Get best of the week

Memory archives: how the brain encodes and reproduces memories



On the one hand, the human brain is quite understandable, on the other hand, it is full of puzzles and questions for which there are no answers yet. And here everything is logical, given that this system is extremely complex both in terms of architecture and in terms of ongoing processes and the relationship between them. If we compare the brain with a computer in classics, then in addition to processing information, it also performs its storage. Any memory is removed from the archives of memory under the influence of some kind of stimulus: a familiar aroma, melody, words, etc. However, the question remains - where is this archive and what contributes to its opening? Scientists from NINDS (National Institute of Neurological Disorders and Stroke) examined the brains of patients with drug-resistant epilepsy to identify and try to explain the mechanisms for extracting memories. So how do we rememberwhat happens in the brain at this moment and why the study was conducted with the participation of patients with epilepsy? We learn about this from the report of scientists. Go.

Study basis


First of all, it is worth noting that patients with epilepsy, which is not amenable to drug treatment (medications, unfortunately, cannot control seizures), are participants in another study in which electrodes are surgically connected to their brain to identify the mechanisms of seizure occurrence.

The presence of these electrodes allows a parallel study of memory, since the connection between this disease and memory is quite curious. Researchers recall that back in 1957, a part of the brain was removed to a certain patient with epilepsy in order to save him from attacks. But the procedure had a serious side effect - the patient could no longer form new memories, i.e. he lost the mechanism of episodic memory.

Since then, the theory has arisen that episodic memories are preserved or encoded as structures (patterns) of neural activity. When a person is faced with some kind of stimulus (a familiar smell, sound, etc.), the brain reproduces this activity, thereby allowing him to remember something related to this stimulus. This is reminiscent of the reproduction of a record on which the memory was recorded, and in this case external stimuli serve as the needle of the player. Nevertheless, no matter how beautiful the analogy is, the very mechanism of this process remains poorly understood.

Previously, a study aimed at explaining the mechanisms for extracting memories has already been conducted. Mice acted as experimental subjects and scientists were able to determine that the brain can store memories in unique sequences of action potentials *.
* — , , .
Scientists decided to check the reliability of the results of research on rodents by conducting the same studies on the human brain. Observations of rodent brain activity, in particular the medial temporal lobe, showed that individual neurons generate impulses in sequences when animals study the environment (in a test chamber), and that these sequences are reproduced during rest (when the animal is not sleeping, but of particular physical activity no) and during sleep.

The reproduction of peak activity sequences was interpreted as memory extraction and consolidation, as well as part of the planning mechanism. But this is all in mice, with humans things can be completely different.

Neural sequences reproduced in the medial temporal lobe of mice are associated with rapid fluctuations, which are called “ripples”. Ripples are also related to extracting episodic memory in humans. Therefore, ripples can in theory be associated with memory-relevant repeated reproductions of peak activity in the human brain.

Research results


To test theories, scientists conducted a study on the relationship between cortical ripples and peak activity of individual neurons. The test subjects were 6 people (4 men and 2 women, average age 34.8 ± 4.7 years).


Image No. 1

The main tools for collecting information were: microelectrode array (MEA) for collecting data on the action potentials of individual neurons and micro-local field * from the anterior temporal lobe; electrocorticogram (iEEG) for collecting macroscale signals from subdural electrodes located above the lateral temporal cortex and along the medial temporal lobe ( 1A and 1B ).
The potentials of the local field * are temporary electrical signals generated in the nervous and other tissues through the total and synchronous electrical activity of individual cells (for example, neurons) in this tissue.
IEEG signals detected ripple-type vibrations in MTG and MTL, as well as any potential connection between brain regions.
MTL - medial temporal lobe of the brain;
MTG is the mean temporal gyrus.
The ripples present in the electrocorticogram recordings of the medial temporal lobe were accompanied by ripples in the micro-LFP signals and peaks of activity of individual neurons ( 1C ). The ripples showed an increase in power in the range from 80 to 120 Hz both on a macro-iEEG scale and a micro-LFP scale.

Each pulsation detected in each microelectrode was accompanied by an increase in the activity of individual neurons in this channel ( 1C ). Peak cortical activity is closely related to the onset of detected pulsations on the macro-iEEG and micro-LFP ( 1D ) scales .

Within a single micro-LFP pulse, the peaks obtained from the electrode channel in a certain region of the cortex were tied to the propagation of ripples, which is consistent with the relationship between peak and pulsation activity observed in rodents and humans ( 1E and 1F ).


Image No. 2

Each of the study participants was asked to perform the task of verbally memorizing pair words, which required them to be encoded and then retrieved new associations between pairs of randomly selected words in each test ( 2A ).

By a burst event, scientists mean time indices during which cortical peaks exceeded the threshold based on a population frequency of at least 25 ms. For all participants, bursts had an average frequency of 1.4 ± 0.2 Hz, and each burst included 39.9 ± 6.3% of all identified units (neurons) during this particular session of the task. Burst events occurred repeatedly throughout the entire time the subject presented verbal pairs ( 2B ).

Next, the scientists reordered the neurons in each test in accordance with the model sequence obtained from the relative burst of activity between pairs of neurons during each coding period. This model sequence was used more for visualization than for analysis of the temporal structure of neuron activity in several events during coding periods and search periods. Neurons during individual bursts apparently preserve the same sequential order of the action potential throughout the entire encoding time ( 2C ).

Since repeated sequences of the action potential were observed when the participants in the experiments encoded pairs of words, it was possible to quantify the degree to which the sequences of the action potential of neurons in the burst events were consistent with each other in different trials or differed in something.

For each burst event, the sequence of peak activity between neurons within this particular burst was determined by ordering each neuron according to when its maximum action potential arose in the range of ± 75 ms from the central burst event index.

Several examples of neurons were found that formed a sequence in one test and then reorganized to form a different sequence during the next test ( 2D ).

In order to check how similar any sequence is to any other sequence, a correspondence coefficient was determined that compares the pairwise temporal relations between all neurons that are common to both sequences and takes a value of 1 for perfect forward playback and -1 for perfect reverse playback .

Having determined the average pairwise value of the correspondence coefficient between all sequences in each test, a comparison was made of this average value with the distribution of coefficient values ​​that occurs when comparing all pairwise combinations of sequences in different trials.

Data analysis showed a common parameter for coding and for reproducing memories — the repeating sequences of cortical activity peaks that were observed in all trials, even when participants did not compose a verbal pair correctly.


Image No. 3

Therefore, if successful coding of memory depends on the time sequence of the action potential of neurons, then the extraction of memory should depend on the same sequence ( 3A) During all tests, repeated burst events were observed during coding and search ( 3B ).

In the process of memory extraction, the sequences, apparently, became more and more similar to the coding sequences until the moment when the participant voiced his answer ( 3C ).

It is curious that the sequence repetition data during coding and during memory retrieval increased if the task was answered correctly (word pair). In the case when the participant incorrectly recreated the verbal pair, less was observed ( 3D) However, before the participant voiced the wrong answer, the search sequences were similar to the coding sequences. In other words, the sequence of activation of neurons during memorization of a verbal pair coincided with the activity during voicing of the answer in the correct version more than in the case of an incorrect answer. From this it follows that the brain, if necessary, remember something specific, selects the desired plate with this memory and reproduces it, metaphorically speaking.

If such a mechanism exists, then it must be individual for different memories ( 3F) It was also found that correct coding and information retrieval had a lower burst frequency of the neuron population and a lower Fano coefficient compared to similar ones in the trials with the incorrect answer. This suggests that a successful search involves reproducing exact sequences of neural excitation ( 3G ).


Image No. 4

As mentioned earlier, the burst events observed during the search are closely related to ripple-type oscillations on the macro-iEEG and micro-LFP scales ( 1C) However, only a few of these cortical events are associated with ripples in the medial temporal lobe. Previous studies suggest that cortical burst events associated with similar events in the medial temporal lobe ( 4A ) are at the heart of memory retrieval .

During verbal pair tests, burst events associated with the medial temporal lobe were observed, which showed a higher coefficient of sequence similarity with the coding period than those events that occurred in the absence of activity of the medial temporal lobe ( 4B ).

It was also found that the reproduction of memory in the cerebral cortex, caused by activity in the medial temporal lobe, occurred no later than 100 ms after the onset of this activity ( 4C ).

During the tests, when the participants gave the correct answer, the burst events associated with MTL pulsations showed a significantly greater reproduction of the sequences present during coding compared to unrelated events ( 4D ).

From this it follows that for each coding of memory there is its own sequence of activity of individual neurons. And for the correct reproduction of memories, the brain must reproduce this sequence repeatedly.

For a more detailed acquaintance with the nuances of the study, I recommend that you look into the report of scientists .

Epilogue


In this study, scientists were able to obtain direct material evidence that the reproduction of memories is based on the coordinated reproduction of sequences of action potentials of neurons in the human brain.

When a person remembers something, a sequence of neuronal activity is formed in the brain. When he wants to remember something, in order to successfully extract the desired memory, his brain must reproduce the previously created sequence.

This was confirmed during tests. When the test participants correctly recalled a given verbal pair, the sequence of reproduction (memory) and coding (memorization) coincided, which was not observed in cases of erroneous answers.

According to the researchers, their work can become an additional tool in trying to understand all the features of destructive processes in the human brain that cause impaired memory, consciousness and thinking. If we argue from a more sci-fi point of view, then understanding that there is a certain sequence that can be reproduced can allow us to accurately and quickly reproduce the necessary memories at the right time.

A bit of advertising :)


Thank you for staying with us. Do you like our articles? Want to see more interesting materials? Support us by placing an order or recommending to your friends, cloud VPS for developers from $ 4.99 , a unique analog of entry-level servers that was invented by us for you: The whole truth about VPS (KVM) E5-2697 v3 (6 Cores) 10GB DDR4 480GB SSD 1Gbps from $ 19 or how to divide the server? (options are available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).

Dell R730xd 2 times cheaper at the Equinix Tier IV data center in Amsterdam? Only we have 2 x Intel TetraDeca-Core Xeon 2x E5-2697v3 2.6GHz 14C 64GB DDR4 4x960GB SSD 1Gbps 100 TV from $ 199 in the Netherlands!Dell R420 - 2x E5-2430 2.2Ghz 6C 128GB DDR3 2x960GB SSD 1Gbps 100TB - from $ 99! Read about How to Build Infrastructure Bldg. class c using Dell R730xd E5-2650 v4 servers costing 9,000 euros for a penny?

More articles:


All Articles