Mining the mind: Machine learning in brain research Matthias Treder - PowerPoint PPT Presentation

Introduction ML in BCI ML in brain research Ethics Mining the mind: Machine learning in brain research Matthias Treder 2016-12-16

Introduction ML in BCI ML in brain research Ethics

Introduction ML in BCI ML in brain research Ethics M EASURING BRAIN ACTIVITY image taken from: Astrand E, Wardak C and Ben Hamed S (2014). “Selective visual attention to drive cognitive brain–machine interfaces: from concepts to neurofeedback and rehabilitation applications” Front. Syst. Neurosci. 8:144

Introduction ML in BCI ML in brain research Ethics BCI = MACHINE LEARNING + REAL - TIME NEUROIMAGING

Introduction ML in BCI ML in brain research Ethics BCI = MACHINE LEARNING + REAL - TIME NEUROIMAGING

Introduction ML in BCI ML in brain research Ethics BCI = MACHINE LEARNING + REAL - TIME NEUROIMAGING

Introduction ML in BCI ML in brain research Ethics BCI = MACHINE LEARNING + REAL - TIME NEUROIMAGING

Introduction ML in BCI ML in brain research Ethics C ORRECTING ERRORS IN MENTAL TYPEWRITERS Schmidt, Blankertz, Treder (2012), BMC Neuroscience

Introduction ML in BCI ML in brain research Ethics C ORRECTING ERRORS IN MENTAL TYPEWRITERS Schmidt, Blankertz, Treder (2012), BMC Neuroscience Stimulation C A F e m T i

Introduction ML in BCI ML in brain research Ethics C ORRECTING ERRORS IN MENTAL TYPEWRITERS Schmidt, Blankertz, Treder (2012), BMC Neuroscience Classi fi cation Stimulation output correct C A C F error e m T i

fi fi Introduction ML in BCI ML in brain research Ethics C ORRECTING ERRORS IN MENTAL TYPEWRITERS 20 error correct [ µ V] 10 0 − 100 0 100 200 300 400 500 600 700 800 900 1000

Introduction ML in BCI ML in brain research Ethics C ORRECTING ERRORS IN MENTAL TYPEWRITERS 20 error correct [ µ V] 10 0 − 100 0 100 200 300 400 500 600 700 800 900 1000 Accuracy ROC classi fi er A ROC classi fi er B 1 1 1 0.9 0.9 0.9 0.8 0.8 0.8 0.7 0.7 0.7 0.6 Hit rate 0.6 Hit rate 0.6 AUC gbo / auc: 0.83 gbo / auc: 0.97 bad / auc: 0.85 bad / auc: 0.97 0.5 0.5 0.5 iae / auc: 0.75 iae / auc: 0.96 gbq / auc: 0.62 gbq / auc: 0.96 0.4 0.4 0.4 gbt / auc: 0.8 gbt / auc: 0.93 0.3 0.3 0.3 iac / auc: 0.83 iac / auc: 0.91 gbn / auc: 0.82 gbn / auc: 0.87 0.2 0.2 0.2 gbw / auc: 0.87 gbw / auc: 0.84 iau / auc: 0.69 iau / auc: 0.79 0.1 0.1 0.1 ClassifierA mk / auc: 0.69 mk / auc: 0.75 ClassifierB fat / auc: 0.95 fat / auc: 0.99 0 0 0 Mean 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Participant False alarm rate False alarm rate

Introduction ML in BCI ML in brain research Ethics ML IN MEMORY RESEARCH

Introduction ML in BCI ML in brain research Ethics ML IN MEMORY RESEARCH Pre- 22 92 13 34 encoding 83 37 40 49 6 77 2 49 12 Encoding later remembered later remembered later forgotten later forgotten Post- 10 72 encoding 31 71 84 18 10 33 98 70 7 89 23 ? ? Retrieval remembered forgotten remembered forgotten

Introduction ML in BCI ML in brain research Ethics ML IN MEMORY RESEARCH

Introduction ML in BCI ML in brain research Ethics N EUROETHICS ◮ Cellular, molecular, cognitive neuroscience

Introduction ML in BCI ML in brain research Ethics N EUROETHICS ◮ Cellular, molecular, cognitive neuroscience ◮ Incidental findings in healthy participants

Introduction ML in BCI ML in brain research Ethics N EUROETHICS ◮ Cellular, molecular, cognitive neuroscience ◮ Incidental findings in healthy participants ◮ Altering brain function (eg deep brain stimulation)

Introduction ML in BCI ML in brain research Ethics N EUROETHICS ◮ Cellular, molecular, cognitive neuroscience ◮ Incidental findings in healthy participants ◮ Altering brain function (eg deep brain stimulation) ◮ Brain enhancement

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: E XPECTATIONS ◮ Traumatic brain injury: impaired capacity to judge costs/benefits

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: E XPECTATIONS ◮ Traumatic brain injury: impaired capacity to judge costs/benefits ◮ Motor skills automatic, BCI operation requires sustained attention and cognitive effort

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: E XPECTATIONS ◮ Traumatic brain injury: impaired capacity to judge costs/benefits ◮ Motor skills automatic, BCI operation requires sustained attention and cognitive effort ◮ High expectations due to media bias

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: E XPECTATIONS ◮ Traumatic brain injury: impaired capacity to judge costs/benefits ◮ Motor skills automatic, BCI operation requires sustained attention and cognitive effort ◮ High expectations due to media bias ◮ Psychological harm: distress, loss of behavioral control

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: E XPECTATIONS ◮ Traumatic brain injury: impaired capacity to judge costs/benefits ◮ Motor skills automatic, BCI operation requires sustained attention and cognitive effort ◮ High expectations due to media bias ◮ Psychological harm: distress, loss of behavioral control ◮ Selection: Equal opportunity for all vs failed expectations

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: R ESEARCH VS CLINICAL PRACTICE

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: R ESEARCH VS CLINICAL PRACTICE

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: R ESEARCH VS CLINICAL PRACTICE Improvement: proof-of-concept and parameter tuning with volunteers, transfer to patients image taken from: McCullagh P, Lightbody G, Zygierewicz J, Kernohan W. (2014). “Ethical challenges associated with the development and deployment of brain computer interface technology.” Neuroethics 7:109–122.

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: B ENEFITS AND RISKS Trade-off: Levels of invasiveness ◮ EEG (noninvasive): e.g. motor cortex signals smeared

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: B ENEFITS AND RISKS Trade-off: Levels of invasiveness ◮ EEG (noninvasive): e.g. motor cortex signals smeared ◮ ECog (epidural implant): risk of infection and hemorrhage

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: B ENEFITS AND RISKS Trade-off: Levels of invasiveness ◮ EEG (noninvasive): e.g. motor cortex signals smeared ◮ ECog (epidural implant): risk of infection and hemorrhage ◮ Microelectrode array (subdural implant): risk of tissue death

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups ◮ Minimally conscious: residual awareness

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups ◮ Minimally conscious: residual awareness ◮ Locked in: conscious but lack capacity for voluntary body motion

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups ◮ Minimally conscious: residual awareness ◮ Locked in: conscious but lack capacity for voluntary body motion ◮ Informed consent

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups ◮ Minimally conscious: residual awareness ◮ Locked in: conscious but lack capacity for voluntary body motion ◮ Informed consent ◮ ‘Talking’ via ML algorithm, rather than to the patient directly

Introduction ML in BCI ML in brain research Ethics E THICS IN BCI: C OMMUNICATION ◮ Patient groups ◮ Minimally conscious: residual awareness ◮ Locked in: conscious but lack capacity for voluntary body motion ◮ Informed consent ◮ ‘Talking’ via ML algorithm, rather than to the patient directly ◮ Caregivers might overestimate system (yes-no responses)

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML Privacy of thoughts ◮ unconscious intentions [Zander et al., Dec 2016, PNAS]

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML Privacy of thoughts ◮ unconscious intentions [Zander et al., Dec 2016, PNAS] ◮ psychological traits

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML Privacy of thoughts ◮ unconscious intentions [Zander et al., Dec 2016, PNAS] ◮ psychological traits ◮ neuromarketing

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML Privacy of thoughts ◮ unconscious intentions [Zander et al., Dec 2016, PNAS] ◮ psychological traits ◮ neuromarketing ◮ brain fingerprinting

Introduction ML in BCI ML in brain research Ethics M INING THE BRAIN USING ML Privacy of thoughts ◮ unconscious intentions [Zander et al., Dec 2016, PNAS] ◮ psychological traits ◮ neuromarketing ◮ brain fingerprinting ◮ ‘thought police’?