A Review of Biosensors and Their Applications

This extensive review presents different types of biosensors with nano and microscale dimensions used for biological applications. A biosensor is an analytical device, used for the detection of an analyte, that combines a sensitive biological component with a detector element (transducer) [1–3]. The sensitive biological element could be a bioreceptor, a biomimetic component, or a biologically derived material that interacts or binds with the analyte under study. A wide range of biological materials are used for the biosensors such as aptamers, tissue, microorganisms, organelles, mammalian cells, bacteria, enzymes, antibodies, and nuclei acids [3–15]. The transducer, which transforms the biological signal into a physio-chemical signal, has different detection options: electrochemiluminescence [16–19], optical [20,21], electrochemical [22,23], fluorescence [24,25], piezoelectric [26], etc., resulting from the interaction of the analyte with the sensitive biological element. Figure 1 provides an illustration of the principle of biosensors.

The use of biosensors was pioneered by Clark and Lyons [27]. In the early 1960s, they developed an electrode that could measure the oxygen concentration in blood. In this research, the electrode was introduced in perfusion blood during an open-heart surgery procedure. Biosensors have the capability to respond to specific biological signals and process them into measurable quantities for human comprehension. Recently, novel biosensors that can be implantable or wearable were developed [28]. In general, a biosensor consists of three main components. These include the biological sensing component for recognition of the biological signal, the transducer for converting the biological signal into an electric signal, and the signal processing system, which amplifies and displays the output in a desired format for human understanding [29]. The applications of biosensors are diverse; they have proven their usefulness in disease diagnosis in human healthcare delivery, agriculture, homeland security, food security, environmental, and industrial monitoring, as well as bioprocessing [28].

Biological sensors can monitor in real time, the vital signs of individual patients for a long period of time, and provide them with personalized health solutions. For achieving these goals, the biological sensors could be skin-integrated in wearable systems, or implantable medical devices. These biosensors are important to monitor different biophysical parameters of the patients such as; blood pressure, blood glucose level, heart electrical signals, pulse rate, and respiration rate [30]. The engineers and scientists have consistently collaborated to improve the biosensors' performance with emphasis on biocompatibility, specificity, reliability, durability, and consistency, as well as sensitivity for diagnosis, monitoring, and treatment of several health conditions.

In addition to real-time monitoring of patients and decisions connected to patient health complications, there is also the need for biosensors to simultaneously detect multiple analytes and/or respond to more than one stimulus, within biological fluids, within or outside the human body [29]. Unlike many other sensors used in other fields of application, biosensors (both implantable and non-implantable) are designed with unique features such as biocompatibility, biodegradability, and/or the ability to be bioresorbable, and miniaturization due to the biological environment in which they function. For instance, a skin-integrated device must be flexible, stretchable, lightweight, ultra-thin, and reliable. These devices can conform to the geometry and/or shape of the mounting surface. Furthermore, any regular movement or modification of the skin must be allowed without causing any discomfort to the patient [28].

This review paper is focused on several aspects of biosensing such as a general overview of biosensors, classification of biosensors based on the transduction methods, characteristics of biosensors, biosensor design considerations, biosensor applications, advancement in biosensors due the nanomaterials discovery, wearable biosensors, and machine learning application in biosensing.

Biosensors have the capability of responding to the presence of a specific biological analyte, and the ability to quantify the analyte, subsequently translating it into a signal that is human-readable [3,31–33]. The sensory element is biochemical in nature, and the biosensor operation is based on reactions with the analyte of interest. A biosensor consists of three main components: (1) bioreceptor–detector layer of immobilized biomaterial; (2) physio-chemical transducer, which transforms the biological response into a measured electric signal; (3) electronic system for amplifying and recording the signal for human interpretation [34,35]. The bioreceptor is simply a biological molecule that comes into direct contact with the analyte of interest. Generally, it is immobilized on the active surface of the transducer. The responsibility of the bioreceptor is to detect the presence of the analyte and subsequently bind it. This results in physiological changes, which then also modify the physicochemical properties of the transducer in the vicinity of the biological receptor. This phenomenon leads to changes in the physical properties of the transducer, which are translated into an electrical signal for human understanding [36]. These main components of the biosensor are illustrated in Fig. 1.

According to Inshyna et al. [37], biosensors can be classified based on either the type of transducer or type of bioreceptor used for their construction. According to this classification, we have electrochemical, optical, calorimetric, piezoelectric, acoustic, and electronic biosensors. Furthermore, based on the type of bioreceptors, biosensors can also be grouped as follows: enzymatic, cellular, tissular, DNA sensors, immunosensors, and aptamer sensors [34,38,39].

Electrochemical biosensors exhibit high sensitivity, selectivity, and the capability to detect. Electrochemical biosensors are based on classical electrochemistry, with the electrodes introduced in an electrolyte. The electrochemical reaction occurs at the working electrode surface between bioreceptor and analyte producing detectable electrochemical signals in terms of voltage, current, impedance, and capacitance. The devices used for electrochemical measurements are relatively simple, easily miniaturized, require low power, and are sensitive for biological applications.

In experimental electrochemistry, the three-electrode cell is one of the most common configurations used to study electrochemical reactions. It consists of a counter electrode, working electrode, and reference electrode, as illustrated in Fig. 2. The electrode where oxidation/reduction takes place (where the potential is controlled) is the working electrode. The reference electrode has a stable potential; no current flows through it. The potential of the working electrode is measured relative to it. The potentiostat is used to supply a constant potential to the working electrode, regardless of the chemical changes taking place on the working electrode at that time. Due to the advancement of microelectromechanical systems’ (MEMS) microfabrication, the classic electrochemical cells could be miniaturized at microscale [40]. Figure 3 is an illustration of three electrodes miniaturized electrochemical cells that can be used for the detection of biomarkers in blood or biological fluids.

Based on the transduction principle, electrochemical biosensors are categorized as (a) voltametric, (b) amperometric, (c) impedimetric, (d) conductometric, and (e) potentiometric [42,43]. These sensors convert the information associated with electrochemical reactions (the reaction between an electrode and analyte) into an applicable qualitative or quantitative signal [41]. In this review, we will discuss voltametric and impedimetric biosensors because there are many applications for these sensors.

The working electrode could be a simple planar electrode or contain an array of micropillars fabricated on the planar working electrode. The micropillar electrode has a larger active area for detection to enhance the sensitivity of detection (Fig. 4).

Voltametric and amperometric techniques are characterized by applying a potential to a working electrode versus a reference electrode and measuring the current. The term voltammetry is used for those techniques in which the potential is scanned over a set potential range. The current response is usually a peak or a plateau that is proportional to the concentration of the analyte. In amperometry, changes in the current generated by the electrochemical oxidation or reduction are monitored directly with time, while a constant potential is maintained at the working electrode with respect to a reference electrode. The voltametric/amperometric biosensors generally are used for enzyme-linked immunosorbent assays (ELISA) detection, and there is extensive research in this area [40,44].

Impedimetric biosensors measure changes in conductance and capacitance at the sensor surface as the selective binding of the target occurs. In this way, the impedimetric transducers can detect biological molecules or the behavior of live cells. For applications regarding live cells, the impedimetric transducer is named Electric Cell-substrate Impedance Sensing (ECIS), which is a real-time and label-free detection method to analyze the behavior of cells. The ECIS technique was pioneered by Giaever and Keese [45–47] and has been extensively studied for over two decades due to its simple structure, easy operation, and sensitivity to many cell behaviors and properties [48–55]. The ECIS technique can measure cell attachment, proliferation, migration, invasion, and cell viability because the impedance measurements are directly responding to cell attachment, growth, and proliferation. When cells attach to the electrodes, cell attachments result in additional impedance to the circuit. Their insulating properties can be detected. The impedance values gradually increase until monolayer formation and then reach equilibrium when the cells are confluent and stable. Apoptotic cells lose the dielectric properties and tend to detach from the sensing electrodes. When the cells detach or lose the dielectric properties, the measured membrane impedance will decrease. As cells grow and cover the electrodes, information about the morphology of the cells and nature of the cell attachment can be extracted from the measured impedance [52–55].

The name “impedance spectroscopy” is derived from the fact that the impedance is generally determined at different frequencies rather than just one. The impedance spectroscopy measurements are generally performed using a small Alternative Current (AC) electric field over a wide frequency range (100–100 kHz). Thus, an impedance spectrum is obtained that allows the characterization of cell size, membrane resistance and capacitance, and cytoplasm conductivity as a function of frequency.

Determining the experimental values, for the capacitance and resistance of the cell membrane, enables cell electrophysiological information collection [52–55]. As an example, blocking the ion channels enables the cell membrane to act more capacitively. The capacitance and resistance characterizing a live cell are significantly different from those of a dead cell. While healthy cells adhere more securely to surfaces, as compared to unhealthy or dead cells, this results in a stronger capacitive coupling between the cells and underlying electrodes. When the cells die, the value of impedance reduces, due to the change in the membrane capacitance and resistance.

The ECIS technique has become convenient for various applications, such as: (1) monitoring cell migration and wound repair [56], (2) monitoring cell proliferation [57], (3) real-time continuous trans-epithelial electrical resistance (TEER) measurements [58], (4) monitoring the signal transduction pathways activated by G-protein-coupled receptors (GPCR) [59], (5) studying cell differentiation events [60], (6) assessing the cytotoxicity of a variety of toxicants [61], and (7) monitoring the cell attachment and spreading [62].

Recently, a review of the ECIS technique used for applications of two-dimensional cell culture was published in the literature [63]. For instance, ECIS has important applications in the study of cancer metastasis [49,53,64]. Mathematical models are used to simulate the impedance of cell monolayer and the influence of electrode dimensions on the detection sensitivity [54,55]. The cell viability was also tested using the ECIS technique, as reported in Ref. [65].

In the past decade, investigation into stretchable electronics has been widely conducted. Dr. John A. Rogers’ group at the University of Illinois, Urbana-Champaign invented the stretchable electronics concept. According to Refs. [66,67], advanced mechanics and materials research have made possible the design of flexible electronics, such as integrated circuits with properties and functionalities comparable to acoustic sensors of the traditional silicon wafer-based ones. The advantages associated with stretchable electronic technology are its ability to conform to irregular surfaces and shapes without any distortion in their functions or deformation. Other advantages include the possibility of micro and nanoscale electronic designs that otherwise are not feasible with the conventional silicon wafer-based type. The advent of stretchable electronics has paved the way for the design of electronic eyeball cameras and deformable light-emitting displays. It is safe to say that soon, low-power sensors can be designed using flexible electronic technology together with piezoelectric materials. Figure 5 shows examples of stretchable electronics.

A stretchable biosensor with the ECIS technique was also demonstrated [68]. Authors reported that ECIS electrodes were embedded in a stretchable device and ECIS measurements on mammalian cells exposed to a cyclic strain were performed. The stretchable ECIS sensor was demonstrated to be capable of real-time monitoring the cell proliferation, while applying mechanical stimulation [68]. Figure 6 shows a stretchable ECIS sensor and details of the components.

Electrochemical biosensors have an important role in many clinical, environmental, industrial, pharmaceutical, defense, and security applications due to their superior sensitivity and selectivity. Recent developments in nanotechnology and material science, as well as being able to custom engineer the biorecognition component, will further push the development of useful and reliable electrochemical biosensor devices. The sometimes-limited shelf life and stability of the biorecognition component, as well as nonspecific binding, continue to be the biggest limitations for these types of biosensors. However, many strategies have helped to overcome or minimize these problems.

The Quartz crystal microbalance (QCM) is a classic example of a piezoelectric biosensor. A standard QCM is based on a thin quartz plate with two aligned circular electrodes fabricated at the top and bottom surfaces of the piezoelectric substrate (Fig. 7). The QCM operates using thickness-shear mode. When an AC is applied between the top and bottom electrodes, the top and bottom surfaces of the piezo plate extend in the opposite direction parallel to the plate surface, which generate a thickness shear mode that propagates through the quartz substrate. Because of the high accuracy and stability of QCM measurements [69–75], this technique has become helpful for scientists, especially for viscosity and density sensing purposes. The drawback of the QCM measurements is the damping that occurs when the biological liquid is added to the top electrode, further decreasing the QCM sensibility. Generally, the QCM resonant frequency variations are monitored in real time and give information about the process that occurs at the piezoelectric crystal surface level.

The QCM operates in the resonant frequency range between 1 MHz and 20 MHz, although at higher frequencies the QCM provides good opportunities for enhancing the sensitivity of the detection. To achieve high resonant frequency, QCM biosensors create technological complications during microfabrication, because the piezoelectric substrate must be very thin [76]. Generally, the QCM detects biological molecules by measuring the changes in resonant frequency that occur when the antigen binds to the antibody receptor [77].

QCM detection offers stability, reusability, and cost-effectiveness. There are many research papers describing the biological detection with QCM. The QCM is an extremely sensitive mass sensor, capable of measuring mass changes in the nanogram range. This piezoelectric thickness-shear-mode resonator has been successfully employed for biological sensing applications, due to the minimal damping of the acoustic wave in liquid. The QCM's sensitive surface can be functionalized with different molecules that respond to different target analytes and have been tested with a wide range of biological applications. The adhesion [78,79] and viscoelastic [80–82] properties of mammalian cell monolayers have been successfully characterized and reported using QCM resonators. QCM was previously used for virus detection. The QCM sensor was able to detect 1 nanogram of orchid virus [83,84]. QCM was also used for the detection of avian influenza virus [85].

An interesting application of the QCM is its combination with ECIS electrodes. Mammalian cells were in vitro cultured on the QCM and ECIS electrodes. This combination of ECIS and QCM devices was used as a water toxicity sensor. The mammalian cells were the sensorial element and their viability after exposure to the toxicant was monitored with both sensors: ECIS and QCM [70,71]. A commercial QCM was used to investigate live cells’ activity in water-based toxic solutions. The QCM used in this research had a resonant frequency of 10 MHz and consisted of an AT-cut quartz crystal with gold electrodes on both sides. This QCM was transformed into a functional biosensor by integrating with polydimethylsiloxane (PDMS) culturing chambers (Fig. 7) [69].

Surface acoustic wave sensors use piezoelectric materials to generate acoustic waves. Surface acoustic waves (SAW) are elastic waves that travel along the surface of the piezoelectric substrate, typically within 1–2 wavelengths from the surface and are sensitive to any modifications on its path of propagation. Anisotropic substrates like Lithium Niobate, Lithium Tantalate, Quartz, and Langasite are commonly used substrates for SAW generation. Parameters like crystal orientation, SAW velocity, coupling coefficient, permittivity and temperature coefficient of delay (TCD) determine the generation of SAW in these substrates [86].

SAW devices for liquid applications have attracted much attention due to their wide applications in the biological field [87–89]. Among various SAW devices, the wave-guided shear-horizontal (SH) SAW device, also referred to as the Love wave device, has been studied for liquid applications due to its high mass sensitivity and sensing capability in liquid environments. A Love mode surface acoustic wave device uses a piezoelectric substrate with inter-digital transducers (IDTs) to produce the surface-localized SH waves. A material layer with a slower acoustic velocity is deposited on the piezoelectric substrate to trap the acoustic energy near the surface and prevent energy loss into the bulk material. An essential application of the Love mode SAW is immunosensing specific biomarkers [87,90–93]. Other than immunosensors, SAW devices were used as mammalian cell and bacteria-based sensors [94]. In Ref. [95], a cell-based Love wave SAW biosensor was developed for real-time marine toxin testing.

A surface-horizontal surface acoustic wave (SH-SAW) device was developed for real-time monitoring of blood viscosity at room temperature. The device was tested with a liquid drop placed on the SAW delay-line path and demonstrated a sensitivity of 3.57 ± 0.3125 kHz shift per centipoise, enabling the potential for high-precision blood viscosity monitoring [96].

SAW sensors have been used in different applications. These include biosensing, chemical and gas sensing, microfluidics, and mechano-biological applications. For biological applications, SAW-based sensors are used for separating, identifying, and controlling biological targets; for instance, biomolecules, or proteins from bio-species, such as bacteria, fungi, or viruses. For chemical and gas sensing, the targeted chemical or gas molecule is attached to the functionalized surface and positioned between two electrodes. A perturbation of the resonant frequency is generated after the targeted biological material is sensed [97].

These types of transducers employ the use of ion-sensitive silicon field-effect transistors (FET). The bio-sensitive layer is placed above the surface of an ion-sensitive membrane that forms a part of the gate of the field-effect transistor. Conventionally, FET-based biosensors with receptors (e.g., antibody) immobilized on the gate region above the active channel of the FETs face an intrinsic issue, that is the severe charge screening effect in high ionic strength solutions, such as in serum or blood samples, leading to low sensitivity for direct detection of protein in the physiological environment. Several research groups have reported that conventional FET-based biosensors can effectively detect proteins in a physiological salt environment, using AC signals in drain-source voltage (Vds), in conjunction with a reference electrode, in a relatively high frequency [98–101]. An extensive review of the FET presents diverse and interesting applications of this type of biosensor [102].

The FET transistors were demonstrated to feasibly measure Aβ fibrils in human serum with concentrations ranging from 100 pM to 10 μM [103]. C-Reactive protein (CRP) was sensed from 1 fM to 100 nM, demonstrating the high sensitivity of this type of biosensor [104]. Nanowires (NW) and nanoribbon (NR) were used in the fabrication of the FET biosensor. The detection in the subthreshold regime of NW FET sensor has the merit of not only improving the conductance response and signal-to-noise ratio but also better detection limit. The research showed that the detection limit of the NW FET device was improved from ∼0.75 pM in the linear regime to ∼1.5 fM in the subthreshold regime [105]. In comparison with the nanowire structure, the larger surface-to-volume ratio of NR was demonstrated to increase the efficient surface area for detection, where the sensing elements were immobilized and therefore proportionally improved the sensitivity of the NR-FET biosensors [106].

Optical biosensors can detect luminescent, fluorescent, colorimetric, or other optical signals produced by the interaction of microorganisms with the analytes and correlate the observed optical signal with the concentration of target compounds. Optical biosensors have the biorecognition element integrated into an optical transducer system. They utilize enzymes, antibodies, aptamers, whole cells, and tissues as biorecognition elements. There are several types of optical biosensors based on optical waveguides, surface plasmon resonance (SPR), photonic crystals, and optical fibers [20,106,107]. Optical biosensors have been used for clinical diagnostics, drug discovery, food process control, and environmental monitoring.

These transducers determine the quantity of heat generated by biological material [31]. An interesting application of the temperature sensor is based on bimaterial cantilever beams, which were employed as highly sensitive temperature sensors for biological applications [108,109]. The bicantilever beam temperature sensor was fabricated from composite materials and operated in deflection mode. To achieve the high sensitivity required for the detection of heat generated by a single mammalian cell, the cantilever beam temperature sensor was microprocessed with a length at the microscale and a thickness in the nanoscale dimension [110]. As a thermogenic sample, the brown fat cells (BFCs) that are related to metabolic heat production were employed. The cantilever beam deflection was measured under a conventional microscope, when six cells of BFCs were situated about 5 µm from the tip of cantilever beam, after the BFCs were thermal stimulated with flowing norepinephrine solution (NE). When the cantilever beam senses the heat, it bends due to the difference in the coefficient of the thermal expansion of the composite structure.

A second cantilever beam was included in a vacuum chamber and operated in the resonant frequency regime. The working principle of the vibrating cantilever beam temperature sensor is based on shifts in resonant frequency in response to temperature variations generated by mammalian cells [110]. The heat sensing of a single BFC cell was demonstrated using this resonant cantilever beam. The temperature measurements were performed by stimulating the activity of BFC by introducing NE solution (1 μM) in the microfluidic chamber that contains BFC culture. The cells were introduced in the microfluidic channel and some cells spontaneously attached to the sample stage side of the cantilever beam. Then, resonant frequency measurements of the cantilever beam side situated in vacuum were performed. The heat production was observed a few minutes after adding NE, and the heat generation continued for approximately 23 min. A temperature change of 0.27 °C corresponds to 1 nJ of heat was measured.

Biosensors as analytical tools have several applications in different fields such as diagnostics, disease monitoring, and toxicity studies. The significant role of biosensors is illustrated in Table 1, which presents several biosensors with transduction methods and applications. The applications of these sensors are interesting, due to the advances and discoveries in the technological areas. There are still numerous challenges of biosensors despite many advances in technology. In some biosensors, a large amount of data are generated quickly at the output, and the analysis of these data requires further processing by an experienced user which can lead to errors. Processing by a person can take time to analyze data, which can greatly reduce the efficiency of the biosensor.

Recently, machine learning (ML) has allowed for enhancement in analytical capabilities of these various biosensing modalities [114]. ML, a subset of artificial intelligence (AI), is a framework allowing algorithms to learn automatically from data. For medical applications, it has been shown that such methods can not only significantly outperform human-engineered expert knowledge, but they are also able to discover new knowledge.

Due to their nature and mode of operation, biosensors are designed with unique characteristics and features upon which their usability and reliability depend.

Sensitivity is the most important characteristic of a sensor. It is the detection limit, which is the minimal amount (or concentration) of analyte that can be detected. This characteristic shows the capacity of the sensor to capture any fluctuations occurring in the targeted analyte if it remains in the vicinity of the sensor. Highly sensitive sensors are affected by fluctuations at low scales such as nanogram and femtogram scales [107,112,115] has been reported that the sensitivity for glucose determination ranges from 0.048 to 3.36 mA L mol⁻¹ cm².

Selectivity means that the sensor detects a certain analyte and does not react with added mixtures and contaminants. This characteristic of a biosensor is based on the ability to bind or communicate with the specific target analyte (molecule) in the presence of others in the same medium or test site. In implantable medical applications of biosensors, selectivity is one of the most important features of the device. This is because most of the analyte candidates in the bloodstream possess similar properties, and therefore, it is important that the bioreceptor part of the sensor communicate only with the analyte of interest [116,117].

The stability of a sensor refers to the signal drifting under constant conditions, which could cause errors. This feature of the biosensor ensures that it can withstand interference or noise from external factors during its operation. Noise in this case can be in the form of humidity that tends to affect the accuracy of the sensor signal in operation [118]. In addition, the temperature of the human body also impacts the effectiveness of the bioreceptor component of the sensor, thereby causing inconsistencies in the overall output of the sensor [119]. Other factors that affect the stability are the degradation of the bioreceptor over time and the affinity of the bioreceptor to the analyte.

Due to the delicate conditions under which biosensing is required, it is necessary that a biosensor produces consistent output results, under the same or similar conditions, using the same analyte. This ability to show repeatable results, whenever the sample is measured, is an important quality of the transducer [118]. Consistent calibration of the biosensors after use in accordance with the manufacturer's instructions will ensure and enhance reproducibility and consistency of results.

A biosensor's response time is the amount of time it takes to read and produce a signal after its bioreceptor meets the specific analyte [36,120]. For example, glucose oxidase–based sensors have a response time between 5 and 30 s [115].

The linearity of a biosensor is its ability to exhibit variation in its output proportional to different analyte concentrations. This is used to determine the resolution of the sensor, that is the measurement of the minimum change in the concentration of the analyte that can generate a corresponding response from the sensor. This feature is useful when sensing a wide range of concentrations for a specified analyte [118,119].

For mass production, it is important that the biosensing parameters are quantitatively validated [121,122]. The analytic hierarchy process was used, to perform a quantitative analysis of the signal produced by a carbon nanotube tube sensor, that consisted of 9.8% noise, and 10.1% error from external factors, which means that only about 80% of the normalized signal was corresponding to the real signal [122].

This characteristic of a biosensor describes the minimum quantity and concentration of analyte that the sensor can detect in the sample. This is mainly used to determine the quality of the sensor, and it is for this reason that many describe it as the most critical feature in the design and selection of biosensors. It is often determined indirectly through a linear calibration function formulated from a linear regression performed on a set of measurements of instruments against the concentration of the analyte [123].

In designing a biosensor, the principal thing is to identify and understand how the specific analyte of interest interacts with the bioreceptor (the component of the sensor that directly encounters the analyte). The following factors such as biological receptor selection, immobilization method, and transducer type must also be taken into consideration when designing a biosensor.

The biological receptor, being an important component of the sensor, is required to be highly sensitive and selective. These features enable it to (a) selectively interact with only the targeted biomolecule in the presence of others and (b) be able to capture the exact characteristics or behavior of the analyte. Understanding and acknowledging the advantages and disadvantages as well as the nuances in the biosensor applications play a major role in the selection and/or design of an appropriate receptor [124,125].

Immobilization is the process of depositing a biological molecule onto a surface, in this case, a transducer. Various methods/techniques are used in the process. In general, the choice of method used is primarily determined by the physicochemical properties of the surface under consideration [126]. Some of the methods include adsorption, entrapment, covalent attachment, microencapsulation, and cross-linking [31,127]. The most popular method used is adsorption because it is simple and inexpensive. The disadvantage is that the deposition done by this method is usually affected by experimental conditions like ionic strength, temperature, and pH due to the weak van der Waals forces forming the bonding [128,129].

The sensitivity of a biological sensor greatly depends on its transduction method. In many applications, high sensitivity and low limit of detection are requested for monitoring a particular analyte of interest. However, in applications where selectivity is important to distinguish between the target analyte and other biological components, sensitivity could be affected. Therefore, finding a good balance between these features in designing or selecting a transducer element is indispensable [127,128].

A few decades ago, when in situ testing and monitoring of biological materials was not possible, samples were taken to labs for analysis. Aside from the fact that this practice provides scientists and engineers with enough time and a convenient environment for the necessary analysis, this tends to be quite expensive and time-consuming, and requires consistent recruiting and training of highly skilled experts to achieve the testing. With the development of biosensors, their applications are in different fields such as medicines, environment, food safety, defense, and drug discovery. Figure 8 illustrates different fields of application of biosensors.

For medical applications, biosensors are designed to interact with the biomolecules of the organism they are residing in. The application of biosensors in the medical field has experienced rapid growth over the years. This is due to the ease with which it has made diagnosis, testing, and monitoring of medical conditions in patients. For instance, Electrochemical biosensing systems coupled with Graphene Quantum Dots (GQDs) biosensors have been proven to enhance the early detection of cancerous cells, as well as monitor the biomarker concentrations of the affected area. This helps medical practitioners to differentiate between normal and cancerous cells [131]. Optical biosensors are also reported by Kaur et al. [132] as an efficient tool for diagnosing different types of cancers in real time owing to their low sensitivity and low detection limit. They are also used to monitor and control blood sugar levels, as well as monitor cardiovascular conditions in real time during and after surgical procedures. Other medical applications include diabetes control and pathogen detection [133]. It was reported by Bohunicky et al. [134] that biosensors can help in early detection of cancer. This can be achieved by constantly monitoring certain protein levels as well as secretions from tumors. This will go a long way to help mitigate the pain and subsequent death of a large number of people carrying the disease.

Cardiovascular disease is the leading cause of death globally. Hence, researchers and medical practitioners have developed various techniques, such as immunoaffinity column assays, fluorometric assays, and enzyme-linked immunosorbent assays for sensing, monitoring, and managing cardiovascular problems [135–137]. However, with the use of biosensors, these expensive and time-consuming techniques can be avoided in case of early detection, monitoring, and treatment of cardiac problems. The emergence of biosensors provides a significant number of advantages over conventional diagnosis assays mentioned because they are established on electrical measurements and employ biochemical molecular recognition elements with the desired selectivity with respect to a particular biomarker under consideration [138,139].

With the adverse environmental pollution comes health problems for those in the immediate vicinity. As discussed prior, sensitivity and selectivity are key features of biosensors that could continuously scan the environment to sense the presence and the quantity of chemical agents, organic pollutants, potentially toxic elements, and pathogens that might pose a health hazard. Typical biosensors for environmental applications make use of antibodies, aptamers, nucleic acids, and enzymes as biological receptors. These biosensors include immunosensors, aptamer sensors, and enzymatic biosensors. Biosensors can also detect pollutants by measuring color, light, fluorescence, or electric current as reported by Refs. [140–145].

In the food industry, biosensors are employed as quality assurance tools. They are useful from the crop-growing stage to the processing of the food. Biosensors are also used for automation in food processing facilities for sorting food and reducing the time and cost of this food process. Biosensors can detect low concentrations of different chemicals in food. Moreover, biosensors are used to detect, monitor, and quantify the contamination of food from other sources such as container surfaces [146–149].

Using nanomaterials, the sensitivity and performance of biosensors can be significantly improved [150]. Nanomaterials with at least one of their dimensions varying in a range from 1 to 100 nm display unique properties as compared to their bulk because their nanometer size enhances their physical properties (electrical, electrochemical, optical, and magnetic). Generally, the nano biosensors use nanomaterials in combination with microscale-size transducers.

The nanoparticles used for biosensors are nanoparticles with the capability of high detection sensitivity such as AuNPs, AgNPs, quantum dots (QDs), magnetic nanoparticles, nanowires, nanotubes or nanoribbons with very high surface area, nanomaterials with high electron conductivity such as carbon nanotubes (CNTs), Au nanowires, thin films at nanoscale such as graphene, polymeric materials as dendrimers, and photonic crystals (PC) [150]. An extensive presentation and classification of nanosensors showing their characteristics, detection mechanism, and application of biological nanosensors is given in Ref. [151]. Examples of nano biosensors’ applications include the environment as well as the industrial sites, detection of ultra-low concentrations of potentially dangerous substances, detection of biochemicals in cellular organelles and medical diagnosis, and monitoring of physical and chemical phenomena in humanly unfriendly environments [151].

The research for nanosensors started in the last decade, and the progress shown in this area has been remarkable. Currently, there is extensive research on the nanoscale phenomena, discovery of nanomaterials, and continuing progress in nanotechnology fabrication tools that will generate further achievements in the field of nanosensors. This can be achieved through the enhanced performance of existing nanosensors and the discovery of novel types of nanosensors based on novel nanomaterials and detection techniques. Table 2 presents the main types of nanomaterials used for nanosensors, transduction methods, and applications of biological nanosensors.

Wearable biosensors and smartwatch technologies have taken health, exercise, and sleep monitoring to a few notches higher in recent years. Their emergence has made the measurement of a combination of physiological data, such as heart rate, heart rate variability, respiration rate, and other vital signals easy to archive due to their user-friendliness. With wearable biosensors, the physiological conditions of the patients are continuously monitored in real-time [175], as illustrated in Fig. 9. These wearable devices worn by patients enable medical personnel to take necessary actions before the effects of the change of the monitored parameters get dangerous. Modern optical and electrochemical-based wearable sensors are made of flexible electronics, wireless data transmission, and self-powering capability [177,178]. Gao et al. [179] have designed a fully integrated wearable sensor for monitoring perspiration to detect glucose, Na⁺, and K⁺ in patients. Yetisen et al. [180] developed a minimally invasive dermal tattoo realized on the skin-based colorimetric biosensor, that is able to detect changes in glucose and pH. Another interesting idea is to detect glucose levels in the tears. Elsherif et al. [181] have developed a wearable contact lens-based sensor for monitoring real-time glucose levels. Several tattoo-based biosensors have been researched with relatively minimal invasion procedures. In this technique, different types of dyes that changed colors in response to pH levels and other stimuli were used [182,183].

A low-cost and high-accuracy wearable piezoelectric-based sensor for blood pressure monitoring was also developed [184]. The piezoelectric sensor was mounted on the subject's wrist above the radial artery and was demonstrated as suitable for continuous, long-term blood pressure-monitoring applications.

Yapici et al. [185] took advantage of the impeccable conductive properties of graphene-clad textiles to develop a fully wearable, cloth-based smart medical garment for monitoring in real-time and for long periods of time the electrocardiogram (ECG) activity of the human body. Figure 10 illustrates the placement of the ECG sensors on the left and right hand and the recording of the electrical signal of the heart [185].

Piezoelectric ZnO was used for the fabrication of a wearable and stretchable piezoelectric power generator, that could harvest energy from the movements of the body (Fig. 11) [186]. The micro-fabricated power generator had a thickness in the micrometer scale to be attached to the skin or garments and stretched by the natural movements of arms, legs, or neck. This power generator demonstrated a maximum power output of 200 µW. It is expected that energy harvested by this device could power wearable sensors [157]. Obviously, wearable sensors with their current capabilities and esthetic packaging features have come to stay, and they will continue to be the basis on which further biosensing devices and their applications will be developed.

Data processing and interpretation is an integral part of biosensing. Conventionally, mathematical models are the tools deployed in data processing in most biosensors. However, for multianalyte-based sensors, this technique has become limited due to their inability to process data beyond two and three dimensions [188].

This has led to the introduction of more sophisticated data processing techniques such as preprocessing and machine learning algorithms in modern biosensing technologies. Neural and non-neural network models are subdivisions of machine learning algorithms, typically used in the design of biomedical signals [188,189].

Like many data mining projects, careful cleaning of the raw data or signal received by a biosensor is critical in enhancing accuracy and efficiency. Typical data manipulation techniques used in the preprocessing stage are normalization, standardization, image analysis, transformation, as well as data metric synthesis [190]. The number and type of preprocessing techniques required largely depend on the nature of the raw data received.

Several choices of non-neural ML algorithms are used in the processing of biomedical signals detected by biosensors. The most popular ones are Principal Component Analysis (PCA) and Linear discriminant analysis (LDA), and Random Forests (RFs), hierarchical cluster analysis (HCA), support vector machine (SVM), and decision trees (DTs). The schematic illustrations of these algorithms are shown in Fig. 12.

The Neural Algorithm is identical to the neuron network of living organisms, such as the human brain. This algorithm consists of a structured network of artificial neurons that are activated one after the other in a defined pattern [188]. The amount of data required to train a neural network is huge, and this is an inherent disadvantage as compared to the non-neural type of algorithm. In most cases, it is challenging to incorporate neural networks into biosensors for the purposes of data collection for physiological monitoring biosensors. Wearable and noninvasive biosensors have helped overcome the issue. Neural networks are subdivided into categories: (1) the Feedforward Neural Network (FNN), (2) Recurrent Network (RNN), and (3) Convolution Network (CNN). Figure 13 illustrates the latest architecture of the Neural Network [193].

In conclusion, the application of machine learning in biosensing technology, and its applications will certainly enhance the need to make them autonomous while improving their accuracy.

This review paper presented biosensors, their applications, and recent advancements in the use of Machine Learning techniques. Several types of biosensors with their transduction methods were discussed with a focus on electrochemical, thermometric, QCM and acoustic biosensors, and wearable biosensors.

Nanomaterials used for biosensing were also introduced. Nanomaterials became important components in bioanalytical devices since they were demonstrated to enhance the performance of biosensors in terms of sensitivity and detection limits down to single-molecule detection. The specific properties of nanomaterials could improve the performance of the classic transduction methods. The combination of different nanomaterials, each with different characteristics, could increase the performances of biosensors even more. Due to the vast number of different nanomaterials, all with their specific properties, only a few examples of nano biosensors were mentioned in this review paper to highlight the principal advantages of such materials. We also discussed biosensors based on nanomaterials, AI, ML, their challenges, and the potential to revolutionize the biosensors industry. Technological advancement in the field led to the development of wearable devices for continuous and real-time monitoring of physiological signals. Their successful applications depend less on technological issues and more on the challenges of practical ethical, legal, and social issues, regarding data collection and usage.

The biosensor industry is interested in integrating the miniaturized biosensors in personalized point-of-care (POC) devices, because of their reliable, sensitive, and fast detection of biomolecules with portable features. The recent advances in the camera and processor of smartphones are being integrated with biosensing applications, and this could increase the precision of healthcare diagnoses. Recently developed wireless technologies are better than their wired counterparts, but they require more innovation in the design of low-power ICs. IC fully integrated within the smartphone eliminates the need for external hardware and the processor unit, so there is a need for more fully integrated smartphone-based sensors for personalized health care. In conclusion, the combination of biosensors, nanomaterial-based technology, and smartphone technology can improve POC-based devices for better health care.

The authors wish to thank S.M. Mahbobur, Caroline Miko, Sarah Gudelis, Ward McHenry, Varun Modak, Arin Ofir, and Anna Taylor of the Spring 2022 ECE Senior Design team of Virginia Tech.

All authors have read and agreed to the published version of the manuscript.

The authors declare no conflict of interest.

No data, models, or code were generated or used for this paper.