The neuroscience of WM capacity and training. Overview. WM relevance. WM in non-human primates

Transcription

1 Overview The neuroscience of WM capacity and training Neural basis of working memory Working memory training Transfer to attention WM, mathematics and development Torkel Klingberg Dept. Neuroscience Karolinska Institutet WM relevance WM in non-human primates WM capacity correlates with: Attention Non-verbal reasoning ability Mathematics Reading WM impairments ADHD Children born prematurely In children after cancer treatment Traumatic brain injury Fuster and Alexander, 97; Kubota and Niki, 97 Funahashi et al. 989

2 Visuo-spatial WM Adapted from Sprague et al. 24 Constantinides and Klingberg (26) Nat Rev Neurosci Human WM Attention and WM WM and attention overlap Verbal STM Verbal WM Visuospatial WM and STM Klingberg et al. 996; Wager and Smith 23; Curtis D Esposito 23; Linden 27 Ikkai and Curtis (2) Jerde et al. (22) Inattention in ADHD and WM Children with ADHD have impaired working memory Kempton et al. 999 Barnett et al. 2 Westerberg et al. 24 Willcut 25 Martinussen et al. 25 Lui and Tannock 27 performance Pingault et al. 2 Gathercole et al. e.g. 23, 24 Working memory training Klingberg et al. 22, 25 Cogmed. Visuo-spatial WM Jaeggi et al. 28 Dual n-back Dahlin et al. 28, Li et al. 28 N-back lists Chein and Morrison, 2 Complex WM tasks 2

3 Cogmed WM training (Klingberg et al. 22, 25) Randomized, double blind design Active control group Repeated performance with feedback (no strategy instruction) Adaptive difficulty close to capacity limit Mainly visuo-spatial WM tasks Intense and extensive à 35 min per day à 5 days/week à 5 weeks Treatment Adaptation of difficulty level Control Easy trials (level 2) TK has no associations with Cogmed or Pearson Klingberg et al. 25 Span-board Cogmed WM training transfer 5,5 items items 5 4,5 4 3,5 3 A 4,4 4,2 4 3,8 3,6 3,4 comparison treatment T T2 T3 Digit-span comparison Span-board Digit-span Visuo-spatial WM tasks Klingberg et al. 22; 25; Thorell et al. 28 Complex WM tasks Holmes et al. 29; Holmes et al. 2; Bergman-Nutley et al. 2; Brehmer et al. 22 Cross modal Thorell et al. 29 Instructions Holmes et al. 29; Bergman-Nutley and Klingberg (24) 3,2 treatment 3 T T2 T B Klingberg et al. 25 Improvement of WM capacity after training d =.6 Meta-analyses of WM improvements after training Dual aims of WM training Improvement of working memory in standard deviations Spencer-Smith 25 Melby-Lervåg 23 Schwaighofer 25 Rapport 23 Cortese 25 Peijnenborgh 25 Experimental tool to study brain plasticity associated with higher cognitive functions Potentially a useful intervention 3

4 Computational models of persistent activity as working memory Neural correlates of WM training Experiment fmri WM training Functional connectivity-> Higher firing rate and Increased capacity Compte et al. (2) Constantinidis, Wang (24) Edin et al. (27) JOCN Edin et al. (29) PNAS Olesen, Westerberg, Klingberg (24) Nature Neurosci. Dopamine and WM training DA BP change vs WM change 3 healthy, young adults 5 weeks of training (Klingberg et al., 25) Measurement of verbal and visuospatial WM fmri during performance of WM tasks PET of D and D2 binding, before and after D SCH2339 D2 Raclopride Linear model: y 2 y = (a + β x 2 ) (a + β x ) =β (x 2 -x ) Quadratic model: y 2 y = (a + β x 2 + β 2 (x 2 ) 2 ) (a +β x + β 2 (x ) 2 ) r 2 =.42; p=.6 r 2 =.75; p=. β estimate p =.2 β 2 estimate p =.5 Mean CTT score Preg Post 2 MEAN CTT McNab, Klingberg et al. Science 29 McNab, Klingberg et al. Science 29 Genetics of cognitive plasticity (Söderqvist et al. (23) J Cog Neursci DRD2 (rs ) AA/AG Training day Percentage change GG N = 25 Age = 6 6 Training 24.7 days (Söderqvist et al. (23) J Cog Neursci 4

5 DAT- Neural basis of WM training DRD2 D2 DAT- c d d Olesen et al. 24 Dahlin et al. 28 Bäckman et al. 2 Constantinides and Klingberg (26) Dual aims with WM training WM training in children with ADHD Parent +Teacher Inattention Parent +Teacher Hyp/Imp Experimental tool to study brain plasticity associated with higher cognitive functions Potentially a useful intervention Attention IQ Mathematics, reading, Change (SD),9,8,7,6,5,4,3,2, post-treat follow -up control treatment Change (SD),9,8,7,6,5,4,3,2, post-treat follow-up control treatment Effect size (Cohen s delta) inattention =.7 Klingberg et al. (25) WM training in old and young healthy adults Young 2-3 years; n=5 Old 6-7 years; n=5 Randomized, blinded, controlled WM training in ADHD RAST Children with ADHD (N = 26) Restricted Academic Setting Task Brehmer et al. 22 Green, Schweitzer et al. (22) 5

6 Children with ADHD (N = 66) Randomized WM training non-adaptive training BRIEF Effect of CWMT training on attention d =.4 WM and mathematics Working memory capacity is correlated with arithmetical performance Henry and MacLean 23; Kyttälä et al. 23; Maybery and Do 23; Alloway et al. 25, 29, Geary et al. 29; Meyer et al. 2 Working memory capacity predict future development of arithmetical ability Jarvis and Gathercole 23; Gersten et al. 25; Bull et al. 28; Dumontheil and Klingberg 22 Working memory and reasoning ability contribute partly unique variance to predict arithmetics Alloway and Alloway 2, Dumontheil and Klingberg 22 Spencer-Smith and Klingberg, 25 WM activity predicts mathematics Development of Math and WM predicts future math scores Dumontheil and Klingberg, 2 Schel and Klingberg (26) Cerebral Cortex 6

7 Effect of WM training on mathematics Combined WM and numberline training Author Participants N Effect size (d) Dahlin et al. 23 Inattention * Holmes et al. 23 Low achievers 75.6 * Holmes et al. 29 Low WM 22. ns Dunnings et al. 23 Low WM 6.3 ns Bergman-Nutley 24 Inattention 34.3 * Roberts 26 Low WM 452 ns Federico Nemmi, Elin Helander, Ola Helenius, Rita Almeida, Martin Hassler, Pekka Räsänen, Torkel Klingberg 3 7 Nemmi et al. (26) Combined WM and numberline training Working memory training Numberline training (NLT) Read/ Read/ Daily training: 3 min/day Read WM Duration: 8 weeks Read/ NLT NLT/WM Average training 38. (3.4) days About 9 h training Age: 6-year olds (8.3 months, SD=3.5) Group : Typically developing children (n = 2) Class-based training Group 2: Children with low WM (n = 98) Individual training Vektor cognitionmatters.org Tal Symbol Place on numberline Length (vector) Counting of object Addition Namn Efternamn April 4,

8 Arbetsminne Pre- and post measures Working memory Span-board forward Span-board backwards Grid task (visuo-spatial WM) Standardized and averaged to one composite WM measure Mathematics WISC verbal arithmetics Addition (without numberline) Subtraction (without numberline) Magnetic resonance imaging (n=58) functional MRI during WM performance structural MRI Standardized and averaged to one composite Math measure Results Math Performance post = β + β WMT + β 2 NLT + β 3 Math bl + β 4 WM bl + β 5 WMT x NLT + β 6 WMT x WM bl + β 7 NLT x Math bl + β 8 Cohort + β 9 Sex + β Population + β Age ε Results Math Performance post = β + β WMT + β 2 NLT + β 3 Math bl + β 4 WM bl + β 5 WMT x NLT + β 6 WMT x WM bl + β 7 NLT x Math bl + β 8 Cohort + β 9 Sex + β Population + β Age ε Math Improvement(res) Math Improvement and WM at Baseline 2 2 WM at Baseline F.WM 8

9 Math and WM related regions Red = predictive information all behavioral measures into account Data mining Universal language of mathematics Created accounts 9

10 Conclusions Future directions Working memory training improves working memory, e.g. remembering instruction d =.6 (.5-.7) improve inattention in everyday life d =.4 (.3-.5) effect for other tasks, i.e. mathematics, IQ, reading, grades, are still inconclusive Better measures of inattention Larger populations in training studies Combined training (WMT+ cog, TMS, drugs, physical, ) Individualized training Continued role of neuroscience 5 6 Fahimeh Darki Henrik Ullman Thanks! Rita Almeida Pernille Olesen Helena Westerberg Fredrik Edin Albert Compte Megan Spencer-Smith Hans Forssberg Federico Nemmi Christos Constantinides Namn Efternamn April 4, Predicting future WM Variable 2 Support vector regression Variable I. II. Predicting WM two years later (T2) Predictor r p t BOLD.52 <.*** r =.78*** t FA.58 <.*** t GMD.29 <.5* Combining MRI predictions in a multiple regression: Predictor p t BOLD <.5* t FA <.** t GMD ns. Model r =.64*** Ullman, Almeida, Klingberg (24) J Neurosci III. Combine MRI prediction with baseline (T) behavior

11 DRD COMT Fronto-parietal activity correlates with current WM capacity Klingberg et al. 22 Kwong et al. 22 Crone et al. 24 Basal ganglia-thalamus, and white matter structure predicts future WM capacity Ullman, Almeida, Klingberg (24) J Neurosci DRD2 DAT- Arithmetics and attention Knops et al. 29