frequency  bands. And  due  to  our  montages  were  to  fill  local. Ever  see  here  is  similar  pattern  but  different. Again,  like  that. Traffic  and  rising  that  are  comparable, the  largest,  but  actually not  different  from  the  baseline. And  it's  been  a  few  time  periods,  began  to  falling  flank, that  is  the  lowest,  but  this  time, this  is  the  outlier. So  it's  a  post  inhibition happening  once  per  cycle  to one  specific  angle  of  the  spindle. Which  means  we  can  now  in  the  future, projects  planned  want  to  repetitively  stimulate  doing the  spindle  at  a  time  when  it's  most excitable  and  produce  plasticity, porcelain  motor  cortex  to  spin  it's opening  a  window  of  plasticity  in  the  brain, which  could  be  a  very  interesting  approach  also for  habilitation  is a  totally  unused  eight  hour  time  window. The  brain  is  doing  a  lot  of  maintenance, a  lot  of  consultation. So  two  mechanisms  utilize  it. The  poppy,  very  interesting  because  we  also plan  to  study  memory  reactivation, model,  level  of  motor  learning,  or  declarative  memory. So  a  couple  of examples  of  how  we can  use  for  different  installations  this  approach, using  TMS,  EEG  to  MEPs  as  a  measurement. Using  the  EEG  to  inform us  about  a  particular  brain  state, in  this  case,  amplitude and  phase  of  an  ongoing  oscillation. But  you  can  think  about  more complex  protocols  in  the  future, more  complex  brain  states. So  we've  already  made the  tradition  from  open  loop  prints  it independent  protocols  to open-loop  brain  state  dependent  protocols. It's  still  not  completely  closed because  alpha  or  slow  oscillation  spindles, then  we  measure  the  MEP, but  we  do  not  change  the  spindle  necessarily. At  least  it  does  look  like  that. Close  loop  would  be  when  we  change the  variable  we're  monitoring. That  we  can  change brain  stage  from  an  undesired  to  desired, from  dysfunctional  to  functional  brain  state and  maintain  it.  And  we're  not  there  yet. We  have  done  some  software  developments  to  help this  kind  of  research  are  facilitated. The  brain  electrophysiological  recording  and  stimulation. Tupac's  the  best  toolbox, open-source  MATLAB  toolbox, which  allows  for  a  lot  of  standard  protocols. I'm  finding  the  motor  hotspot,  automatic  closed  loop. So  detection,  input,  output  curves  everything but  also  all  the  brain  state-dependent ED  to  get  protocols, autism  teammates  if  my  synchronization  and  so  forth. So  it  helps  to  solve  some  of  the  technical  problems. But  some  can  measure in  real  time  and  do  the  calculations and  target  the  stimulator. But  the  thing  is, we  need  to  identify  those  brain  sits  in  real-time. What's  more  complex  ones? Connectivity,  entropy, some  sophisticated  network  analysis. Whatever  you  think  is  interesting  as  a  brain  state. In  the  case  of  Egypt  were  getting  removed the  artifacts  in  real  time. That's  the  tricky  part  you  want  to  simulate. Maybe  each  cycle  of  oscillation is  possibly  at  the  moment. And  the  most  difficult  of  all. How  do  we  change  the  brain  state? Which  part  of  the  network  do  we  need to  stimulate  when  and how  to  shift  from  one  state  to  the  other. Dysfunction  to  a  function. How  do  we  keep  it  there?  So  maybe control  theory  is  going  to  help  us. But  this  is,  I  think  the  biggest  problem if  all  the  technical  thing  is  sort  of  like, how  do  we  change  the  brain? How  do  we  make  it  behave  like  we  would  like? So  this  is  the  biggest  problem. So  actually,  I  hit the  45  min  and  I  haven't  talked  about  ultrasound. If  you  find  it,  add another  5  min  at  least  to  give  you  the  rough  idea, not  going  into  all  the  details. I  have  more  slides  if  you're  interested. I'm  happy  to  talk  about  it  afterwards. But  just  to  give  you  a  quick  idea, you  can  do  a  lot  of  different  things with  transcranial  ultrasound. It's  used  in  surgery  already for  ablation  with  high  intensities, you  can  heat  up  the  tissue  45, 60  degree,  even  ablated. And  if  you  do  this  for  the  demo  SDN, you  can  treat  basically  tremor  with the  subject  walking  in  and out  afterwards  without  opening  the  scar. But  it's  ablation.  Or  use a  low  intensities  with  some  tasty  injected  micro  bubbles. You  can  open  the  blood-brain  barrier. That's  typically  a  good  thing. It  keeps  stuff  out of  your  brain  that  shouldn't  be  in  there. But  it  also  prevents important  pharmacological  treatments  to enter  the  brain  because  molecules  are  too  large, they  can't  pass  from the  bloodstream  into  the  brain  tissue. You  have  these  micro  bubbles and  exciting  ultrasound,  cavitation. They  oscillate  in  size  and  press  against the  tight  junctions  of  cells and  then  opens  them  transiently  and  injected  molecules, pharmacological  compounds,  or  wherever  you  can actually  get  from  the  bloodstream  into  the  brain. With  a  grandmother  actually  explored  this in  mice  and  also  in  humans. Then  your  modulation,  which  is  more  work  to  do. And  I'm  appealing  to  all  these  other  fancy  things. Typically,  haiku  setup  with  high-intensity is  intensely  focused. Ultrasound.  This  is  what  you  would  do  for  ablation. We  keep  the  temperature  below  one  degree  in  the  brain. So  it's  not  a  terminal  effect, most  likely  a  mechanical  effect  where the  ion  channels  in  the  membrane. We  have  applied  the  treatment  in  the  past  fashion. So  go  into  details  here, but  not  too  much  energy  to  be  deficit  at  the  same  time, not  too  much  warming  to  happen. We  only  give  it  a  certain  amount  of  time, like,  let's  say  like  30%  of  the  time, it's  positive  for  some  thousand  hertz. Then  the  protocols  can  be  created, which  actually  opens  up  issues  parameter  space  at the  moment  it  means  you  try  to  figure out  which  parameters  to  what, how  did  we  get  excitatory  or  inhibitory  protocols. So  we're  not  there  yet. Depending  on  the  type  of  constructor gets  smaller  or  larger  foci, we  can  control  them  if  you  have  multiple  elements, can  steer  them  in  distance,  e.g. that's  the  simplest  system  that  we're  using, which  is  the  four  elements  Angela  array, which  we  can  steer  between superficial  and  key  brain  regions, but  just  in  a  straight  line. If  a  cabinet  with  some  ultrasound  gel, which  is  tricky  because  it needs  like  doubt  coupling  to  get  from the  transducer  to  the  bone. That's  the  tricky  part  into  the  brain. Navigation  is  key  because  we  told  it  a  little  bit, then  an  eight  centimeter  depth might  be  millimeter  wide  beam. We  might  just  be  off.  Cb.  Try  some  robots based  navigation  tool  and  use  it  in  the  MRR directly  or  to  the  right  of  light  were  to  individualize. That  doesn't  attach  it  to  the  head. Then  you  have  to  do  acoustic  simulations based  in  bone  imaging, where  the  beam  is  actually  ending up  in  tenure  position  accordingly. Bone  imaging  is  key,  city  would  be  optimal, but  healthy  participants, we  can't  do  that  because  that's  the  radiation. So  we  have  to  use  MR, which  is  not  optimal  for  bone  imaging better  to  improve  that. Then  of  course,  as  always, we  would  combine  it  with other  techniques  for  proof of  target  engagement  with  EEG  to see  whether  we  can  change  oscillations with  FMI  just  seem  to  get with  the  bold  response  modulation  of  task-related  board. And  with  TMS  and  the  motor  cortex. Just  because  you're  always started  the  motor  cortex,  right? To  use  MEPs  to  measure  excitability  of  motor  cortex. While  it's  only  kidding.  And  modulating  the  excitability. Just  want  to  go  from  r, r  timess  ultrasound  transducer, large  coil  because  it's  so  far away  from  the  head  node  because  it's  looped  in Susa  to  publication  shown that  there's  a  suppression  of motivic  potentials  if  you  start  sonic  aiding, a  few  hundred  milliseconds  before  the  TMS  bugs. Long  story  short  like  two  years  later,  two  labs, labs  working  on  that.  Love  different  experiments. This  is  now  archives  and  soon  to  be  resubmitted. This  oppression  was  an  artifact.  Unfortunately. So  ultrasound  is  an  audible  per  se  but  definition. But  if  you  pass  it  1,000  hz, then  you  hear  thousand  hertz  tone. And  it's  conducted  by  the  bone. And  it's  in  a  weird  way. You  might  have  shear  waves  in  the  bone,  etc. And  then  you  hear  a  sound that's  located  somewhere  in  your  head. And  it's  really  difficult  to  mask. We  tried  masked  bone conduction  headphones  and  everything. But  not  go  into  this  detail. Because  they've  been  left  to  try  to  control  sites. The  white  matter  instead. Stimulating  only  with  noise, no  ultrasound,  and  always get  this  suppression  of  the  MVP. So  in  that  case, not  a  transcranial  effect. The  animal  work  is  pretty,  pretty  clear, promising  some  good  work  in  humans  now. But  we  still  have  to  figure  out the  right  parameters  and  control these  perfect  co-stimulation  conference by  ramping  palaces, by  masking  noise  to  make  sure  that  we  are  actually stimulating  the  extra  measuring only  through  brain  stimulation  effects. Okay,  and  here  I  forgot  a  bit  over  time, but  I  wanted  to  show the  exciting  ultrasound  stuff  we  started  to  get  into, different  from  the  electric  and  magnetic  stimulation. I  hope  that  I've  convinced you  that  none  of  us  is  when  the  stimulation  is a  super  powerful  tool  and  a  diverse  tool to  not  only  like  a  neuromodulator, but  to  also  just  perturb  in measure  to  characterize  brain  networks. In  combination  with  fMRI  in  connection  with  EEG, we  can  read  out  particular  characteristics  of these  networks  and  study brain  function  in  health  and  disease. And  particularly  this  simultaneous  combination, which  is  a  bit  of  extra  effort, but  it's  the  lowest  you to  look  into the  actual  neural  effects  of  the  stimulation. And  not  just  hope  that  it  would  affect  the  brain. The  position  we  put  two  timess  squared. So  I'm  thank  all  my  collaborators  and  funding  agencies, of  course,  and  thank  all  of  you  for  being  with me  here  for  a  few  days  to  this. A  lot  of  time  for  questions,  actually. Thanks  a  lot. And  I  should  mention  a  postdoc  PhD  positions  to  fill. So  if  you're  interested  or  know  someone  who  might be  let  me  know. Should  ought  to  be  on  the  website.  Yeah.  Thank  you. Wow,  that's  not  a  great  talk. I  have  a  lot  of  questions,  but  I'll, I'll  start  with  just  one. You  showed  with  the  Alpha, the  pulse  inhibition  and  the  L  with  alpha, sorry,  not  pulse  inhibition. You  asked  whether  it's  posted  inhibition or  pulse  facilitation. And  then  it's  facilitation, at  least  for  the  motor  evoked  potentials. Do  you  think  that  generalizes to  sensory  processing  and  alpha? I  well,  I  don't  think  so. I  think  the  reason  why  we  do  see  facilitation  or maybe  disinhibition  is  that we  measure  from  the  motor  cortex, but  that  oscillation  might  be  extracted  from the  sensory  cortex  with  one  study. Actually  complete  it,  but  not  it.  Analyze  that  part. We  also  did  just  somatosensory evoked  potentials,  EG  potentials, we  made  enough  stimulation  to  the  peaks  and troughs  of  the  Mu  Alpha  to  see  whether maybe  the  incoming  afferent  input  into the  somatosensory  cortex  that might  be  modulated  in  the  opposite  way. So  whether  there's  actually  palettes  inhibition that  during  the, the,  the  trough  that might  be  maximal  inhibition  happening. And  this  much  of  inhibition  causes then  a  reduction  of  the  sensory  to  motor  inhibition, so  to  disinhibition, erythematous  inhibition  of  the  motor  cortex. But  this  speculation,  so  we  don't  have  the  data  yet. Existing  data  on  actual  detection  performance, sensory  detection  performance,  and  its  relation to  mainly  alpha  power. But  it's  very  mixed. So  there's  positive, this  negative  linear  relationship,  inverted,  U-shaped. But  the  face  has  not  been  investigated. So  that's  the  first  time  we're  doing  it.  Yeah. Thank  you  for  the  talk. I  have  just  one  question. You  mentioned  that  you  didn't  use a  noise  masking  during the  TMS  experiments  that  you  present. Right  now. The  wrong. So  we  did  use  Lloyds  market  for  Basel. Use  them  for  all  these  experiments. They're  the  most  important. For  TP  measurements,  if  you  want  to  measure  the  US  bonds. Because  there's  quite  some  overlap with  auditory  evoked  potentials and  which  also  produce  earlier  but  obsolete components  in  100  P2  you  also  see  for  the  TP. So  you  would  need  to  have  successful  noise masking  to  be  reassured  that what  you're  measuring  is  not  an  auditory  conform. The  problem  is  hardly  ever  get  this  fully  masked. Bone  conduction.  Depends  on where  you  stimulate  some  central  sites. You  can  actually  stimulate  with  some  quotes that  pop  up  muscle  that  you  can  hear  it. Quite  often  it  is  audible. And  then  switch  tricky  to  interpret. The  earlier  components  are easier  to  interpret  than  the  later  ones. We  use.  Most  of  these  experiments, you  MEPs  as  the  measurement.  They're  not  affected. Actually,  there  is  the  work  by coupled  to  you  at  all,  2021. It's  human  brain  mapping, probably  the  journal  that  shows  that MEP  is  affected  by  the  clique, by  the,  by  the  TMS. Click  is  a  recent  work,  but  not  the, not  the  single-pass  every  piece  of  pathos  logos. I  could  imagine  that,  especially  for  the  long  interval, reportedly  inhibition  for  the  LLC, I  recall  the  first  part  is  100  milliseconds  before. I  do  believe  that  there's  an  auditory  effect. But  you  can  measure the  default  is  running  down  the  spinal  cord. If  you've  implanted  electrodes  like milliseconds  after  the  domestic  bonds. At  that  time.  There's  no  way any  auditory  information  can  if  reach the  motor  cortex  to  have  an  effect. So  I  think  for,  for  senior  class  in  the  piece, we're  fine  for  PEP  has,  but  of  course  I  agree. There  may  be  issues.  Okay. Thanks. I  figured  out  how  many  it  is  loud  and I  feel  like  I've  many  times  to  ask  you, but  I  feel  like  with  everyone  here, I'll  just  ask  a  big  picture, too  big  picture  questions. One,  your  experiments  are  very  elegant  and  rigorous and  to  the  point  of  almost  challenging  to  extract  the, the,  the,  where  you  take  the  information  from  there. And  you  did  make  some  mention  of  next  steps. But  could  you  talk  a  little  bit  about  where  you  see these  approaches  when  it  comes  to maybe  designing  interventions,  e.g. and  you  kind  of  alluded  to  it, but  maybe  for  those  who  are  trying  to  figure  out what  you  could  do  with  this  type  of elegant  brain  stimulation  imaging  as  it  relates  to maybe  clinical  conditions  or  something  along  those  lines. Yeah.  I  mean, so  I  think  in  basic  science, we  have  to  first  try  and  figure out  how  we  can  optimize  these  approaches  maximally, be  as  specific  as  possible  without  all  the  confounds. And  then  test  with a  which  steps  of personalization  of  state  dependent  stimulation, increasing  specificity. Which  of  them  are  important? Which  not  so  much  an  effect  size will  actually  change  if  you  do  that  or  not,  right? Especially  when  it  comes  to  clinical  use. Of  course,  it  will  need  to  pay  off  right  eye. E.g.  in  this  work, we're  not  trying  to  optimize  these  individual targeting  for  deep  brain  indirect  deep  brain  stimulation. Whether  the  plants  were  to  run  this, we  submitted  crossing  fingers  that  we  get the  funding  in  patient  with  depression, which  the  accelerated  treatment  protocol  and the  same  protocol  is  few  days  of  that  with  TMS  applied before  and  after  our specific  orientation,  optimization,  etcetera. And  then  to  also  the  treatment  with  and without  this  extra  modulation. Because  they  don't,  of  course  asked  me  for  good  reasons, do  we  actually  need  to  do  this? Decreases  the  likelihood  that we  actually  use  it  in  the  clinic. It's  a  fair  point.  We  have  to  check  with  it. It's  worth  it.  If  it's  a  significant  effect. In  terms  of  effect  size. If,  if  more  people  that  respond  to  the  stimulation because  you  are  on  target  in  each  individual  and not  just  30  per  cent  of  them. Right  then  it  might  be  worth  it. Especially  because  all  you  need is  a  structural  MRI  and  a  10-minute  resting  state  scan. And  the  rest  is  processing  that  can  be  streamlined. They  can  put  into  the  Cloud  that, that  could  be  done  quickly,  eventually,  right? But  firstly,  to  figure  out whether  it  has  a  benefit  or  not. So  I  always  like  to  say  the  study from  that  exactly  looks  into, okay,  what  is  the  benefit of  using  more  precise  targeting  for? Interference  with  the  mental  rotation  task, which  that  well  established  in  the  lab. And  they  use  either  dependent  on  the  right,  the  cortex, either  for  an  individual  functional  fMRI  localizer to  find  exactly  the  right  spot  in  each  individual. Or  when  for  anatomically  coordinates  are  just MNI  coordinate  or  a  neutral  position  P4. That's  what  you  would  do  if  no  clue  if  I'm  MRI, just  measure  and  assume  the  left  motor  cortex  before. That's  the  parietal,  then  the  criterions. Okay,  how  many  subjects  do we  need  to  get  a  significant  effect? There  was  like  five  for  the  half-mile  localizer. It  was  like  40,  45. If  you  just  went  for  a  neutral  position before  and  then  you  can  do  the  calculations. Is  it  worth  equation so  many  more  subjects  to  get  you  effect? If  you  have  the  chance  to  pay  for  an  MRI. If  it's  about  political  tripping, of  course,  then  Christmas. Is  it  ethical  to  do  such  a  bad  targeting? Understand  that  you  want  to  make  their  treatment available  for  many  people. But  if  it  only  works  in small  percentage  because  most  of  the  time  you are  off  target  for  this  simulation, then  I  would  say  it's  not  really ethical  to  do  that,  right? So  it's  a  trade-off  and  it  depends  on  the  case. So  long  answer.  There's  a  question  on  Zoom  from  Cyrus. Cyrus,  if  you  want  to  unmute  yourself, then  you  can  feel  free  to  ask. Yeah.  Can  you  hear  me?  Yes. Yeah. So  I  was  thinking, is  it  possible  to  after  you  take the  CT  and  have  a  image  of  the  brain  that  you can  kind  of  modulate  the  ultrasound  so  it  targets a  big  area  such  as  the  default  mode  network  and then  better  prepared  stuff,  specific  region. Yeah,  that's  a  very  good  point because  you  are  asking  actually for  a  less  focused,  focused  ultrasound,  right? And  that's  exactly  the  point of  discussion  with  some  colleagues  recently because  the  big  benefit  of  it, so  focus,  we  can  stimulate  the  amygdala  or  small. In  principle,  if  get  the  targeting right  and  the  bone  imaging  can target  small  sub  nuclei  eventually  anywhere. But  if  we  wanted  to  estimate  the  critical  region, is  this  small  focus  going  to  make  an  effect  or  not? It  may  even  be  that  our  output  measures  like  an  EEG. We  might  not  see  an  effect  because the  new  population  will  be  synchronized  so  small. So  in  principle,  we  would,  one  could  not. The  default  mode  network  is  a  large  area, right?  Yeah,  it's  huge. And  as  multiple  sites  with  media,  right? Because  you  could  submit  multiple  regions  and you  could  use  like  for transducers,  stimulate  different  sites. But  each  of  them  is  large. So  each  side  would  need  larger  simulation. So  far  there  is  no  mighty,  mighty  transducer  work. I  know  of  colleagues  playing  that  at  the  moment. So  it's  definitely  going  to  happen. But  if  we  wanted  to  make  the  beam  larger, the  focus  larger,  the  problem  is  that the  energy  is  dissipated  distributed  over  large  areas. So  it  would  need  more  energy  that passes  through  the  bone  and  then  you  heat  up  the  bone. What  about  if  you  cover  a  large  area with  several  small  focused  region? Select  a  layout  like  a  carpet  of perhaps  20  different  stimulations  all  over. Cortex. Yeah,  that's  a  good  idea. And  it's  also  what  e.g. insert  electrode  is  M  Israeli  company, 1,024  transducers  over  like  a  hemisphere. You  have  to  couple  them  or  to  the  skin, which  means  taking  those  subjects  here, the  hair  as  V-shaped, which  is  fine  if  you  go  for  one  treatment  to  get your  central  tremor  basically  ablated  and you  go  home  and  to  regrow your  hair  if  you  want  to  go  like repeated  times  for  neuromodulatory  sessions for  treating  depression  or  whatever, you  might  not  want  to  shave  your  hair. So  then  the  captain  gets  really  difficult. Disadvantage  of  large  transducer  areas  is  the  coupling. But  I'm  sure  that  the  feet  will  work  this  out. So  it's,  it's  very  young  field. And  the  good  thing  is  like  based  on  the  physics, we're  far  away  from  the  ceiling,  right? And  so  it's,  a  lot of  engineering  questions  will  be  solved. Does  actually  good  place  to  be here  to  pitch  some,  some  of  those  problems. Thank  you  so  much. Interesting  talk.  And  I think  that  was  all  of  our  time  for  questions, but  one  more  round  of  applause  for  Dr.  Berg.  Thank  you. Thank  you  so  much.