Targeted Memory Modulator: A Hybrid fMRI-EEG-FUS Modality for Reading and Modulating Specific Human Memories

Inspired by fictional neurotechnology character’s creation Tony Stark’s “B.A.R.F”, this paper proposes a real world hybrid system for healing trauma by reading and modulating specific memory engrams. The system integrates three complimentary neuroimaging modalities, including high-field functional magnetic resonance imaging (fMRI), high-density electroencephalography (EEG), and transcranial focused ultrasound (FUS). Human memory is the foundation of all. It is how humans form their identity, emotion, and learning. However, no technology has ever been able to pinpoint and directly modulate a specific human memory without invasive surgery. Most of the time, neuroimaging tools only offer a partial view of what we want to know, for example, fMRI provides precise spatial localization, and EEG captures the timing of neural events. These technologies have their limitations as well, such as the lack of temporal resolution that fMRI provides, while EEG unable to locate the sources of neural activity. Furthermore, there is recent advancement in transcranial focused ultrasound or what is known as tFUS. Legon et al. explained, “tFUS can focally modulate sensory evoked brain activity and cortical function in humans” (2014). This aligns well with the aim of this invention, that is tFUS as a third dimension, non-invasive, targeted neuromodulation. This paper proposes a system using three modalities integration of fMRI, EEG, and tFUS. A system that could locate, read, and modulate specific memory engrams (the physical traces of memories encoded across distributed neural circuits). By combining high-resolution imaging with millisecond-scale electrophysiology and precise ultrasound targeting, this hybrid system aims to make real-time memory mapping and modulation to it.

One modality that we use is a high-field fMRI as the “finder”. We can think of this as a GPS in the brain. Based on Ugurbil (2021), “ Ultrahigh magnetic field (UHF) of 7 Tesla has played a critical role in enabling higher resolution and more accurate (relative to the neuronal activity) functional maps.” (2021). These advances can help to show the brain map in macroscopic scale that can illustrate images from a seed in the brain and its networks that form the entire brain. With this ultrahigh field strength, it allows submillimeter spatial resolution, sufficient to differentiate activity patterns within the hippocampus that is responsible for distinct aspects of memory. Based on Ekstrom et al. (2009), it confirmed that the hippocampal area, including the subregion and the subiculum, as well as the surrounding cortices plays very significant roles in learning and memory. Moreover, the way that fMRI works is that it measures blood oxygen level-signals, which indirectly reflect local neuronal activation. But it does have limitations, because the blood-oxygen-level-signals peaks several seconds after the neural activity. Logothetis et al. explained that “ The subsequent signal increase is delayed by 2–3 s, followed by a ramp of 6–12 s to a plateau or peak value for long (>10 s) or short (<10 s) pulses, respectively” (2001). This finding shows that fMRI could be unsuitable for capturing the millisecond-scale dynamics of memory replay. Therefore, we can take fMRI to provide the spatial address, but not the temporal narrative of the engram.

The second modality we are using is high-density EEG. To complement fMRI, EEG provides the “when” portion and acts as the content reader of the memory. Since memory is all about network of connection and their neural connectivity that activated and inhibited, Fell and Axmacher (2011) argued that “This rhythmicity is reflected in oscillations of the extracellular field potential that can be measured through recordings of local field potentials and through electroencephalography (EEG)”. For this rhythm of connectivity to make sense of a memory, the brain also performs synchronization that stores the association of one thing to another between regions that represent specific stimulus to a memory. In addition, performance in EEG can also help this technology to detect and evoke specific brain waves that correlates with memory encoding and retrieval. Based on Hanslmayr et al., “ increases in synchronized activity in the theta (around 5 Hz) and gamma (>40 Hz) frequency ranges play an important role for memory formation and retrieval via shaping synaptic plasticity and coordinating the reactivation of memories” (2012). It gives the system a concrete, measurable biomarker for memory. This can give a function of the technology to look specifically at the theta-gamma signals as the signature of recall with high precision. With these two combined, it can capture electrical field changes generated by postsynaptic potentials, offering millisecond precision.

The last modality that we can apply on is tFUS as the “rewrite” tool. Using ultrasound technology as part of this invention can get us to directly interact with the nervous system and neuron connectivity that is associated with the memory. Tyler et al., argued that “ Ultrasound (US) as a means of exciting and reversibly suppressing neuronal activity was shown to be effective on a gross level several decades ago” (2018). We can apply this concept to the engram system by interacting with the specific neurons, with the support of data from EEG and fMRI to touch and alter the selected memory. In this invention, the purpose of tFUS is to direct and non-invasively modulate deep brain structures that correlates with the memory and the reactivity to the memory. Barksdale et al (2025) discussed the use of tFUS as an intervention in neuromodulation, and said that, “the focused administration of low-intensity, high frequency sound waves to the brain, is a novel method of reversibly augmenting brain function”. However, this method does bring advantage, focused ultrasound can deliver a spatial precision and adjustable focal depth to the specific brain part that is associated with memory that is targeted. In a study experiment, Mahdavi et al (2023) use tFUS to modulate amygdalar activity to see the effect among patients with generalized anxiety disorder, and the result shows there are some improvements in anxious symptomatology, in which there was a significant decrease in scores after eight weeks treatments. This aspect from tFUS gives us an insight on is a strong component to this invention.

In Tony Stark’s B.A.R.F. model, the technology was presented as a simple, wearable device. While a fully wearable system is not feasible for high-field fMRI, the integration philosophy of B.A.R.F., a single, unified system that combines multiple neural technologies could be achievable in a clinical lab setting. The architecture of the Engram System relies on synchronizing three components (fMRI, EEG, FUS) within a high-field magnetic environment. This presents a major engineering challenge, as the components can interfere with one another. The solution is a custom-designed, "tri-modal" head coil where all three systems are physically integrated and co-registered to the same 3D coordinate space.

The first one is EEG-fMRI integration. When only using a powerful high-scale fMRI scanner that creates large electrical artifacts could contaminate the EEG signal, with this integration, researchers noted that, “ The present approach offers a principled framework to integrate fMRI and EEG, and promises to provide high resolution and precision in both time and space” (Yang et al, 2010). This will help us to get a clean and accurate recorded data in real-time. And the second integration is FUS-fMRI integration. Fomenko et al. (2020) present that they were able to integrate tFUS that is MRI-safe, “ A custom two-element annular array ultrasound transducer (Sonic Concepts Inc, Bothell, Washington) operating at a fundamental frequency of 500 kHz and housed in a MRI-safe non-ferromagnetic brass cylinder measuring 38 mm in diameter and 10 mm thick was used” . They successfully built and tested a FUS transducer from "non-ferromagnetic brass" that can operate inside an MRI and not distort the magnetic field. This proves that physical integration is feasible.

This integrated hardware enables a powerful "closed-loop" workflow:

• Phase 1: Locate (fMRI). The fMRI provides a continuous, real-time map of brain activity. The system identifies the 3D coordinates (the "address") of the memory engram in the hippocampus/amygdala.
• Phase 2: Trigger (EEG). The system simultaneously monitors the high-speed EEG data, "listening" for the specific theta-gamma "song" (Hanslmayr et al., 2012) that signals the memory is being recalled.
• Phase 3: Modulate (FUS). This is the closed-loop intervention. The computer is programmed to fire a FUS pulse only when both conditions are met: the EEG detects the "trigger," and the FUS is aimed at the fMRI's "address." This is EEG-triggered, fMRI-guided neuromodulation, an event that is impossible without all three modalities working in perfect synchrony.

The data from this tri-modal system give primary output that could possibly visualize all three streams, including fMRI view that shows a real-time 3D brain map, and using EEG view to get a real-time spectrogram that shows the power of different neural oscillations, and lastly, a tFUS view that act as a control panel for the target coordinates and tracking pulse intensity. This all comes together when during activation and repeated therapeutic sessions, changing the emotional modulation of memory trace could be successful.

The integration of fMRI, EEG, and tFUS demands high performance real-time processing. Each modality generates data per second and that many signal removal and classification required for this technology to run smoothly without any possible interference. Using ultrahigh fMRI and tFUS that are recently advanced can be difficult to achieve and to combine. The estimated costs would be millions of dollars. The ethical considerations can also raise a flag. While much literature and research spoke about the minimum effects of MR-guided tFUS, the long-term effects of repeated modulation are unknown. Moreover, it does bring some more questions, since memory is part of one’s identity, does alter it can be considered as ethical.

There are few applicable situations where we can implement this technology. Since we are accessing memory and are able to reconstruct them, this technology can be one of the implications that could be used for people with Post-Traumatic Stress Disorder or PTSD. This technology could dampen the hyperconnectivity between the hippocampus and amygdala that underlies traumatic flashbacks, and provide a direct therapeutic approach to the brain. Beyond therapy, the system offers an unprecedented window to the biology of memory formation and retrieval. We can use it to observe real-time activation of individual memory traces. Researchers could also use it to decode the language of memory in the human brain.