Introdution
Click to copy section linkSection link copied!
A memristor, short for “memory + resistor”, is an electronic device that has both resistance and memory functions, which can be designed as a resistance-based memory. (1) This theoretical concept was first proposed in 1970, leading to a rapid development of memristor-based devices in the subsequent years. (2) Memristor has a typical nonvolatile resistance switching behavior, which can be well used in information memory, also called resistance random access memory (RRAM). (3) Structural engineering studies have shown that the RRAM device has a sandwich structure, a layer of semiconductor/insulator sandwiched between two conductive electrodes, similar to other memory devices, such as phase change memory (PCM)…
Introdution
Click to copy section linkSection link copied!
A memristor, short for “memory + resistor”, is an electronic device that has both resistance and memory functions, which can be designed as a resistance-based memory. (1) This theoretical concept was first proposed in 1970, leading to a rapid development of memristor-based devices in the subsequent years. (2) Memristor has a typical nonvolatile resistance switching behavior, which can be well used in information memory, also called resistance random access memory (RRAM). (3) Structural engineering studies have shown that the RRAM device has a sandwich structure, a layer of semiconductor/insulator sandwiched between two conductive electrodes, similar to other memory devices, such as phase change memory (PCM) based on phase change materials and spin-transfer torque magnetic random access memory (STT-MRAM) based on spintronic materials, (4−6) as shown in Figure 1. It should be noted that, although some advantageous characteristics can be achieved in these three types of devices, the single device that combines them together is only RRAM. Therefore, RRAM is considered one of the most promising electronic devices in the post-Moore era.
Figure 1
Figure 1. Comparison of various performance parameters of the three resistance-based memories, including RRAM, PCM and STT-MRAM.
In particular, RRAM has a readily available structure similar to that of a biosynapse, which provides a physical basis for simulating the performance of a biosynapse using memristor based on the analog resistive switching effect. (7) Therefore, since the first device preparation by the HP laboratory in 2008, memristors have been widely studied in the fields of information memory and neuromorphic computing due to their nonvolatile memory, high integration density, low power consumption, and synaptic-like behavior. (8−10) With the increasingly mature preparation of memristors, they have shown great potential applications in many fields, such as nonvolatile memory, artificial neural networks, artificial synapses, neuromorphic computing, and intelligent electronic skin (e-skin). Therefore, it is highly necessary to continue developing the application of memristors in emerging fields such as biological detection, medical diagnosis, and brain–computer interfaces.
In addition, due to the fact that traditional sensors only have a single signal acquisition function, they have shown a huge application bottleneck in the era of high-tech intelligence. Here, we should consider whether memristors also have signal sensing functions due to their structural isomorphism with biosynapses. Fortunately, unlike traditional sensors that are limited to signal acquisition, memristors seamlessly integrate sensing, nonvolatile memory, and neuromorphic computing within a unified physical structure (usually a crossbar array), fundamentally breaking the von Neumann bottleneck and truly achieving integrated sensing-memory-computing technology into a device unit. This integration eliminates the need to transfer data back and forth between discrete components, thereby improving the integration density of device efficiency and computing speed. As shown in Figure 2, a qualitative comparison between traditional sensor systems and memristive-sensing architectures reveals that memristors with integrated sensing-memory-computing have significant advantages in many applications. In other words, memristors integrate sensing, memory, and computing functions into a single device unit, while sensors only perform sensing and require many additional components to achieve memory and computing capabilities. Therefore, memristors with integrated sensing-storage-computing functions will demonstrate greater potential in the biomedical field, greatly promoting the rapid development of smart medicine.
Figure 2
Figure 2. Comparison between traditional sensor-based systems and memristor-based sensing-memory-computing integrated systems.
At the same time, since the variable conductance of the memristor can simulate the synaptic weight by adjusting the charge transmission under the applied voltage, the memristor can be implemented as a compact hardware for biosynaptic application. (11) Biosynapses are the critical parts of neurotransmitter transmission between presynaptic neurons and postsynaptic neurons in the human brain, which are involved in information transmission and processing and are considered to be the biological basis for learning and memory. (12) The memristor that functions as the electronic device is closest to the neuron’s biosynapse relative to other devices. In building and simulating neurons, memristors have been considered one of the most advanced and best implementations to achieve the artificial synaptic device, and its development is of significant important for neuromorphic engineering. (13−15) Therefore, memristors can simulate the dynamic behavior of biological neurons and directly interact with bioelectric signals, giving their information unique advantages in biosensing, health monitoring, disease diagnosis, intelligent treatment, and neuromorphic system, (16−20) which may be intuitively applied to disease diagnosis and health monitoring using memristor-based brain-like chips in the biomedical field, including blood glucose testing, pulsating hypertension monitoring, and artificial synapse reconstruction, and in the field of neuroscience, as shown in Figure 3. In recent years, the application of memristors in the biomedical field has gradually attracted attention because they can be used in brain–computer interfaces, bioinformatics sensing, and medical detection.
Figure 3
Figure 3. Memristor-based neuromorphic chips can be applied for the disease diagnosis and health monitoring in biomedical fields.
This Perspective systematically reviews the key application progress of memristors in biomedical fields and explores their development directions. Specifically, a series of effective solutions are proposed to address the existing problems and challenges in current research, which will help advance the application of memristors in the field of intelligent medicine. Finally, we provide a series of unique perspectives on the application of memristor-based neuromorphic chips in the field of neuroscience, which can greatly promote the rapid development of smart medicine.
Application of Memristor in Biomedical Field
Click to copy section linkSection link copied!
Health Monitoring
In previous work, it was believed that the high sensitivity and nonlinear response characteristics of memristors make them suitable for detecting bioelectric signals, such as electrocardiogram (ECG), electroencephalogram (EEG), and electromyography (EMG). (21) Compared with traditional CMOS-based sensors, memristor arrays can achieve higher density signal acquisition and perform preliminary signal processing directly at the sensing end, reducing data transmission requirements. (22) In addition, through functional modifications such as antibody, DNA probe, hydrogel, or enzyme immobilization, memristors can be used to detect specific biomolecules such as glucose, virus, bacteria, proteins, and nucleic acids. (23) During the detection process, the molecules or ions among the detected materials usually undergo chemical reactions with the molecules in the functional layer of the device, which can cause fluctuations in the memristive effect of the device. The change in memristive behavior of device can reflect the concentration and types of target molecules, achieving accurate detection with high sensitivity. In particular, flexible and biocompatible memristors can be attached to the skin or implanted into the body for long-term and real-time health monitoring, indicating memristors have unparalleled advantages in the applications of biomedical field. (24,25)
Memristors can be used to construct highly sensitive memristive sensors, which can achieve more accurate medical signal detection and recognition at the cellular level. (26) In fact, many physiological system diseases are directly related to cellular vitality/homeostasis issues, while brain diseases are often associated with abnormal discharge. (27,28) A memristor is a nanoscale electronic component capable of simulating synapses, providing unique features to address the size and power limitations of traditional electronic devices. (29) The most noteworthy point is that, due to the characteristic value of memristors being the resistance state of the device, it is a parameter that is less prone to fluctuations compared to other electronic parameters. Therefore, utilizing the high resistance state (HRS) and low resistance state (LRS) of memristors to reflect physiological changes or indicators may have unparalleled unique advantages in biological detection and disease diagnosis.
In previous report, it was demonstrated for the first time the use of memristive devices in clinically relevant environments, where communication between two populations of neurons depends on specific activity patterns in the source population. Aguiar et al. proposed the application of memristive devices in performing pattern detection calculations through the cultivation of hippocampal neurons in vitro, (30) and the principle of experimental process is shown in Figure 4A. In this research, they demonstrated a highly reproducible real-time adaptive controlling system using a monitoring computing driving paradigm, which can be obtained by the automatic detection and suppression of epileptic seizures in epilepsy patients. Based on this work, the cure of epilepsy patients is just around the corner; by monitoring abnormal neural discharges in the affected area, nerve resection and even artificial nerve replacement can be achieved.
Figure 4
Figure 4. (A) Memristor-based neuromodulation device for real-time monitoring and adaptive control of neuronal populations. Reproduced from ref (30). Copyright 2022, American Chemical Society. (B) An energy-efficient high-fidelity image reconstruction with memristor arrays for medical diagnosis. Reproduced from ref (31). Available under a CC-BY license. Copyright H. Zhao et al. (C) Schematic diagram of a reversible implantable memristor for health monitoring applications. Reproduced with permission from ref (32). Copyright 2024, Elsevier. (D) Schematic diagram of experimental design and the clinical application scenario of the Ag/WO3/Ti memristor for rapid detection of brain tumor cells. Reproduced with permission from ref (33). Copyright 2025, Elsevier.
It should be first mentioned that Professor Wu’s team proposed an innovative memristive image reconstructor (MIR) that greatly accelerates the image reconstruction of Discrete Fourier Transform (DFT) by using a computing-in-memory (CIM) model, (31) as shown in Figure 4B. At the same time, they also proposed a high-precision quasi-analogue mapping (QAM) method and a universal complex matrix transfer (CMT) scheme, which can respectively improve mapping accuracy and transmission efficiency. High fidelity magnetic resonance imaging (MRI) and computed tomography (CT) image reconstruction were also demonstrated, and software equivalent quality and DICE score can be achieved after segmentation using the nnU Net algorithm. It is worth noting that the as-proposed MIR improves energy efficiency and normalized image reconstruction speed by 153 times and 79 times, respectively, compared to the graphics processing unit (GPU). Thus, this work demonstrates that MIR is a promising high fidelity image reconstruction platform that can be used for future medical diagnosis and greatly expands the application of memristor-based CIM beyond artificial neural networks. Based on this work, it is expected to apply memristors to real-time monitoring of organ health in vivo by implanting memristor chips into living organisms.
In addition, it is undeniable that real-time monitoring for the health status of transplanted liver is a very important medical tool. In our current research work, a highly stable implantable memristor with Ag/BaTiO3/MnO2/FTO structure was successfully fabricated by magnetron sputtering method, and the device can maintain the good resistive switching characteristics for over 1200 cycles. (32) Specifically, the device successfully demonstrated a three-stage response to biological signals after implantation in SD (Sprague–Dawley) rats, which can be achieved lesion monitoring of liver tissue, as shown in Figure 4C. Importantly, the memristor exhibits significant reversibility, maintaining the same stable memristive performance even after extraction from rats body, indicating that the memristive device has reversible usability and will be able to achieve repeated use for implantable monitoring. Therefore, this study provides a new perspective for the application of memristors in the biomedical field, expanding the potential of implantable memristors in intelligent medical fields such as health monitoring and auxiliary diagnosis. Inspired by this work, it will be more promising applications to design a magnetic field sensitive memristor chip that is expected to achieve directional detection controlled by an external magnetic field.
Currently, malignant tumors pose a significant global challenge in daily life, public health, and economic development, with brain tumors being one of the deadliest tumors. It can be seen that early and accurate detection of brain tumors will play a crucial role in achieving timely and appropriate treatment strategies, which is expected to improve the cure rate of patients. Based on this reason, we presented a novel and innovative approach for the rapid detection and differentiation of brain tumor cells using a memristor device with Ag/WO3/Ti structure, (33) as shown in Figure 4D. In this work, we have fabricated a stable memristor and demonstrated its ability to identify different brain tumor cell types, specifically distinguishing between highly invasive and low invasive glioma cells based on changes in the I–V curves and resistance states. This study successfully establishes a proof-of-concept for the memristor’s feasibility in differentiating glioma cell types based on their electrical signatures to distinguish between highly invasive (U251 and U87) and low invasive (LN-18 and SHG44) brain tumor cells is a significant finding with direct clinical relevance. Moreover, preliminary in vivo implantation experiments in mice to assess biosafety and biocompatibility are a valuable addition for future clinical translation, which helps contextualize the advantages and limitations of the memristor-based approach. It can be seen that these results support rapid detection of brain tumor cells and have broad clinical application prospects, thus paving the way for clinical translation in rapid diagnostics, surgical guidance, and postoperative monitoring. Therefore, this work has opened up a new opportunity for the application of memristors in the field of neuroscience research.
Medical Image Processing
From image analysis to lesion diagnosis, artificial intelligence (AI) technology has aroused great interest in the medical community. (34) However, the progress of artificial intelligence in medicine has been hindered by the lack of available medical image data sets and labor-intensive labeling processes. (35) Fortunately, memristors have significant advantages in medical image recognition, mainly reflected in the following aspects: (i) The resistance state changes of memristors are similar to the connection strength of biological synapses, which can simulate synaptic weights through different resistance values and achieve dynamic adjustment of information transmission. This characteristic enables memristors to more accurately simulate the learning and memory mechanisms of biological neurons when constructing artificial neural networks. (ii) Memristors have nonvolatile and nanosecond switching speed, with power consumption only one-tenth or even lower than traditional memory. In image recognition tasks, its fast response characteristics can significantly improve real-time processing capabilities while reducing device energy consumption. (iii) Memristors can integrate memory and computing functions, breaking the bottleneck of memory and computing separation in the von Neumann system. In image recognition scenarios, data can be directly computed in storage units to avoid data transmission delays and improve overall system efficiency. (iv) Through array design, it can be expected that memristors can achieve large-scale neural connections and support parallel computing. In biological image recognition, multiple memristors can simultaneously process image features from different regions, achieving efficient feature extraction and classification.
Here, Kim et al. demonstrated that the use of state-of-the-art generative adversarial networks (GANs) can generate a large number of annotated and realistic chest X-ray images, which utilize the noise generated by random memory calculations of memristor crossbar arrays, (36) as shown in Figure 5A. In this work, the memristors based on polymer films with high thermal resistance can increase the randomness of the tunneling distance of randomly broken conductive wires through excessive Joule heating, thereby generating the true random numbers required to create naturally diverse images in GANs. Further, using StyleGAN2 adaptive discriminator augmentation (ADA), high-quality chest X-ray images with and without pneumothorax can be successfully enhanced while maintaining a good Frechet starting distance score. In a word, this work provides an economical and efficient solution for preparing privacy sensitive medical images and labels to develop innovative applications of artificial intelligence algorithms in medical field. Therefore, the application of medical imaging based on memristor chips will be expanded to more fields, and the improvement of existing medical detection methods will also be a more important aspect.
Figure 5
Figure 5. (A) A stochastic memristor array for machine learning-based medical diagnosis. Reproduced with permission from ref (36). Copyright 2024, Wiley-VCH. (B) A real time hybrid medical image encryption algorithm combining memristor-based chaos with DNA coding. Reproduced from ref (39). Available under a CC-BY license. Copyright A. S. Demirkol et al. (C) The medical image reconstruction based on a sustainable biomemristor designed logic gate circuit. Reproduced with permission from ref (40). Copyright 2024, Elsevier.
In addition, image encryption can be effectively used to protect medical data on public networks, demonstrating crucial applications in the healthcare industry and medical information encryption. (37,38) Due to the complex dynamic characteristics of memristors, they are often used to develop new chaotic systems to improve the efficiency of chaos-based encryption algorithms. For example, Demirkol et al. proposed a novel chaotic circuit model based on locally active memristors and introduced a real-time hybrid image encryption application developed using the Jupiter programming environment (Figure 5B). (39) In this work, the as-proposed hybrid algorithm combines chaos based on memristors with the DNA (DNA) encryption algorithm using diffusion obfuscation technology. They first proposed a new compact noninductive chaotic circuit, derived the model equation, and then numerically verified its chaotic dynamics by studying the phase diagram, Lyapunov exponent, and bifurcation diagram. Further, a chaotic circuit was implemented using discrete element experiments and the experiments were evaluated through NIST testing. At the same time, various indicators were used to evaluate the performance of encryption algorithms, including histograms and correlation analysis, differential attacks, and information entropy, as well as data loss and noise attacks, proving the security and applicability to real-time encryption systems for medical image and data security.
In the research of memristors, using natural biomaterials to prepare memristors has many advantages, such as sustainability, no pollution, biocompatibility, and nontoxicity. Thus, it will be very meaningful research if natural biomaterial-based memristors can be applied in the biomedical field and realize medical engineering applications. In our current work, a sustainable biomemristor with Ag/mugwort:PVDF/ITO structure was prepared by spin coating and magnetron sputtering methods using the nature biomaterial mugwort as a core functional layer of the device. (40) It was found that the biomemristor represents good resistive switching behavior with obvious rectifying effect under an applied voltage from −1.0 to 1.0 V. Based on such physical characteristics, a new logic gate circuit was constructed using the as-prepared memristor, and an innovative 3–8 decoder encryption circuit based on memristor was designed, which can achieve unified rule encryption and decryption of data, medical images, and medical information, as shown in Figure 5C. Therefore, this work achieves the integration of memristors with traditional electronic technology and expands the application of sustainable biomemristors in digital circuits, data encryption, and medical image security.
Brain–Computer Interface
It is well-known that the synaptic plasticity of memristor makes them suitable for neural regulatory systems such as deep brain stimulation (DBS) and epilepsy prediction. (41) the memristors can achieve nonvolatile memory and dynamic signal processing by simulating the discharge process of neurons, of which the resistance transition mechanism mainly relies on the migration of electrons and ions in the functional layer under the action of an electric field, changing the functional layer material structure (such as the formation/breakage of conductive channels), thereby dynamically adjusting the resistance of the device. Based on this principle, the memristor array can monitor real-time changes in electroencephalogram (EEG) signals and convert the signals into instructions through an adaptive neuromorphic decoder. Therefore, the brain–computer interface (BCI) based on memristors can achieve efficient signal processing and interaction by simulating the firing process of brain neurons, thereby realizing hardware-based brain–computer information output and monitoring. The application of memristors in brain–computer interfaces includes the following advantages: (i) The energy consumption of brain–computer interfaces based on memristor chips is only 1000th of that of traditional CPU systems, indicating the operating energy consumption is low. (ii) Memristor implements a single step decoding strategy, which filters signals and extracts features with low computational complexity and fast speed. (iii) Through the dual loop collaborative evolution framework, brain signals and decoders can adapt naturally, it can be achieved high accuracy and stable human–computer interaction. (iv) The structure of memristors is similar to that of synapses in the human brain and can be reduced to the nanometer scale, making them more suitable for neuromorphic calculations in the brain.
At present, the brain–computer interface is a very promising strategy in the field of neuroscience research, which can restore lost motor function and explore the mechanisms of brain function. (42) As the volume of recording electrodes continues to shrink and the number grows exponentially, the signal processing capability of brain–computer interfaces shows great backwardness due to the high power consumption and slow processing speed caused by traditional von Neumann architecture and digital computing, which fundamentally differs from the working principle of the human brain. (43) Based on the above reasons, Professor Wu’s research group proposed a neural signal analysis system based on memristors, which utilizes the reasonable characteristics of memristors in biology to efficiently analyze signals in the simulation domain, (44) as shown in Figure 6A. As a proof of concept demonstration, the memristor array was used to filter and recognize epilepsy related neural signals, which can achieve a high accuracy of 93.46%. It is worth noting that, compared to the most advanced complementary metal oxide semiconductor (CMOS) systems, this memristor-based brain–computer interface system has improved power efficiency by nearly 400 times. Therefore, this work demonstrates the feasibility of using memristors for high-performance neural signal analysis in the next generation of brain–computer interfaces, which will greatly promote the research progress of neuroscience and intelligent robots.
Figure 6
Figure 6. (A) Memristor-based neural signal analysis system for brain–machine interfaces. Reproduced from ref (44). Available under a CC-BY license. Copyright Z. Liu et al. (B) A memristor-based adaptive neuromorphic decoder for brain–computer interfaces. Reproduced with permission from ref (45). Copyright 2025, Nature Publishing Group.
Specifically, a practical brain–computer interface should be able to decipher brain signals and dynamically adapt to brain fluctuations. In addition to the artificial neural network based on memristors discussed above, which can achieve a brain–computer interface system, a decoder that can be flexibly updated and has energy-saving decoding capabilities is also needed. In recent work, Professor Wu’s team innovatively reported a neuromorphic and adaptive decoder for brain–computer interfaces, based on a 128k unit memristor chip, (45) as shown in Figure 6B. This method has a hardware efficient one-step memristor decoding strategy, allowing the interface to achieve decoding performance equivalent to software. Besides, researchers have also demonstrated that the system can be used for real-time control of drones in four degrees of freedom, leading to the development of an interactive update framework that enables the memristor decoder and constantly changing brain signals to adapt to each other. Finally, the coevolutionary ability of the brain and memristor decoder was successfully demonstrated in an extended interactive task involving 10 participants, which was about 20% more accurate than an interface without coevolution. From these results, it can be found that neural networks based on memristors can be effectively applied to the construction of human–machine interface systems. Therefore, this work can not only achieve efficient and low-power processing of human brain information but also have the potential to expand a wider range of application functions in biomedical field.
Neuromorphic Diagnosis and Treatment Systems
With the popularization of artificial intelligence technology, medical auxiliary diagnostic systems developed based on artificial intelligence technology can quickly and accurately analyze a large amount of medical imaging data, such as X-rays, CT, MRI, etc., which can provide some valuable diagnostic recommendations for doctors. However, the electronic devices based on the traditional von Neumann architecture have faced many bottlenecks, indicating that the development of new chips integrated memory and computing is imperative. Fortunately, memristor neural network-based neuromorphic diagnosis and treatment system can achieve more natural prosthetic control. (46) Memristors have a similar sandwich structure to neural synapses, which have natural structural advantages in the preparation of implantable neuromorphic diagnostic and therapeutic systems. (47) Specifically, the implantable diagnostic and therapeutic systems have the potential to address diseases that currently lack effective treatment options.
At the same time, the implantable diagnostic and therapeutic devices need to have good biocompatibility, which can provide enhanced performance indicators of low power consumption, high precision, small size, and minimal delay to achieve continuous intervention in brain function. Caterina et al. demonstrated a diagnosis and treatment system based on memristors, which was used to support single trial detection of brain signals that are meaningful to behavior within a time range of real-time closed-loop intervention, (48) as shown in Figure 7A. In this study, neural activity in the ventral tegmental area of the reward center in rats trained to associate music tones with rewards was recorded by using the threshold characteristics built into memristors to detect nontrivial biomarkers in local field potentials. It was found that the proposed method can continuously and accurately detect >98% of biomarkers, while keeping the power consumption of each channel below 4.14 mW. Therefore, the ability of implantable diagnostic and therapeutic systems based on memristors to process real-time in vivo neural data paves the way for low-power chronic neural activity monitoring and biomedical implantation.
Figure 7
Figure 7. (A) Signal processing flow in auditory cue detection using memristor devices. Reproduced from ref (48). Available under a CC-BY license. Copyright C. Sbandati et al. (B) Human-machine interaction using implantable memristors for in-memory computing and physiological signal monitoring. Reproduced with permission from ref (49). Copyright 2024, Elsevier. (C) Memristor-based pressure information processing system and as-designed pulmonary hypertension detection system based on memristor array for construction of a fully connected neural network based on memristors for the feedback and processing of blood pressure signals. Reproduced with permission from ref (50). Copyright 2025, Wiley-VCH. (D) The memristor-based intracranial pressure (ICP) monitoring and warning system can detect potential intracranial hypertension following craniotomy for various neurological diseases. Reproduced from ref (51). Copyright 2025, American Chemical Society.
In current work, we demonstrated a flexible memristor by using ZrOx thin film as functional layer, which shows broad application prospects as an implantable device. (49) This work realizes the interaction between implantable memristors in biomedical and processing systems, which proves the feasibility of flexible memristors in the field of implantable biomedical applications, and promotes the rapid development of implantable neuromorphic diagnosis and treatment systems, as shown in Figure 7B. Furthermore, we have developed a memristor with Ag/MnO2/BaTiO3/FTO structure by using MnO2/BaTiO3 heterojunction as functional layer, and then implanted it into Sprague–Dawley (SD) rats. Using polydimethylsiloxane (PDMS) encapsulation, the device can operate continuously in vivo for up to 4 weeks and without significant performance degradation, demonstrating excellent stability and biocompatibility. (50) In addition, based on the pressure response characteristics of the as-prepared memristive device, a memristive sensor array for pulmonary artery blood pressure monitoring was designed. The front-end memristor sensor array can collect and feedback pressure signals, while noise reduction is achieved through memristor logic circuits, and ultimately the information is processed by a memristor neural network using neuromorphic computing, as shown in Figure 7C. Therefore, this work demonstrates the potential of implantable memristors in pulmonary artery pressure monitoring and provides new inspiration for the design of efficient, real-time, and reliable implantable neuromorphic diagnosis and treatment systems in medical health monitoring.
In recent work, a memristor was prepared using the WO3/MnO2 heterojunction as the functional layer of the device, and its pressure signal encoding ability in vitro was successfully verified by integrating it with a pressure sensor, (51) as shown in Figure 7D. Specifically, through neuromorphic computing processing, it can achieve imaging recognition of intracranial pressure (ICP). Furthermore, through extensive testing and characterization, it was found that the sensor–memristor integrated system can be used for the acquisition and encoding of intracranial pressure signals, and it has good biocompatibility in animal bodies through biocompatibility characterization. This study provides the first evidence of the potential application of an implantable resistive memristor–sensor integrated system in monitoring ICP after craniotomy, and emphasizes its role in enhancing neurosurgical care. Therefore, this work provides innovative insights for designing efficient, real-time, and low-power implantable brain disease monitoring devices for medical health monitoring.
Challenges and Prospects
Click to copy section linkSection link copied!
Current Challenges
Due to its simple sandwich structure, easy preparation process, synaptic-like behavior, and advantages of integrated memory and computing, memristors have attracted the attention of many researchers in biomedical applications. (52−59) At present, although memristors have made rapid progress in the biomedical field, there are still many challenges on the way to widespread application. (1) Biocompatibility. For implantable electronic chips, it is essential to develop safer packaging materials to avoid immune reactions caused by long-term implantation. (2) Biological stability and reliability. For implantable electronic devices, when they are implanted into the body, the biological environment (such as body fluids, mechanical deformation) may affect the performance of the memristor. Therefore, it is crucial to prepare memristors with good biological stability. (3) System integration. In order to accelerate the clinical application of memristors, it is necessary to optimize the compatibility between memristors and existing medical electronic devices. (4) Good flexibility. For memristors implanted into living organisms, they are usually subjected to pressure from internal activities, so good flexibility is the primary requirement for implanted electronic devices. (5) Size factor. For electronic devices implanted into living organisms, their size should not affect the activities of the organism. Of course, in addition to these major challenges, there may be another factor that needs to be considered, such as the toxicity of materials, the rejection of devices and organisms, and the metabolism after device failure. Therefore, it is undoubtedly necessary to overcome various challenges in order to prepare a perfect implantable memristive device and successfully apply it into the biomedical field.
Prospects
With the current research progress, the main trends in the application of memristors in the biomedical field are as follows: (1) Biological hybrid systems. Since memristors can simulate synaptic behavior; they can directly interact with living neurons for neural repair. (2) Self-powered medical equipment. Due to the easy integration of memristors with other electronic devices, self-powered implantable devices can be achieved by combining energy harvesting technologies such as biofuel cells. (60) (3) Intelligent medical treatment. Edge computing based on memristor neural network can realize real-time personalized diagnosis and treatment. (4) Intelligent drug delivery. The intelligent drug controlled release system based on memristors can respond to specific biological signals (such as pH or inflammatory factors) to achieve precise drug delivery. (5) Brain–computer interface. Due to the isomorphism between the sandwich structure of memristors and neural synapses, memristor-based brain–computer interfaces have significant development advantages. (6) Brain signal recording system. It should be mentioned that using memristor-based neural networks to record brain signals may be a very promising research direction. Due to the excellent compatibility between memristor neural networks and human brain neural networks, recording brain signals through memristor-based neural networks can effectively monitor abnormal discharges in the brain, thereby achieving the diagnosis of some brain diseases (Figure 8), such as epilepsy.
Figure 8
Figure 8. Memristor-based neural networks are applied for the disease diagnosis and health monitoring of brain through a brain signal recording system.
Finally, it should be mentioned that brain diseases such as cerebral infarction, brain tumors, epilepsy, stroke, etc. have become the most threatening and deadly types of diseases to human health. (53) Especially, the diagnosis and treatment of brain diseases are very complex compared to other types of diseases, as the brain is currently considered the organ with the most complex physiological structure. The most effective way to overcome these diseases is to monitor, detect, and treat them early. Therefore, there is an urgent need to develop convenient, low-cost, accurate, and patient friendly methods for brain health monitoring and disease diagnosis, in order to effectively reduce the fatal risk of such diseases. The most intuitive idea for the structure of brain monitoring devices is to mimic the structure of the human brain neural network to prepare an artificial neural network, in order to achieve a brain–computer interface with the human brain neural network. A viewpoint worth mentioning here is that the application of memristor-based neuromorphic chips in the field of neuroscience will be the best bridge between memristors and clinical medicine. (54) Because the structure and size of memristors are highly similar to biological neurons, they have natural advantages in the preparation of brain–computer interfaces. (55) Specifically, the neuromorphic computation based on memristors is very similar to the way the human brain stores and processes information, (56) which can be applied to artificial neuron and sensory memory. (57−59) Therefore, neuromorphic devices based on memristors can be effectively applied to the diagnosis and treatment of brain diseases, as shown in Figure 9, which will effectively promote the unprecedented development of neuroscience and smart medicine.
Figure 9
Figure 9. Application of memristor in neuromorphic diagnosis and treatment system.
Conclusions
Click to copy section linkSection link copied!
The application of memristors in the biomedical field is shifting from laboratory to clinical practice, demonstrating enormous potential in biosensing, neuromorphic computing, and intelligent diagnosis and treatment. In the future, with the advancement of materials science and micro-nano processing technology, memristors are expected to drive the development of a next generation of intelligent medical systems, providing innovative solutions for precision medicine and neural engineering in artificial neuron and sensory memory applications. Specifically, due to its excellent compatibility with biological neural synapses, memristors will have unparalleled development advantages in the field of neuroscience through brain–computer interface systems.
Author Information
Click to copy section linkSection link copied!
Bai Sun - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; 
https://orcid.org/0000-0002-5840-509X; Email: [email protected]
Xiaoliang Chen - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; https://orcid.org/0000-0002-5674-9884; Email: [email protected]
Jinyou Shao - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Email: [email protected]
Junchao Zhang - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China
Chuan Yang - School of Physics and Technology, Wuhan University, Wuhan 430072, China
Guangdong Zhou - College of Artificial Intelligence, Brain-inspired Computing & Intelligent Control of Chongqing Key Lab, Southwest University, Chongqing 400715, China; https://orcid.org/0000-0002-5824-9488
Linfeng Sun - Centre for Quantum Physics, Key Laboratory of Advanced Optoelectronic Quantum Architecture and Measurement (MOE), School of Physics, Beijing Institute of Technology, Beijing 100081, China; https://orcid.org/0000-0001-5851-8206
Zelin Cao - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China
Bo Yang - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China
Xiangming Li - Frontier Institute of Science and Technology, and Interdisciplinary Research Center of Frontier Science and Technology, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; Micro-and Nano-technology Research Center, State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China; https://orcid.org/0000-0001-8986-8495
B. Sun, J. Zhang and C. Yang contributed equally to this work. B. Sun, J. Zhang and C. Yang drafted the perspective manuscript. B. Sun, J. Zhang and G. Zhou made revisions. L. Sun, Z. Cao, B. Yang, and X. Li made some discussions. X. Chen and J. Shou gave some useful suggestions. All authors gave their approval to the final version of the manuscript.
The authors declare no competing financial interest.
Acknowledgments
Click to copy section linkSection link copied!
The authors gratefully acknowledge financial support from the Key Research and Development Program of Shaanxi (2022GXLH-01-12), Key Projects in Shaanxi Province (2023JC-XJ-15), Shaanxi Science and Technology Program (2025JC-TBZC-05; 2025YXYC011), Shaanxi University-Enterprise Shared Special Program (SXXZGY202401), National Natural Science Foundation of China (52375575; 52350349), and the basic research project of Xi’an Jiaotong University (xtr062025002; xtr052025010; xzy022024001; xzd012024058; xtr072024022).
This article references 60 other publications.
1
Lanza, M.; Pazos, S.; Aguirre, F.; Sebastian, A.; Le Gallo, M.; Alam, S. M.; Ikegawa, S.; Yang, J. J.; Vianello, E.; Chang, M.-F.; Molas, G.; Naveh, I.; Ielmini, D.; Liu, M.; Roldan, J. B. The Growing Memristor Industry. Nature 2025, 640, 613– 622, DOI: 10.1038/s41586-025-08733-5 1.
2
Harabi, K.-E.; Hirtzlin, T.; Turck, C.; Vianello, E.; Laurent, R.; Droulez, J.; Bessiere, P.; Portal, J.-M.; Bocquet, M.; Querlioz, D. A Memristor-Based Bayesian Machine. Nat. Electron. 2023, 6, 52– 63, DOI: 10.1038/s41928-022-00886-9 1.
3
Li, Y.; Tang, J.; Gao, B.; Yao, J.; Fan, A.; Yan, B.; Yang, Y.; Xi, Y.; Li, Y.; Li, J.; Sun, W.; Du, Y.; Liu, Z.; Zhang, Q.; Qiu, S.; Li, Q.; Qian, H.; Wu, H. Monolithic Three-Dimensional Integration of RRAM-Based Hybrid Memory Architecture for One-shot Learning. Nat. Commun. 2023, 14, 7140, DOI: 10.1038/s41467-023-42981-1 1.
4
Chen, X.; Hwang, C. S.; van de Burgt, Y.; Santoro, F. Introduction to Memristors and Neuromorphic Systems. Mater. Horiz. 2024, 11, 3462– 3464, DOI: 10.1039/D4MH90052A 1.
5
Salinga, M.; Kersting, B.; Ronneberger, I.; Jonnalagadda, V. P.; Vu, X. T.; Le Gallo, M.; Giannopoulos, I.; Cojocaru-Miredin, O.; Mazzarello, R.; Sebastian, A. Monatomic Phase Change Memory. Nat. Mater. 2018, 17, 681– 685, DOI: 10.1038/s41563-018-0110-9 1.
6
Worledge, D. C.; Hu, G. Spin-Transfer Torque Magnetoresistive Random Access Memory Technology Status and Future Directions. Nat. Rev. Elec. Eng. 2024, 1, 730– 747, DOI: 10.1038/s44287-024-00111-z 1.
7
He, K.; Liu, Y.; Yu, J.; Guo, X.; Wang, M.; Zhang, L.; Wan, C.; Wang, T.; Zhou, C.; Chen, X. Artificial Neural Pathway Based on a Memristor Synapse for Optically Mediated Motion Learning. ACS Nano 2022, 16, 9691– 9700, DOI: 10.1021/acsnano.2c03100 1.
8
Kumar, S.; Wang, X.; Strachan, J. P.; Yang, Y.; Lu, W. D. Dynamical Memristors for Higher-Complexity Neuromorphic Computing. Nat. Rev. Mater. 2022, 7, 575– 591, DOI: 10.1038/s41578-022-00434-z 1.
9
Wang, T.; Cui, Z.; Liu, Y.; Lu, D.; Wang, M.; Wan, C.; Leow, W. R.; Wang, C.; Pan, L.; Cao, X.; Huang, Y.; Liu, Z.; Tok, A. I. Y.; Chen, X. Mechanically Durable Memristor Arrays Based on a Discrete Structure Design. Adv. Mater. 2022, 34, 2106212 DOI: 10.1002/adma.202106212 1.
10
Duan, X.; Cao, Z.; Gao, K.; Yan, W.; Sun, S.; Zhou, G.; Wu, Z.; Ren, F.; Sun, B. Memristor-Based Neuromorphic Chips. Adv. Mater. 2024, 36, 2310704 DOI: 10.1002/adma.202310704 1.
11
Zhang, C.; Chen, M.; Pan, Y.; Li, Y.; Wang, K.; Yuan, J.; Sun, Y.; Zhang, Q. Carbon Nanodots Memristor: An Emerging Candidate Toward Artificial Biosynapse and Human Sensory Perception System. Adv. Sci. 2023, 10, 2207229 DOI: 10.1002/advs.202207229 1.
12
Zhu, S.; Chen, Y.; Zhou, G.; Zhao, H.; Zhang, Y.; Xu, M.; Zhao, Y.; Sun, B. In-depth Conduction Mechanism Analysis of Programmable Memristor and Its Biosynaptic Applications. Mater. Today Nano 2024, 28, 100543 DOI: 10.1016/j.mtnano.2024.100543 1.
13
Wang, S.; Gao, S.; Tang, C.; Occhipinti, E.; Li, C.; Wang, S.; Wang, J.; Zhao, H.; Hu, G.; Nathan, A.; Dahiya, R.; Occhipinti, L. G. Memristor-Based Adaptive Neuromorphic Perception in Unstructured Environments. Nat. Commun. 2024, 15, 4671, DOI: 10.1038/s41467-024-48908-8 1.
14
Jeon, K.; Ryu, J. J.; Im, S.; Seo, H. K.; Eom, T.; Ju, H.; Yang, M. K.; Jeong, D. S.; Kim, G. H. Purely Self-rectifying Memristor-Based Passive Crossbar Array for Artificial Neural Network Accelerators. Nat. Commun. 2024, 15, 129, DOI: 10.1038/s41467-023-44620-1 1.
15
Xu, Z.; Li, Y.; Xia, Y.; Shi, C.; Chen, S.; Ma, C.; Zhang, C.; Li, Y. Organic Frameworks Memristor: An Emerging Candidate for Data Storage, Artificial Synapse, and Neuromorphic Device. Adv. Funct. Mater. 2024, 34, 2312658 DOI: 10.1002/adfm.202312658 1.
16
Ren, H.; Li, F.; Wang, M.; Liu, G.; Li, D.; Wang, R.; Chen, Y.; Tang, Y.; Wang, Y.; Jin, R.; Huang, Q.; Xing, L.; Chen, X.; Wang, J.; Guo, C.; Zhu, B. An Ion-mediated Spiking Chemical Neuron Based on Mott Memristor. Adv. Mater. 2024, 36, 2403678 DOI: 10.1002/adma.202403678 1.
17
Yang, C.; Wang, H.; Wang, K.; Cao, Z.; Ren, F.; Zhou, G.; Chen, Y.; Sun, B. Silk Fibroin-Based Biomemristors