The Science of DreamsPresentation

byCharles Beaman

MD/PhD StudentUT Health

Are Dreams Important?2 Nobel Prizes

Otto Loewi Chemical Transmission of

Nerve Impulses

Neils BohrStructure of Atoms and

Quantum Mechanics

How Do We Measure Dreams?

How Do We Measure Dreams?St

ages

of S

leep

Two Types of DreamsSt

ages

of S

leep

NREM

REM

Alpine Racer 2

Dreaming in RatsNow we can actually LISTEN to the Neurons in the brain

Dreaming in Rats

Rats in the Maze

Dreaming in Rats

NREM Dreaming

REM Dreaming

• Compressed in Time Scale (1 sec of dream equals 10 of reality)• Practice Learned skills

• Expanded Time Scale • Simulations?• Testing Future Possibilities

Why Do We have Nightmares?

How Does My Lab Measure Dreams?

Electrocorticoraphy

• Intractable Epilepsy Patients

• Patients in the Hospital for about 1 week

• We can use this time window to study Sleep

Local Sleep

The Dream-Reading Machine

Neural Decoding of Visual Imagery During Sleep

T. Horikawa, M. Tamaki, Y. Miyawaki, Y. KamitaniATR Computational Neuroscience Laboratories, Kyoto, Japan

Task Design

Awoken every 5-6 minutes

Outline of Sleep (Nap) Experiment

• 3 subjects• 1 pm to 5:30 pm• fMRI scans + EEG, EOG, EMG, and ECG• Usually last 90 minutes over 7 days > 200

awakenings with visual report• Subject awakened after single epoch of alpha-

wave suppression and theta-wave (ripple) occurrence (Stage 1 sleep)

Success with Awakening at Appropriate Time

235 awakenings

198 awakenings

186 awakenings

Example of Verbal Reports

Reports lasted 34 +- 19 seconds

They also collected a “Vividness” and subjective timing of each event, but did not use this data

Non-visual reports were classified as: thought (active thinking), forgot, non-visual, and no report

Visual Content Labeling - WordNet

Based on Synonymy – 117,000 “synsets” that are sets of related words

They assigned all reports to synsets.

Base Synsets – Common, specific

semantically exclusive and specific

Visual Stimulus Experiment

• Used ImageNet – 240 images per base synset

• Placed in center of screen, subjects freely viewed images without fixation

• fMRI recorded for each base synset

• 9 second stimulus block, 6 images sampled from one synset, .75 s with .75 s interleaved blanks

• Followed by 6 s rest period

• ~40 blocks per base synset were recorded

Area of Brain Studied• Higher Visual Cortex – ventral region covering

lateral occipital complex, fusiform face area, and parahippocampal area (1000 voxels)

• Lower Visual Cortex – V1 to V3 (1000 voxels)• Subareas (400 voxels)

Pairwise DecodingBinary classifier was first trained on fMRI data to 2 base synsets, then tested on sleep samplesContaining exclusively 1 of the 2 synsets

Multilabel Decoding

Videos

• http://www.sciencemag.org.ezproxyhost.library.tmc.edu/content/suppl/2013/04/03/science.1234330.DC1/1234330s1.mov

• http://www.sciencemag.org.ezproxyhost.library.tmc.edu/content/suppl/2013/04/03/science.1234330.DC1/1234330s2.mov

Sleep Types