Screening and Biosensor-Based Approaches for Lung Cancer Detection

Early diagnosis of lung cancer helps to reduce the cancer death rate significantly. Over the years, investigators worldwide have extensively investigated many screening modalities for lung cancer detection, including computerized tomography, chest X-ray, positron emission tomography, sputum cytology, magnetic resonance imaging and biopsy. However, these techniques are not suitable for patients with other pathologies. Developing a rapid and sensitive technique for early diagnosis of lung cancer is urgently needed. Biosensor-based techniques have been recently recommended as a rapid and cost-effective tool for early diagnosis of lung tumor markers. This paper reviews the recent development in screening and biosensor-based techniques for early lung cancer detection.


Introduction
Lung cancer is a major health problem in the United State and worldwide [1]. Every year, approximately 1.3 million new lung cancer cases and about 1.2 million lung cancer deaths occur in Europe and North America [2][3][4]. Previous studies have indicated that tobacco smoke [5], environmental pollution [6], second-hand smoke [7], industrial substances [8], and genetic factors [9] may cause lung cancer. Compared to some other common cancers such as breast cancer, lung cancer continues to have a much lower survival rate [10]. Early diagnostic of lung cancer with suitable treatment significantly improves the five-year survival rate [11]. Chemotherapy and radiation therapies are commonly applied for small cell lung carcinoma (SCLS) [12], while surgical treatments are normally provided for non-small cell lung carcinoma (NSCLS) [13]. Eastern Europe has the highest lung cancer mortality rate in males, while Northern Europe and America have the highest mortality rate in females [14]. Increases in new lung cancer cases are expected in some developing countries such as China and India in the next few years [15].
Researchers have extensively studied lung screening methods, including chest radiograph (CRG), computed tomography (CT), low-dose CT (LDCT), magnetic resonance imaging (MRI), and positron emission tomography (PET). These techniques have some drawbacks, such as being expensive and having low sensitivity for identifying cancer cells at early stages. Annual CRG was reported as not helpful in reducing the mortality of lung cancer [16]. CT has been considered as the gold standard lung screening tool, which offers information of tumor features such as size, characterization and tumor growth. 3D CT image offered assessment of the chest wall, diaphragm, and mediastinum invasion, in addition to staging of the tumor. Radiations produced from CT also increased the cancer risk [17]. To solve this limitation, LDCT was applied for lung imaging and it reduced 20% of lung cancer mortality [18]. However, LDCT continues to have a high false positive rate (up to 96.4%) [19]. 18F-Fluorodeoxyglucose PET/CT was applied in oncological imaging but produced inaccurate results [20,21]. Magnetic induction tomography (MIT) has been recently proposed for early disease detection with advantages of low-cost and high-sensitivity [22]. Apart from imaging approaches, biopsy is another common way to identify lung cancer; however, it is expensive and requires trained physicians [23]. Autoantibodies can detect lung cancer cells about five years earlier than autoradiography because tumor growth is associated with gene and protein changes [24]. Biomarker-based techniques for early diagnosis of tumor markers have attracted much attention [25]. The major limitations of these techniques include being time-consuming and expensive, and having low-sensitivity for low marker concentrations [26]. Therefore, it is necessary to develop a rapid, low-cost and high-sensitive method for early diagnosis of lung cancer. In recent years, various biosensors have been developed to analyze tumor markers for early diagnosis of various diseases including lung cancer [27][28][29][30].
This paper summarizes the recent achievements in screening and biosensor-based approaches for lung cancer detection. Several MIT sensor systems, including their benefits and limitations as well as future research directions are also reviewed. The paper is structured as follows: Section 2 presents the current available lung screening approaches; Section 3 describes the MIT based approaches for imaging of biological objects; Section 4 reviews some recently developed biomarker and biosensor based techniques for lung cancer detection; Section 5 presents some current trends and future perspectives of lung cancer detection techniques; and Section 6 concludes this paper. Table 1 compares some most popular screening methods for imaging of lung. Abnormal chest imaging is the conventional diagnostic method for lung cancer detection, which can be done by CT, LDCT, MRI, PET and bone scans [31]. CRG is the conventional screening modality for diagnosis of lesion and pulmonary nodule in subjects with high risk of lung cancer. It employs X-rays to produce lung images, which are associated with low sensitivity, high radiation exposure dose, and poor image quality. CRG is not helpful in reducing the mortality of lung cancer. Several clinical trials were conducted to investigate CRG and CRG with sputum cytology in the 1970s. The obtained results showed that both approaches have poor sensitivity and are not helpful in reducing lung cancer mortality [32,33].

Clinical Lung Screening Modalities
CT can detect features of nodules such as characterization, size and tumor growth. 4D CT significant impacts lung cancer management by allowing more precise targeting of the administered radiation [34]. The major drawbacks of CT are cost, unsafe radiation exposure, high false-positive rate, and cannot be performed on a routine basis. The radiations produced from CT also increase the risk of cancer in children because they are more sensitive to radiation-induced carcinogenesis [35]. Additionally, there was poor correlation relationship between CRG and CT [36].
LDCT produces much lower radiations than CT, which has potential to become an effective tool for early lung cancer detection in high risk individuals [37,38]. The iterative image approach was developed to reduce dose in LDCT [39]. Lung LDCT was associated with a reduced mortality in high-risk population [40], while annual LDCT increased lung tumor incidence [41]. In a previous report, studies showed that LDCT detects more early-stage lung nodules and cancer cells than CRG, but it is not helpful in reducing lung cancer mortality [42][43][44]. Up to 90% of new lung cancer cases were detected using LDCT [45]. The national comprehensive cancer network published guidelines

MRI
MRI is a powerful tool for lung imaging without ionizing radiation. However, it was not recommended for regular lung screening because it provides insufficient anatomic information, and is time-consuming and expensive. Published results showed that lung MRI detects up to 90% of nodules 4-8 mm in diameter and 100% of nodules >8 mm in diameter [57]. The respiratory-gated proton MRI was investigated for small mice neoplastic lesion detection [58]. Investigators also studied the optimized proton MRI sequences for imaging of lung [59].
MRI with ultra-short echo-time (UTE) pulse sequence is extremely efficient for limiting motion artifacts. MRI with UTE helps to enhance the MRI signal intensity of pulmonary tissue and reduces lung susceptibility artifacts. Researchers have investigated MRI with gradient echo and MRI with T 2 -weighted fast spin echo for diagnosis of lung nodules (3-4 mm in diameter) [60]. It was found that MRI with UTE is more sensitive than MRI with 3D dual-echo GRE for small pulmonary nodule detection (4-8 mm) [61]. However, further investigations are needed to improve the spatial resolution and robustness of lung MRI.
Lung MRI has a higher false-positive rate (up to 95%) than LDCT, and the detection rate and sensitivity of MRI are satisfactory. It is hard to detect lung nodule with MRI due to the low intensity of nodule. Fink et al. [62] investigated the sensitivity of MRI for lung nodule detection by using different pulse sequence. Cieszanowski et al. [63] applied MRI with T 1 -weighted sequence and MRI with T 2 -weighted sequence for diagnosis of small lung nodules, respectively. Their research findings showed that the false-positive rate is affected by types of sequence. MRI comprising two or three sequences is helpful in reducing the false-positive rate. It was found that 3T MRI has more difficulty detecting ground glass opacities than 1.5T MRI [64]. 1.5T MRI with SSFP sequences successfully detected ground glass opacities in 75% of lung fibrosis subjects [65]. MRI with T 2 -weighted fast spin echo for ground glass infiltrates detection was similar or even better in immunocompromised subjects [66].

Breath Test
Breath analysis or breath test is one of the most noninvasive screening approaches for early diagnosis of lung cancer, which has gained attention due to its easy and fast data collection process [67,68]. Breath gas mainly contains nitrogen and carbon dioxide produced by respiration. It consists of more than 100 gas species with different concentrations, which may provide useful information for early diagnosis of disease [69]. However, breath test has some drawbacks, such as insufficient accuracy because exhaled air has many crucial volatile organic compounds (VOCs) at very low concentrations and there is no clear protocol for breath sampling [70]. To solve these challenges, many researchers have investigated gas chromatography-mass spectrometry (GC/MS), however, this method is expensive and it is not easy to build an implementation system. VOC patterns (multiple VOCs) have been reported as useful biomarkers for exhaled breath analysis for disease detection [71][72][73]. Various algorithms have been developed to measure VOC patterns for lung cancer detection, including forward stepwise discriminant analysis [74], support vector machine (SVM) [75], weighted digital sum discriminator [76], logistic regression [77], partial least-squares regression [78] and random forest classification [79].
Recently, Sakumura et al. [80] developed a prototype system to measure exhaled breath made of simple GC columns, a simple gas-condenser unit and SnO 2 -based semiconductor gas sensors. To find the most effective combination of VOCs for lung cancer detection, they analyzed the VOCs measurements from lung cancer patients and healthy subjects. They also performed a subsequent statistical analysis to detect lung cancer by combining various VOCs. Their research findings indicated that the five-element VOC pattern of CHN, methanol, CH3CN, isoprene, and 1-propanol is sufficient for 89% screening accuracy [69].

Magnetic Induction Tomography and Measurement Systems
Recently, MIT has been proposed as a new cost-effective imaging tool for diagnosis of various diseases including lung cancer and brain stroke [81][82][83]. Similar to electrical impedance tomography (EIT), MIT measures the electromagnetic properties of biological object. Compared to EIT, MIT can obtain more information on insulating tissue such as ribcage [84]. As shown in Figure 1, a MIT system normally contains several coils (excitation coils and detection coils) arranged around the object to generate and measure the scattered magnetic field, and a host computer with matched software to produce images [85]. During data collection, an excitation coil generates primary field B 0 in the conductive medium, the induction of eddy currents accompanies the interaction in the medium itself as the primary field propagates and penetrates the medium. The secondary field ∆B is also known as the magnetic perturbation field [86].
Al-Zeibak et al. [87] applied the MIT theory to distinguish fat tissue and water-bearing fat free tissue. Most MIT hardware systems are dependent on external devices and power amplifiers to produce high-power signals with less noise. MIT-based techniques have not been extensively investigated in clinical environments due to the limited image resolution, and they have not met the standard for widespread commercialization. To solve these problems, investigators have developed several approaches with particular focus on the development of clinical applicable MIT systems.

Breath Test
Breath analysis or breath test is one of the most noninvasive screening approaches for early diagnosis of lung cancer, which has gained attention due to its easy and fast data collection process [67,68]. Breath gas mainly contains nitrogen and carbon dioxide produced by respiration. It consists of more than 100 gas species with different concentrations, which may provide useful information for early diagnosis of disease [69]. However, breath test has some drawbacks, such as insufficient accuracy because exhaled air has many crucial volatile organic compounds (VOCs) at very low concentrations and there is no clear protocol for breath sampling [70]. To solve these challenges, many researchers have investigated gas chromatography-mass spectrometry (GC/MS), however, this method is expensive and it is not easy to build an implementation system. VOC patterns (multiple VOCs) have been reported as useful biomarkers for exhaled breath analysis for disease detection [71][72][73]. Various algorithms have been developed to measure VOC patterns for lung cancer detection, including forward stepwise discriminant analysis [74], support vector machine (SVM) [75], weighted digital sum discriminator [76], logistic regression [77], partial least-squares regression [78] and random forest classification [79].
Recently, Sakumura et al. [80] developed a prototype system to measure exhaled breath made of simple GC columns, a simple gas-condenser unit and SnO2-based semiconductor gas sensors. To find the most effective combination of VOCs for lung cancer detection, they analyzed the VOCs measurements from lung cancer patients and healthy subjects. They also performed a subsequent statistical analysis to detect lung cancer by combining various VOCs. Their research findings indicated that the five-element VOC pattern of CHN, methanol, CH3CN, isoprene, and 1-propanol is sufficient for 89% screening accuracy [69].

Magnetic Induction Tomography and Measurement Systems
Recently, MIT has been proposed as a new cost-effective imaging tool for diagnosis of various diseases including lung cancer and brain stroke [81][82][83]. Similar to electrical impedance tomography (EIT), MIT measures the electromagnetic properties of biological object. Compared to EIT, MIT can obtain more information on insulating tissue such as ribcage [84]. As shown in Figure 1, a MIT system normally contains several coils (excitation coils and detection coils) arranged around the object to generate and measure the scattered magnetic field, and a host computer with matched software to produce images [85]. During data collection, an excitation coil generates primary field B0 in the conductive medium, the induction of eddy currents accompanies the interaction in the medium itself as the primary field propagates and penetrates the medium. The secondary field ∆ is also known as the magnetic perturbation field [86].
Al-Zeibak et al. [87] applied the MIT theory to distinguish fat tissue and water-bearing fat free tissue. Most MIT hardware systems are dependent on external devices and power amplifiers to produce high-power signals with less noise. MIT-based techniques have not been extensively investigated in clinical environments due to the limited image resolution, and they have not met the standard for widespread commercialization. To solve these problems, investigators have developed several approaches with particular focus on the development of clinical applicable MIT systems.

Gradiometer
Gradiometer is attractive for magnetic field measurements in noisy environments, which is commonly applied in MIT systems to measure the gradient of magnetic field. Axial gradiometer, planar gradiometer (PGRAD) and asymmetrical gradiometer are the three main types of gradiometers. The axial gradiometer has been employed in various MIT systems to remove the primary field. Karbeyaz et al. [88] applied a single coaxial gradiometer to move over the phantom using the MIT system. Riedel et al. [89] investigated the precision and sensitivity of an axial gradiometer containing five PCBs covered with shielding layers to avoid capacitive coupling.
Xu et al. [90] developed a multi-channel hemispherical system with several modifications to the circuit of Riedel's system. They improved the stability of the Riedel's system by using a cancellation senor to reduce the phase drift. They also used an amplifier with high CMRR for capacitive coupling rejection, whereas shielded cables were applied to remove unwanted signals. Various PGPADs have been developed for MIT applications; the thin-film PGRAD is more attractive to measure the magnetic field, and can be fabricated with high intrinsic balance owing to the precision photolithographic methods applied to fabricate these devices.
Ketchen et al. [91] first developed a thin-film PGRAD, which was made of parallel and series configured pickup loops coupled to the superconductivity quantum interference device (SQUID) inductance. Since then, various planar gradiometers with baseline less than 2 cm have been investigated. Stolz et al. [92] developed a PGRAD with two series-configured pickup loops transformer coupled to a thin-film SQUID. Cantor et al. [93] developed a similar long-baseline planar gradiometer.
Scharfetter et al. [94] developed a coil PGRAD system with the receiver coil and excitation coil located at the same position. The primary signal was subtracted from the captured signal (measured by reference coil). The signal to noise ratio was improved significantly by applying the planar gradiometer. Rosell et al. [95] evaluated the coil PGRAD system analytically and experimentally. PGRAD offered a robust and stable cancellation technique capable of reducing carrier signal in the absence of conductivity perturbations while maintaining essentially the same absolute sensitivity for local perturbations. PGRAD system required fewer electronic devices than the coil system. Compared to PGRAD system with 16 coils, the PGRAD system with 16 gradiometers has lower sensitivity.
Scharfetter et al. [96] investigated the PGRAD and solenoid coil using the single channel MIT system. PGRAD was less sensitive to far-field electromagnetic interference and produced phase errors due to the thermal mismatches between gradiometers halves. The same research team also developed a multi-channel MIT imaging system including zero flow gradiometer (ZFGRAD) that combines the benefits of PGRAD and zero flow coil (ZFC) [97]. ZFGRAD offered better immunity to far magnetic perturbation than PGRAD and ZFC (up to 12 times). ZFC and ZFGRAD exhibited their maximum sensitivities near the tank borders on the side of the excitation coil, whereas the PGRAD achieved more sensitivity near the receiver side.
Merwa et al. [98] developed a MIT system comprising 16 excitation coils and 32 receiving coils. The MIT system with 32 receiving coils offered good localization of perturbed sphere. The system with 16 PGRADs offered better localization; however, two places had ghost images with opposite sign mirrored true image with respect to the x-y plane. The point spread function was applied to solve the "location unrecognized" problem. Such function defined the propagation of image due to a point source or object as related to the location and geometry of the perturbation [99].

Excitation Coil
Coil sensor plays an important role in MIT systems, which is sensitive to the flux. Many sensors have been investigated for implementing MIT systems, such as Rogowski coils, gradiometers, vibrating coils, tangential field sensors and needle probes [100]. Air-core coil and ferromagnetic core coil are the most commonly available excitation coils, and helpful in reducing the effect of the primary field. The ferromagnetic core coils can overcome the low sensitivity of air-core coil and act as flux concentrator. Various types of receiving coils have been developed to reduce the effects of primary field, such as excitation and receiver coil with back-off coil, planar array, axial gradiometer, PGRAD-differential coil, ZFC-receiver coils, and ZFPGRAD-planar gradiometer [101,102]. Stawicki et al. [103] developed an excitation coil that contains a conducting shield to protect the primary field. The ferrite core was located at the center of screen and it could concentrate the primary field in the core material itself; such design increased the magnetic field. Barba et al. [104] used a similar measurement system to conduct their investigations.

Sensor Arrangement
The image quality of MIT system is highly dependent on the sensor array configurations. Watson et al. [105] recommended that the primary field from the planar array can be done with a sensor coil. Igney et al. [106] enhanced Watson's work by using a shielded PCB printed coil. Such design improved the efficiency of the insensitivity to the primary field effects. Eichardt et al. [107] investigated the performance of MIT systems by using different sensor arrays including cylindrical and hemispherical sensor arrays. Compared to the cylindrical system, the hemispherical MIT system offered higher sensitivity, and larger coil sensors were more sensitive in relation to standard setups. The sensitivity was increased by increasing the distance between each two-coil element.
Gursoy et al. [108] studied six sensor arrays and they found that the sensor arrays affect stability and image quality significantly. They also developed a fast-deterministic algorithm to optimize the sensor arrays, which has the ability to produce the most suitable sensor array, even though it does not guarantee finding the global optimum. Scharfetter et al. [109] proposed an active marker system for artifact suppressions during object movements. The system contained an elastic belt with several small loops that controlled by the data collection unit. More accurate images were obtained when the assumed SNR close to the determined. Results showed that MIT is possible for tracking of object boundaries. However, future investigations are needed with particular focus on the development of biomarkers and the optimization of working frequencies.

Recent Development of MIT
MIT has recently been applied for medical applications with particular focus on imaging of lung, brain, heart, liver tissue and biological tissues [110][111][112][113][114]. Gabriel et al. [115,116] investigated various biological tissues over a wide frequency range. The simplicity of characterizing passive electrical properties of biological tissues offered an alternative way for imaging of biological objects. Table 2 compares various newly developed MIT systems. Patz et al. [117] developed a MIT with direct digitizing signal measurement (DDSM) module to detect cerebral stroke. The National Instrument (NI) PXI system was applied to conduct the experimental work with working frequency of 10 MHz. The amplifier was applied to improve the measurement phase delay. Referring to Figure 2, researchers at the University of Bath have developed a 16-channel MIT system for biomedical applications [84]. In their experimental setups, the NI system was applied to accomplish the signal driving, switching and data acquisition; such design simplified the system while providing satisfactory performance. The system was experimentally validated to detect a saline bottle. Results showed that the MIT system has potential for biomedical imaging applications [118]. Watson et al. [105] developed a MIT with phase-stable amplifier for biomedical application. The phase-stable amplifiers and the gradiometers configurations need to be mutually exclusive, and the highest measurement precision could be achieved by utilizing both approaches. The MIT image quality would be improved by increasing the number of coils, however, such method also increased the system cost, complexity and operation time. A rotational MIT system containing a transceiver RF coil was developed for biomedical application. Compared to conventional systems, the proposed rotational system offered a better field penetration depth towards the center of image. The exiting MIT systems suffer from several limitations. The capacitive coupling between excitation and receiving coils affects the measured data from receivers; therefore, it is important to eliminate capacitive coupling to represent the actual results [119].
More recently, Wang et al. [120] developed a numerical model of holographic electromagnetic induction (HEI) system for biomedical applications with particular focus on lung cancer detection based on MIT technique. The system (see Figure 3) is made of 16 coils, and each coil worked as transmitter and receiver. Various simulations were conducted with several realistic human thorax models and results demonstrated that HEI could detect arbitrary shaped lung tumors with random sizes and locations. The research outcomes offered crucial priority information that can be exploited to improve MIT based approaches. Watson et al. [105] developed a MIT with phase-stable amplifier for biomedical application. The phase-stable amplifiers and the gradiometers configurations need to be mutually exclusive, and the highest measurement precision could be achieved by utilizing both approaches. The MIT image quality would be improved by increasing the number of coils, however, such method also increased the system cost, complexity and operation time. A rotational MIT system containing a transceiver RF coil was developed for biomedical application. Compared to conventional systems, the proposed rotational system offered a better field penetration depth towards the center of image. The exiting MIT systems suffer from several limitations. The capacitive coupling between excitation and receiving coils affects the measured data from receivers; therefore, it is important to eliminate capacitive coupling to represent the actual results [119].
More recently, Wang et al. [120] developed a numerical model of holographic electromagnetic induction (HEI) system for biomedical applications with particular focus on lung cancer detection based on MIT technique. The system (see Figure 3) is made of 16 coils, and each coil worked as transmitter and receiver. Various simulations were conducted with several realistic human thorax models and results demonstrated that HEI could detect arbitrary shaped lung tumors with random sizes and locations. The research outcomes offered crucial priority information that can be exploited to improve MIT based approaches. Watson et al. [105] developed a MIT with phase-stable amplifier for biomedical application. The phase-stable amplifiers and the gradiometers configurations need to be mutually exclusive, and the highest measurement precision could be achieved by utilizing both approaches. The MIT image quality would be improved by increasing the number of coils, however, such method also increased the system cost, complexity and operation time. A rotational MIT system containing a transceiver RF coil was developed for biomedical application. Compared to conventional systems, the proposed rotational system offered a better field penetration depth towards the center of image. The exiting MIT systems suffer from several limitations. The capacitive coupling between excitation and receiving coils affects the measured data from receivers; therefore, it is important to eliminate capacitive coupling to represent the actual results [119].
More recently, Wang et al. [120] developed a numerical model of holographic electromagnetic induction (HEI) system for biomedical applications with particular focus on lung cancer detection based on MIT technique. The system (see Figure 3) is made of 16 coils, and each coil worked as transmitter and receiver. Various simulations were conducted with several realistic human thorax models and results demonstrated that HEI could detect arbitrary shaped lung tumors with random sizes and locations. The research outcomes offered crucial priority information that can be exploited to improve MIT based approaches.

Biomarkers for Lung Cancer Detection
Biomarkers predict the response to certain types of therapy such as surgery and chemotherapy or estimate the risk of future relapse. Genetic and proteomics-based biomarkers are two major types of biomarkers, which can be identified through tumor cells, urine, sputum, blood, or other body fluids [3]. Table 3 demonstrates numerous lung cancer markers . Table 3. Lung cancer markers.
CEA is a common proteomic biomarker to distinguish malignant tissue and benign tissue. The CEA level range of 2.5-5 ng/mL was observed in healthy subjects, and the highest level of CEA offered a useful prognostic indicator [142]. NSE was applied as the putative serum marker of SCLC with relatively higher sensitivity and specificity compared to CEA [143,144]. A subject is suspected suffering from SCLC if the NSE level is greater than 35 ng/mL. Annexin II and ENO1 are other common lung cancer biomarkers [145]. Up to date, it is still not possible to detect lung cancer with a specific biomarker, as most identified markers are nonspecific indicators. Therefore, a protein biomarker panel was applied for the accurate detecting of disease consisting of CEA, retinol binding protein (RBP), R1-antitrypsin (AAT), and squamous cell carcinoma (SCC) antigen. The proteomicsbased biomarker panel detected about 88% of subjects with lung cancer and 82% of subjects without cancer [146]. The sensitivity was improved with a biomarker panel containing CEA and some specific biomarkers such as ENO1, SCC, NSE, and CYFRA21-1 [147].

Biomarkers for Lung Cancer Detection
Biomarkers predict the response to certain types of therapy such as surgery and chemotherapy or estimate the risk of future relapse. Genetic and proteomics-based biomarkers are two major types of biomarkers, which can be identified through tumor cells, urine, sputum, blood, or other body fluids [3]. Table 3 demonstrates numerous lung cancer markers . Table 3. Lung cancer markers.
CEA is a common proteomic biomarker to distinguish malignant tissue and benign tissue. The CEA level range of 2.5-5 ng/mL was observed in healthy subjects, and the highest level of CEA offered a useful prognostic indicator [142]. NSE was applied as the putative serum marker of SCLC with relatively higher sensitivity and specificity compared to CEA [143,144]. A subject is suspected suffering from SCLC if the NSE level is greater than 35 ng/mL. Annexin II and ENO1 are other common lung cancer biomarkers [145]. Up to date, it is still not possible to detect lung cancer with a specific biomarker, as most identified markers are nonspecific indicators. Therefore, a protein biomarker panel was applied for the accurate detecting of disease consisting of CEA, retinol binding protein (RBP), R1-antitrypsin (AAT), and squamous cell carcinoma (SCC) antigen. The proteomics-based biomarker panel detected about 88% of subjects with lung cancer and 82% of subjects without cancer [146]. The sensitivity was improved with a biomarker panel containing CEA and some specific biomarkers such as ENO1, SCC, NSE, and CYFRA21-1 [147].
CYFRA21-1 protein has been recognized as the most sensitive biomarker to identify NSCLC, especially squamous cell carcinoma [148]. Previous research findings have confirmed the sensitivity and specificity of CYFRA21-1 as a tumor marker [149]. CYFRA21-1 gene has been reported as the most robust DNA-based biomarker for NSCLC [150]. Subjects with advanced NSCLC observed high-level serum CYFRA21-1 [151]. NSCLC patients with high-level serum CYFRA21-1 are poor prognosis. High serum CYFRA21-1 level may be a useful noninvasiveness marker to identify NSCLC risk, however, this statement requires further clinical investigations.

Gene Biomarkers
This section reviews the most commonly available genetic lung cancer biomarkers, including p53, p16, K-ras, telomere length and telomere-related genes and microRNAs. p53 mutation occurred in 50% of subjects with NSCLCs and the spectrum changes around 34-82%. p53 expression occurred in about 58% of subjects with lung cancer [152]. Additionally, there was a correlation between bcl-2 and p53 over-expression for lung cancer patients [152].
p16 plays a crucial role in regulating cell cycle. p16 methylation normally occurs in lung cancer patients especially chromate lung cancer and smokers. Approximately 21-51% of NSCLC patients observed p16 methylation, and about 54-100% of NSCLC patients observed p16 loss of heterozygosity [153,154]. p16 gene is normally associated with lung cancer, which can be affected by tobacco smoking, and radon and plutonium exposure. It was reported that p16 methylation is highly related to past smoking and duration of smoking [155].
Ras genes are responsible for cancer-causing activities of Harvey (H-ras) and Kirsten (K-ras) sarcoma viruses. Ras mutation was observed in about 20-25% of cancer patients and up to 90% of patients with specific cancers [156]. K-ras, H-ras and N-ras code 21 kD proteins are the three main types of human Ras genes; they are members of Ras superfamily of GTPases and have a crucial role in cell proliferation and signal transduction. Approximately 60% of Ras mutations are confined to codon 12 of K-ras [157]. Approximately 78% of lung cancer patients observed K-Ras mutation, and subjects with NSCLS, pleural effusion, sputum, serum and bronchoalveolar lavage fluid also observed K-Ras mutation [158].
Lung diseases are also highly related to telomere-related genes such as TERT, TERF2 and POT1 [159,160]. The shortening in telomere length is related to cancers and the genetic alterations in telomere-related genes change the risk level of cancer due to telomere length. The risk of lung cancer is increased with the longer tertile of telomere.
The serum miRNAs have potential to detect some diseases include lung cancer. Comprehensive studies of miRNAs have been performed for diagnosis of lung cancer [161]. Several miRNAs have been investigated extensively for target lung marker detection [162]. Additionally, serum miR-206 and miR-133b were applied to detect lung carcinogenesis [163]. High hsa-mir-155 and low hsa-let-7a-2 were used to identify lung tumor markers [164]. MiR-449c with the direct target of c-Myc has been investigated to identify NSCLS, which could suppress cancer cells growth in vivo [165]. It was found that tumor growth is associated with silence and overexpression of miRNAs, and overexpression of miRNAs is suitable for early lung disease detection.

Biosensors for Lung Cancer Biomarker Detection
Enzyme-linked immunosorbent assay (ELISA) is the conventional approach for biomarker detection. It requires a labeling process, which hinders monitoring of the probe/target interaction rapidly. To overcome this challenge, investigators have studied many high-sensitive and label-free detection techniques such as Field effect transistors (FETs), surface plasmon resonance (SPR) and quartz crystal microbalance (QCM). Among these techniques, FETs are more attractive because they are compact, inexpensive and able to integrate many sensors on the same chip. Over the years, researchers worldwide have paid attention to developing label-free and high-sensitive biosensors for early diagnosis of diseases such as lung cancer. Table 4 lists some recently developed biosensors for lung tumor marker detection [166][167][168][169][170][171][172][173][174][175][176][177][178][179][180].
SPR-based biosensors have been developed for biomolecular interaction [173][174][175][176][177][178][179][180][181][182][183][184][185]. SPR-based sensors excite surface plasmon from the interface and measure the refractive index changes, which can be classified as label-free and real-time affinity reaction detection systems. A collimated polychromatic light beam from a halogen lamp passes through an optical prism and contacts a thin gold layer at a defined angle of incidence. Upon the incidence of the thin gold film, each light beam excites a surface plasmon at a certain wavelength. The reflected light is collected by an instrument with measured channels. Various SPR and FET based biosensors have been developed for assaying CYFRA21-1 protein [182,183].
Wang et al. [186] developed a high-precision optical system based on magnetic ELIA to detect CYFRA21-1. The system has potential to become a powerful tool for the rapid detection of lung cancer marker with advantages of compactness and high-sensitivity. Recently, Ribaut et al. [187] developed an innovative plasmonic optical fiber immunosensor to detect lung cancer marker cytokeratin 17. The proposed optical fiber immunosensor was tested on human lung biopsy, and the accurate detection of biomarkers in soft matters including tissues could be performed with plasmonic optical fiber grating immunosensors. Their research outcomes offered significant contributions toward diagnosis of biomarkers in tissues in clinical environments.

Piezoelectric Biosensors
Piezoelectric biosensors are helpful for bioanalytical applications with several advantages such as easy-to-make, cost-effective and high-sensitivity. Piezoelectric quartz crystal (PQC), a thin slice of quartz derived from a single crystal with optimal chemical, electrical and mechanical properties, is suitable for analytical applications. A vapor deposition of gold or silver was fabricated on a PQC sensor to serve as electrodes.
QCM-based sensors have been applied for point mutation detection of lung cancer [188][189][190][191][192][193]. They measure frequency changes in quartz crystal resonators based on adsorbate recognition, and mass changes caused by selective binding can be detected by the corresponding changes in crystals. Quartz is the most popular crystal for analytical applications due to the mechanical, electrical and chemical properties of quartz, which demonstrate piezoelectricity effectively. QCM has the ability to diagnosis small objects, such as molecular weight ligands, cells, viruses, proteins and nucleic acids [191]. Piezoelectric immunosensors with specific antibodies have been applied to detect cancer makers. Further, minimal mass threshold for unit change of frequency in conventional QCM has been improved using nanoparticles, which led to ultrasensitive detection of analytes up to few attomoles.

Electrochemical Biosensors
In recent years, various electrochemical biosensors have been developed for lung cancer marker detection [194][195][196][197]. Electrochemical-based sensors normally contain semiconductors and screen-printed electrodes, which can detect various molecules including proteins, antibody, DNA, antigen and heavy metal ions. Electrochemical biosensors are highly sensitive for diagnosis of lung cancer markers. Recent advances in electrochemical nano-biosensors offer a promising tool for diagnosis of molecules with some advantages, including low-cost, more accurate, fast-response and high-sensitivity.
Altintas et al. [195] developed a magnetic particle-modified capacitive sensor to detect some cancer markers such as CEA, CA15-3 and hEGFR. Experimental validations have been conducted to evaluate the proposed sensor. CEA and hEGFR were detected in the concentration range of 5 pg/mL to 1 ng/mL while CA15-3 was detected in the range of 1-200 U/mL with high specificity. The experimental results showed that the proposed sensor may become a useful tool for early cancer detection.
Recently, Tabrizi et al. [196] developed a highly sensitive electrochemical aptasensor based on carbon-gold nano-composite modified screen-printed electrode. VEGF165 was observed in lung cancer patients by using the proposed sensor, which might have potential to become a powerful tool for lung cancer detection. More recently, a new aptamer-based electrochemical sensor was developed by Zamay et al. [197] to identify lung cancer. The sensorcould identify cancer-related targets in crude blood plasma of lung cancer patients.

Current Trends and Future Perspectives
The currently available lung screening approaches are effective but have some drawbacks, as detailed above. MIT technique has potential to become an additional or alternative method to CT for lung disease detection. MIT-based approaches have some challenges, including heavy computational imaging algorithm, unrealistic thorax phantoms, difficult hardware systems for clinical use, and spatial resolution. To overcome these challenges, it is necessary to develop a high dynamic hardware implementation system to detect the small difference of scattered electromagnetic field.
Many investigators have employed more MIT coils to increase image resolution. This method also increased the mutual coupling signals between coils, which may reduce the accuracy of detection. Moreover, the cost and complexities of the implementation systems are also increased using the method of increasing coils. To solve this problem, a single coil can be applied to replace the multi-coil array. Recent studies have suggested that optimization of coil array configurations offer some potential benefits such as high image resolution, low-cost, and reduced operating time. Multiple in multiple out technique may also helpful in reducing complexity of the hardware system. However, further MIT investigations should be taken to improve the image algorithm and the implementation system with particular focus on the development of low-cost and compact RF coils and coil arrays to improve the image quality.
Biosensor-based techniques with particular focus on diagnosis of tumor markers have gained worldwide attention in the past few years. Up to date, biosensor and biomarker based techniques are still immature and the obtained evidence is too restricted for early diagnosis of lung cancer. Proteomic biomarkers have been applied within a panel of protein biomarkers, but they were not recommended for diagnosis with individual biomarker. The individual marker was not helpful for clinicians to obtain sufficient information about cancer tissues such as cancer stage, treatment and patient information. The major drawback of biosensor-based techniques is related to integration of lung cancer detection in primary healthcare. Among all these biosensors, QCM biosensors are more suitable and reliable for clinical surgery. The limitations of biosensors also include small target size, marker levels, and the possibility of high non-specific binding in the case of serum or real patient samples. Nano-biosensor techniques for molecules detection offer great potential for early lung cancer detection, however, these techniques are immature for clinical trials. More investigations should be provided to improve the sensitivity, accuracy, and multiplexing capacity of biosensors in the future.

Conclusions
This manuscript reviewed the recently developed imaging and biosensor-based techniques for early diagnosis of lung cancer. Recent developments in screening techniques such as CT, PET, MRI and breath analysis, as well as MIT-based approaches, were addressed with particular focus on for early lung cancer detection. Clinical trials of MIT-based approaches for lung disease detection caused worldwide excitement, which confirmed that MIT has potential to become a low risk alternative or additional clinical tool to CT for lung disease detection. Some recently developed biosensor-based techniques for diagnosis of target lung tumor markers were also reviewed. However, MIT and biosensor-based approaches are still immature for clinical applications across large populations.