A RAMAN CELL BASED ON HOLLOW CORE PHOTONIC CRYSTAL FIBRE FOR HUMAN BREATH ANALYSIS

................................................................................................................................... ii Table of


Chapter 1: Introduction
In this introductory chapter, the current status of lung cancer and its screening methods, the history and current status of breath analysis and the potential of Raman spectroscopy serving as a breath analyzer will be reviewed.

Anatomy of the lung
The lungs are essential organs in our respiratory system and gases are exchanged through breathing to keep one alive. It is located at the thoracic cavity and is composed of three lobes on the right and two on the left, which are all connected by the trachea and bronchi. When the air is inhaled, it passes through the nose or mouth to the windpipe and then divides into two bronchi to the left or right lung. Each of the bronchi branches into bronchioles and terminates at numerous alveoli. An alveolus is an air cavity supported by thin and elastic membranes to separate blood capillaries and inhaled air. This boundary allows gases to diffuse across the membrane between the blood capillaries and inhaled air. It has been estimated that there are over 300 million of alveoli in the lung. The alveolar surface area of an adult human can vary from 97 to 194 m 2 with a mean of 143 m 2 , which is about half the size of a standard tennis court (1, 2).  (12). This shows that effective screening tests for lung cancer in early stages, which is more likely to be treatable, would greatly improve the survival rates of lung cancer patients.

Current lung cancer screening methods
Screening refers to simple medical tests to help find diseased individuals from the asymptomatic public. Follow-up diagnostic tests are typically needed for confirmation (13).
A successful screening tool should be able to reduce the mortality rate of the disease with the least degree of pain and invasiveness together with reasonable clinical sensitivity, specificity and cost (14,15). Chest X-ray is one of the most common lung cancer initial imaging tests and it is often useful for discovering small resectable tumours, but chest x-rays do not reduce the number of new advanced stage lung cancer cases nor the mortality (15). CT gives more anatomic information that may be useful to provide guidance for the needle biopsy.  (18,19). Currently, there is no lung cancer screening tool generally accepted by health professions as they are usually costly, inefficient and considerably invasive, which causes long waiting time to patients and places a burden to the health care system (19). Some are considered as ineffective because of the high false-positive rates and cause unnecessary invasive confirmation tests to verify the results. A low cost, efficient and non-invasive lung cancer screening tool would be a great advancement in lung cancer control. Breath tests have been recognized as a promising candidate for lung cancer screening (6).

Physiology of human breath
Breathing is a major process in the body to perform gas exchange. For example, homeostatic processes maintain a constant condition of oxy-and deoxyhemoglobin concentrations. The body needs oxygen and sugar to provide energy to stay alive. In cellular respiration, oxygen is consumed and carbon dioxide is produced as waste. Red blood cells transport and deliver these dissolved gases to and from cells in our body via blood flow. Gas exchange takes place in the lung, more precisely at the alveolar membranes. A human exhaled breath (tidal volume) can be generally separated into two parts. The first part is called the dead space air, roughly 150 ml for a healthy subject, where inhaled air is trapped in upper airway and there is no gas exchange between blood and breath air. Alveolar breath is roughly 350 ml of gases which comes from the lung where gas exchange has taken place (20).
There are three basic processes of respiration: Ventilation is the first part of respiration which involves inhalation and exhalation. Ventilation is coordinated by contraction and relaxation of the diaphragm and intercostal muscles located at the rib cage. Air is inhaled from the nose and mouth followed by the airways and reaches alveoli by the pressure difference created by the change of lung volume. External respiration is the second part of the respiration process, in which gases are exchanged at the alveolar membranes. Gases are exchanged through the walls of numerous small air sacs and to the blood capillaries. Molecules that dissolved in blood will then be brought to other parts of the body by the circulatory system. The released carbon dioxide and left over molecules in air are then exhaled back to the atmosphere by the same route that they were inhaled. The third part is called internal respiration which describes the gas exchange between the body cells and blood (1).

Background of breath analysis
In the ancient days Hippocrates instructed students to smell patients' breaths. For example, a fruity smelling breath can mean high risk of diabetes and a fishy smelling breath may point the finger to liver impairment (21). The inhaled dry air composition by volume is approximately 78% N 2 , 21% O 2 and 0.03% CO 2 , and other trace components. Human exhaled breath usually contains approximately 78% N 2 , 16% O 2 and 4% CO 2 and other substances in trace amounts that vary from person to person (22,23).
The first detailed quantitative study of human breath analysis was in the 1970s using gas chromatography (GC) by Pauling et al (24). After Pauling's finding, Phillips et al stated that there are over three thousand different substances in human exhaled breath (25). These substances, which constitute a distinctive "breathprint," are in trace amounts and occur because of various biochemical processes in human body. By correctly analyzing the exhaled "breathprint", the physiological status of the body can be revealed and provide results quicker than the conventional body fluid tests (20,26 (24,(27)(28)(29)(30)(31). Volatile compounds are known to appear in blood and to be able to cross the alveolar membrane in the lung and thus be present in expired breath. This makes breath analysis a promising tool to serve in disease screening/diagnosis (29,32), drug assay/monitoring (33), cancer screening/detection (34)(35)(36)(37), monitoring patients at surgery (33), etc.
Volatile organic compounds (VOCs) are a subset of volatile compounds which are organic, such as benzene and isoprene. VOCs can be found in exhaled breath in the range of parts per trillion to parts per million by volume (pptv to ppmv), from the environmental cycle and production in the human body. Measuring endogenous VOCs can be useful to monitor the biochemical processes of the body and it is believed that the products of the biochemical processes of patients with and without lung cancer can be different and therefore distinguishable by breath analysis (38). This field has been interesting to researchers since there has been an overwhelming need to offer efficient, non-invasive and sensitive screening tools for patients in early-stage lung cancer to reduce mortality.
By examining the breath collected from lung cancer patients and control groups, experimental studies have suggested biomarkers for lung cancer screening, as listed in Figure   1 (27,(35)(36)(37)(39)(40)(41)(42)(43)(44)(45). Most of the early findings of exhaled components were verified by mass spectrometry-based tools, because of their strong analytical power in molecular identification. However, researchers generally do not obtain consistent results. There are over 50 different lung cancer-related VOCs suggested in ten publications from 1999 to 2010, with between 2 and 22 VOCs per study. The exhaled biomarkers of lung cancer have long been studied, however there is a lack of solid evidence to correlate these chemicals to lung cancer.
The inconsistency of lung cancer biomarkers could be due to collection method, location and diet, etc, which are discussed in section 4.3.

Current human breath analysis techniques
Various forms of breath analyzing techniques have been designed for different purposes: An alcohol breath analyzer is commonly used worldwide to deter operation of a motor vehicle by an impaired driver. It is purposely designed for blood alcohol level testing only and has no way to work for other types of exhaled substances. Electronic nose (46), is an interesting and cost-effective development. The semiconductors arrays can generate series of electronic signatures from the analyte but electronic nose has difficulty to give accurate quantifiable information to correlate VOCs. Interestingly enough, researchers have demonstrated breath analysis for lung cancer patients by trained dogs (47), but this approach may not be suitable for clinical settings. However, these non-spectroscopic methods for breath analysis do not give signals to specify analytes and therefore cannot be used for molecular identification purposes.
Spectroscopic breath analysis tools offer advantages for the identification and quantification of exhaled substances. There are two major types of spectroscopic methods for breath analysis, mass spectrometry and laser spectroscopy. Mass spectrometry has been widely accepted as a powerful analytical technique that measures the mass-to-charge (m/z) ratio of charged molecules, and it has a historical term called mass spectroscopy. Laser spectroscopy is a general term to describe analytical measurements based on the optical-medium interactions such as photon absorption and photon scattering using a laser source.

Mass spectrometry-related techniques for breath analysis
Mass spectrometry (MS) has long been playing an important role in various fields, such as food, medicine, forensic, sports, for the identification and quantification of unknown

Laser absorption spectrometry-related techniques for breath analysis
Laser absorption spectroscopy (LAS) techniques, which measure the photon absorption properties of an analyte, have been known for decades but have only become available for trace amount detection for breath analysis recently. Due to the fast development of semiconductor lasers and photodetectors, it has undergone recent advancements that enable high sensitivity and selectivity for real-time breath analysis for clinical uses (52)(53)(54). There are major laser spectroscopic techniques with sufficient sensitivity for breath analysis that have undergone increasing investigations over the past decade, including cavity ring down spectroscopy (CRDS) (55,56), tunable diode laser absorption spectroscopy (TDLAS) (57,58) and photoacoustic spectroscopy (PAS) (59).
Cavity ring down spectroscopy (CDRS) is a very sensitive optical device and can be used to monitor trace gas levels (60). It relies on long interaction pathways provided in the optical cavity by at least two highly reflective mirrors in specific wavelength. Photons bounce back and forth within the ultra-low-loss cavity and thus the effective optical interaction length can be increased more than 10,000-fold (54). Ring-down decay time is the measurement of photon leakage of the cavity in a particular frequency band to measure the sample concentration of a particular analyte. The ring-down time decays exponentially and is independent of laser intensity output instability. Absorption process can be shown by gradual decay of the laser intensity and hence the concentration of the sample gas.
Tunable diode laser absorption spectroscopy (TDLAS) is a combination of the traditional laser absorption techniques with tunable diode lasers, which measure the absorption characteristics of the sample molecules by scanning across its particular central absorption peak. TDLAS can give not only sample concentration information but also the temperature and pressure of the gas analyte. TDLAS can be integrated with a Herriott cell (multipass) consisting of concave mirrors. The laser beam comes out of the cavity at a different angle than the incoming beam, unlike the high finesse optical cavity of CDRS. Its effective beam path is considerably shorter than CDRS, but the Herriott cell can still lengthen the effective absorption path by 10-to 100-fold and offers a limit of detection in the order of 10 1 pptv (54).
Photoacoustic Spectroscopy (PAS) (59) is a spectroscopic technology based on indirect measurement of photon-molecule resonance (which converts light energy into sound energy).
The setup usually involves microphones to detect sound waves produced during photon absorption by molecules. Therefore, the laser beam is not directly measured and the background is relatively weaker compared to conventional laser-based spectroscopy.
Each laser-based technique has its own advantages and limitations, and they complement each other. CRDS is precise and reliable, but it is untunable and limited in a narrow frequency range. Moreover, it requires an expensive narrow bandwidth laser and highly reflective mirrors for particular molecules; TDLAS is tunable and offers the possibility to access different frequency ranges in the mid-IR region, but the cost of a tunable diode laser is still the main consideration; PAS employs indirect acoustic measurements to avoid major laser background noise, but the noise issue is still challenging for such weak photoacoustic conversion. Failure or difficulty with multiplexing is one of the biggest challenges for LAS- To satisfy the condition of total internal reflection, the refractive indexes n 1 has to be greater than n 2 . In this case, light propagates in the silica core with low loss of refracted light escaping the fibre. Light rays with incident angles greater than θ max would probably not propagate through the fibre and are most likely absorbed by the fibre jacket. The acceptance cone is a geometrical description of the acceptable light rays which are coupled into the core of the fibre with the largest incident angle, which has a calculated value called numerical aperture (NA). In conventional index-guiding fibres, light is always confined in the solid core but has low chance to interact with gas molecules around the fibre, and thus is not effective for gas sensing.  (67). So the wall of the core acts as a highly reflective surface to photons with these particular wavelengths and guides light through the fibre obeying the law of reflection. The hollow core of HC-PCF is gas-fillable with gas molecules so the fibre can be treated as a tiny hollow tube.

Raman scattering in HC-PCF
Spontaneous Raman scattering is a multi-directional process and the conversion efficiency is low in general. Focusing a laser beam to a focused spot would increase the power density of the laser and thus enhance the number of photon-molecule interactions in a small volume. Figure 4a shows a free space setup with a converging lens to focus the laser beam to a small focused spot. This free space setting has a small interaction volume and is inefficient for Raman-scattered photon collection as the signal is unguided. Therefore, most of the signals are lost and cannot reach the detector properly. Furthermore, filters are hard to set up in this configuration to get rid of the detector saturation problem caused by the laser photons. Figure   4b shows a laser beam focused to a hollow capillary fibre with an internal reflective coating.
This type of hollow capillary fibre can be filled with liquid or gas sample and allow Raman signal to be generated and guided in the hollow core. However, this type of capillary normally has strong leakage of the coupled light and the NA of the capillary is usually low.
Therefore, the generation and collection efficiencies of the Raman-scattered photon are still low. A research group in Japan has tried using this type of capillary fibre for gas analysis and

Hypothesis and objective
Lung cancer is the leading cause of cancer death in Canada according to the Canadian Cancer However, these techniques may not be easily applicable to clinical settings as they are usually expensive and large in size.
Raman spectroscopy is a promising candidate for breath analysis because of its ability of molecular identification. Raman scattering is one kind of photon-molecule interaction, photons interact with molecules causing excitation and de-excitation, which generate fingerprint-type spectra due to the unique molecular kinetic modes of the analyte. However, the efficiency of Raman scattering is low and the unguided signals in free space are weak.
Photonic Crystal Fibre (PCF) is new technology in fibre optics that allows light to travel through a hollow core with low transmission loss by band-gap guidance. The hollow core of PCF can be filled with gas phase samples and thus confine photons and Raman active media in a small space to enhance Raman signal strength. Spectroscopic information can be obtained from the breath-filled PCF to perform molecular identification. Therefore, the Raman spectral analysis of the exhaled breath samples from lung cancer patients and healthy controls may give clues to develop a lung cancer screening tool.
Our goal is to develop a simple, cost-effective and non-invasive tool based on Raman spectroscopy for breath analysis and potentially lung cancer screening. In this project, the plan is to develop a prototype of a Raman-gas cell based on HC-PCF for spectral analysis.
Second, the performance of this Raman-gas analyzer based on the spectroscopic signals obtained from reference gases and human breath sample is investigated. Third, the detection limit of this Raman-gas analyzer is compared with a recent publication based on hollow capillary fibre (70). The plausibility of detecting lung cancer-related exhaled biomarkers based on the detection limit of the prototype is investigated.

Chapter 2: System development
In this chapter, the design and construction of the Raman cell based breath analyzer prototype are described. The Raman-gas cell system contains mainly two parts: photonics - the optical system provides laser illumination and the collection of the Raman signals for spectroscopic analysis; and the gas filling system -the tubing and fitting system provides pressure difference in a closed space for gas loading and unloading to the HC-PCF.

HC-PCF
The fibre used in this project was a photonic crystal fibre with a hollow core (HC-800-01).
This fibre is made of silica and the holey region is made with unit cells (hollow capillaries).
The hollow core was made by removing the 7 unit cells in the middle of the photonic crystal (PC) cladding. The core is hollow through the entire fibre so that it can be filled with gas by suction. Raman signals generated in the hollow core of the HC-PCF can be guided through the optical components and collected by a Raman spectrometer for analysis. *The mode field diameter of the core is measured by the average of x-and y-near field intensity of the fibre output profile.  Poor cleaving quality can lead to unsatisfactory laser beam coupling efficiency. The cleaved PCF tip should not be cleaned by any cleaning agent because the holey region may easily be filled with any liquid by the capillary effect. It is also important not to touch the fibre tip throughout the experiment because dirt can attach on the fibre tip and absorb laser energy, leading to poor laser coupling or even damage to the fibre tip. Therefore, the fibre should be inserted into the vacuum fibre optics feedthrough before cleaving to avoid leaving any dirt on the tip during the process.

Gas chamber and tubing system
The HC-PCF can provide laser guidance and at the same time be gas-fillable. This hollow fibre can be treated as a hollow tube-cell with 10 µm diameter. Pressurized or atmospheric pressure gas samples can be pumped through the hollow core of the HC-PCF integrated with the gas chamber and tubing system. In order to provide a sealed environment for the gas cell, gas chambers are needed on both ends of the fibre without blocking the transmission of light.
Apertured and sealed gas chambers are therefore designed to serve this need. It is mainly composed of two parts, the fibre optics feedthrough and customized gas fitting. The fibre optics feedthrough is to hold the fibre into position and it is sealed by tightening a plastic sealant. The fitting is 90 degree street elbow 1/8" with NPT thread made of brass. A 2-step hole was drilled with two different drill heads (6.4 mm and 3 mm) to support the sapphire window and provide an aperture for light transmission. Figure 7 shows the schematic diagram of the gas chamber design. The sealed feedthrough is connected to one end of the apertured fitting by 1/8" NPT thread. The sapphire window (SAW12, Newport), 6.35 mm in diameter and 0.5 mm thick, is sealed and glued onto the brass fitting by silicone sealant.
Sapphire windows are ideal for applications where high pressure, vacuum, or corrosive atmospheres are considerations. The threshold pressure of this sapphire window is ~ 1600 psi but the highest pressure tested on this configuration was ~ 250 psi, which is the limit of the reference gas regulators. The other end of the brass fitting is attached to a push-to-connect tube fitting with 1/8" NPT thread. Both ends of the fibre have the same gas chamber setting and are connected by the plastic tube through the push-to-connect fittings. The tubes are connected to a vacuum pump and the gas source for gas unloading and loading by pressure difference. Not shown in Figure 5, pressure gauges are used to monitor the filling process and internal pressure; ball valves are employed to control the gas flow and discharge process. All fittings, valves and pressure gauges are connected by the 1/8" outer-diameter plastic tubes. The incoming laser beam can be focused by an aspheric lens and the laser cone hits the sapphire window attached on the custom brass gas fitting then reaches the HC-PCF. To couple the laser into the core of the fibre, the whole gas chamber was bolted down and aligned by a xyz-translation stage. Fibre coupling can be achieved by maximizing the coupled laser power monitored by a laser powermeter on the emission end. To achieve the best laser coupling, the fibre portion sticking out of the feedthrough should be kept in an optimal position such that it is as close as possible to the sapphire window but not in contact.
Otherwise, the optical guidance and gas filling process can both be affected.

Chapter 3: Results and analysis
In this chapter, several reference gas samples (oxygen gas, hydrogen gas, carbon dioxide gas) obtained from Praxair were tested to verify the calibration and response of the system; human breath was tested to identify some major exhaled components; ambient air was tested to find out the limit of detection of this prototype.

The measurement of oxygen gas
A tank of pure oxygen gas was connected to a CGA-540 gas regulator with the maximum safety working pressure 250 psi. The oxygen gas Raman spectrum in this spectral region contains a vibrational line at 1556 cm -1 with 1.2 differential Raman cross-sections (71). The fine structures on both sides of the central vibrational peak are the rotational lines (72). In our experiment, a sharp peak was observed at 1557 cm -1 , along with peaks with lower intensities and symmetrical pattern extending 150 cm -1 from both sides of the sharp peak, shown in Figure 8. Zooming into the fine structure, the rotational lines were almost evenly spaced ~ 12 cm -1 apart, the intensities of the lines were maximum at ~ 50 cm -1 away on both sides from the central vibrational peak. This measurement was taken in 0.3s exposure time with 200 psi gas pressure.

The measurement of hydrogen gas
A tank of pure hydrogen gas was connected to the tubing system by a CGA-350 gas regulator with the maximum safety working pressure 125 psi. The published Raman spectrum of the hydrogen gas has four rotational lines within the spectral region of our spectrometer (73).
The four lines observed in this experiment were at 584 cm -1 , 814 cm -1 , 1034 cm -1 and 352 cm -1 in decreasing order of the peak intensities, shown in Figure 9. The vibrational line of hydrogen gas at 4161 cm -1 is out of the transmission band of the fibre and the spectral range of the spectrometer, therefore it is not observed. The other broad peaks and bumps in this spectrum are due to the Raman characteristic of the fused silica originated from the fibre materials. The measurement was taken in 1s exposure time with 100 psi gas pressure.

The measurement of carbon dioxide gas
This tank contains 5% carbon dioxide gas in nitrogen gas. The outlet of the CGA-500 regulator was connected to the tubing system to supply this gas sample for analysis. The pressure of the gas was set to near atmospheric pressure such that it was close to the condition of human exhaled breath. The Raman spectral feature of carbon dioxide gas has two vibrational lines within the fingerprint region, at 1286 cm -1 and 1388 cm -1 with Raman cross-sections 0.89 and 1.4 respectively and generating the intensity ratio 0.64 (71). The two lines observed in this experiment were at 1286 cm -1 and 1388 cm -1 in increasing order of the peak intensities, shown in Figure 10. The intensity ratio measured in this experiment was around 0.61. The nitrogen gas peak is outside the spectral region of the spectrometer and therefore not observed. This measurement was taken in a total exposure time of 10 s.

The measurements of the ambient air
The filling process of the ambient air is done by having the two ends of the fibre tip exposed to the air. Typical ambient air contains approximately 21% oxygen and 0.03% carbon dioxide gas. This measurement was taken in 5 s exposure time. The Raman features of the oxygen molecules, 1556 cm -1 line and the rotational signatures are observed, shown in Figure   12. However, in this experimental condition, the carbon dioxide gas Raman features are not clearly observed. It is attributed to the low concentration level of the carbon dioxide gas in the ambient air. The weak Raman signals are obscured by the measurable background noise.
To reveal the overtaken Raman signals originated from low concentration molecules, it is recommended to improve the spectral signal-to-noise ratio, for example, by increasing the spectral integration time and measurement accumulation. deviation. Using the signal-to-noise ratio obtained from the carbon dioxide peaks, assuming the minimum detectable signal to be the statistical variance of the spectrum (whose signal-tonoise ratio is equal to 1) and the equivalent measured concentration of the sample, the lower detection limit in this optical configuration is roughly 15 parts per million by volume (ppmv) of the carbon dioxide molecules under ambient conditions.

Chapter 4: Discussion and conclusion
In this chapter, the challenge and potential of this project are discussed and summarized to draw conclusions for the future studies. The performance of this Raman-gas cell based on a photonic crystal fibre is compared to the recent published Raman-gas cell based on a capillary hollow fibre. The future direction and improvements to ameliorate the performance of the system to detect analyte in ultra-trace amounts are discussed. The predicted challenges of breath analysis for clinical application are also discussed.

Comparison of a HC-PCF based Raman breath analyzer to that of based on a hollow capillary fibre
The HC-PCF based prototype was designed and developed at BC Cancer Research Centre in Vancouver, BC. The specifications and performance of this prototype were compared to another breath analyzer with a reflective material coated capillary fibre-based breath analyzer reported by Y.Okita et al (70), shown in Table 2. In this published experiment, a polycarbonate based hollow capillary optical fibre with a silver inner coating was employed to be the light guide and gas cell of the system. The excitation laser beam at 532 nm was focused to the inner silver coated capillary fibre (700 µm diameter) by a 70 mm focal length lens, while the other end of the fibre was opened. Back-scattered light (i.e., photons scattered in backward direction and traveling in the opposite direction to the incoming excitation laser) goes back by the original route. The laser beam was then collimated by the same focusing lens for filtering and re-focused to a Raman spectrometer for spectral analysis.
Attenuation: The internal silver coating helps to reduce attenuation loss of both the laser excitation and Raman signal. However, it is recommended to keep the capillary fibre straight in order to minimize the attenuation loss. This publication shows that the total power loss of this capillary fibre set-up (1.1 m long) can be up to more than 80% for the bending radius ~ 14 cm. This means that the strong attenuation of the inner silver coated capillary fibre due to bending would constrain the smallest possible size of the breath analyzer. In contrast, the HC-PCF typically has better optical guidance. The HC-PCF in this project (HC-800-01) has less than 10% loss per meter in attenuation within the transmission band of the fibre, and it is almost insensitive to bending. The experiment was run with a 2.5 m long HC-PCF coiled up onto a fibre wheel with a radius of a few centimeters.
Scattering modes: The back-scattered mode for Raman collection would theoretically give higher signal intensity, especially for fibres with strong attenuation, for example, capillary fibres. In contrast, both the laser and Stokes Raman photons had low attenuation in HC-800-01. Therefore, there should be no significant difference in terms of the Stokes Raman signal intensity in both forward-or back-scattered mode for fibre length less than ~ 5 m. The additional higher-order surface modes (considered as noise) decay exponentially in the cladding structure (74,75) and therefore can be minimised in the forward-scattered mode. It was experimentally proven that forward-scattered mode can offer better signal-to-noise ratio over the back-scattered mode in this optical configuration.
Numerical aperture (NA): Hollow capillary fibres usually have low NA compared to the typical value of 0.22. This is because coupled light with high NA (small angle to the normal line) would introduce extra absorption and scattering by the wall of the capillary and thus reduce reflection efficiency, which is governed by Fresnel's equation as a function of the incident angle. In this case, the focal length of the focusing lens has to be long enough to produce a low NA (since the NA is inversely proportional to the focal length of the lens).
HC-800-01 has a NA ~ 0.2, which does not require a lens with a particularly long focal length. The focusing lens used in this project was 18.4 mm, which is optimized for the coupling efficiency and the space taken by the gas chamber, in which it is ready for a compact design.
Detection limit: The detection limit of the capillary fibre-based system was roughly 0.2% (2000 ppmv equivalently) by extrapolating the Raman intensity as a function of the oxygen concentration to the measurable noise level, as the lowest measured concentration is ~ 0.25%. The detection limit of this project for the HC-PCF-based breath analyzer was found by measuring the carbon dioxide level in air (~ 0.03%) and the statistical variance of 150 spectra. The detection limit of this single pass configuration was ~ 15 ppmv, which is over 100-fold improvement over the capillary system.

Future work to improve the device sensitivity
Most of the trace VOCs in exhaled breath are within the upper parts per billion by volume (ppbv) to the lower parts per trillion by volume (pptv) range. Isoprene, one of the potential lung cancer-related biomarkers in human exhaled breath, has concentrations on the order of magnitude of 10 2 ppbv. The Raman spectrum of isoprene has a very strong peak at 1638 cm -1 with cross-sections around 50 (76). To observe this peak from an exhaled breath sample, the Raman system has to have at least a single digit ppmv detection limit. In this case, it needs roughly 10-fold improvement over the sensitivity of the HC-PCF-based breath analyzer presented in this project. Possible improvements for the future work are discussed in this section.
Sample pre-concentration: Sample pre-concentration can artificially increase the concentration of particular types of analyte in a sample before the analytical procedure.
Therefore, the signal-to-noise ratio and sensitivity of the device can be enhanced. This allows trace amounts of analyte with concentrations below the detection limit to possibly be detected. Solid-phase microextraction (SPME) is one of the pre-concentration tools used most frequently in VOC analysis (77)(78)(79). SPME contains a fibre that is adhesive to certain types of VOCs and the fibre can in principle adsorb targeted molecules based on the concentration proportion once the equilibrium is reached. Heating is then used to desorb the VOCs from the fibre. The fact that it is convenient and inexpensive allows researchers to selectively enhance the concentration of particular types of VOC. Sensitivity for certain VOCs in human breath can be further improved by integrating the low-temperature glassy carbon with SPME (80). To make SPME useful in this project, a customized heating desorption chamber may be required to integrate it with the Raman-gas cell.
Interaction length: The Raman peak intensity depends on the number of successful Ramanscattered photons in the core of the HC-PCF. Excitation photons and sample molecules are confined in a tight space in the core of the fibre. One way to enhance the Raman intensity is to increase the length of the HC-PCF. This would increase the number of photon-molecule interactions in the fibre core and thus improve the signal strength with an optimal fibre length. Moreover, a multipass configuration (81) can also increase the effective interaction length by re-focusing the output excitation beam back into the Raman-gas cell such that the excitation beam can pass the Raman-gas cell back and forth multiple times to increase the number of Raman interactions. In this case, the photon-molecule interaction length can thus be enhanced to produce a better signal-to-noise ratio in a compact system.

Exposure time and statistical noise:
The spectral signal-to-noise ratio of a Raman peak is a function of the exposure time. In Raman spectroscopy, the charge-coupled device (CCD) camera of the spectrometer introduces noise to each readout. By maximizing the exposure time in a single spectrum (the longest possible integration time before the CCD is saturated by the intensity of the peak of interest), the readout noise of the CCD camera can be minimized in each spectrum. On the other hand, the counting statistics error would also be introduced to each measurement. The counting statistics error can be minimized by repeating the experiment and averaging the collected spectral data. The averaged spectrum would represent values that are closer to the true mean in each pixel column of the camera. But the statistical variance reduction would reach a plateau when the readout noise of the CCD becomes a dominant factor.
Laser power: The laser power is a measure of the number of photons in a laser beam in a given time frame. The larger the number of photons confined to the core of the fibre, the higher the chance of having more successful Raman scattering and propagate through the fibre. This HC-PCF can transmit high laser power with a good beam quality at a very high power density up to 10 14 W/cm 2 (82). Therefore, employing a laser with higher power would be an alternative way to improve the sign-to-noise ratio.

Predicted challenges in breath analysis for clinical application
Researchers have been trying to make breath analysis more practical and applicable in clinical settings. However, the large variety of results in literature for the same test is one of the major problems preventing breath testing from being widely applied in the medical field.
Source of predicted challenges and potential solutions in clinical application are discussed: Mouth and nose exhalation: Researchers have demonstrated that concentration levels for some VOCs are affected by the route of exhalation, i.e., the mouth, nose or both. It was shown that some VOCs could be produced in the oral cavity instead of endogenously exchanged on the alveolar interface (83,84). Therefore, the route of exhalation may affect the measured concentration of the analytes of interest.
Dead space air: Dead space air is the part of inhaled air which does not perform gas exchange, and it is usually the portion stays in the conducing airways during the breathing cycle. Breath sampling is a general sample of the exhaled breath and it is not easy to separate the dead space volume. The volume of breath samples from each patient can be different, therefore, the dead space volume can dilute the endogenous VOCs' concentration by different degrees. End tidal breath sampling (collection at the end of exhalation) for breath analysis is an interesting technique recently proposed. Certain target molecules from the end tidal breath would be less diluted by the dead space volume during the sampling. It has potential to provide an alternative pre-concentration method without chemical manipulation to the sample to avoid contamination and absorption competition. Researchers have shown the possibility of real-time buffered end tidal measurements of breath VOCs followed by MS-based techniques (85). Moreover, concentration of endogenous substances can vary considerably not only from person to person, but also day to day and even breath to breath for the same individual. Collection of breath sample can be done in a single breath but more accurate results can be done in averaging a series of breath samples (31).
Personal/medical issue: There is a lack of well-recognized standardization of endogenous VOCs' concentration profile in human exhaled breath. The difference in concentration level of some types of molecules can vary across gender, age, diet/eating habit, drug (86-88) or even the illness or disease. The fact that some patients are less able to offer full exhalation should be taken into account. For instance, patients with late-stage lung cancer may only be able to offer a small vital capacity due to the weak exhalation force. As inhaled air sitting at more upper airways would be exhaled before the lower airways, molecules released by a deeper sitting tumour would be less exposed to the fresh air and less likely to be exhaled compared to those from the higher airways.
Measuring tools: To measure ultra-low concentrations, the testing tools must be cleaned and sterilized before use to remove potential contaminants, e.g., residues from the manufacturing process, cleaning agents, etc. To make sure the endogenous VOCs from the alveoli breath are measured and distinguished, some other common compounds that could be found in the oral cavity may also need to be eliminated during the analysis, e.g., oral hygiene products, food.
Sample manipulation (such as pre-concentration) may also introduce calculation error due to the selective/non-linear absorption.
The combination of VOCs and their corresponding concentrations could vary significantly between different sample sources and instrumental responses. The background problem can generally be minimized by subtraction and supplying air for inhalation with known concentration of VOCs. However, these processes are less convincing if the concentration change of VOCs are tiny compared to the background level. Moreover, complex dependent issues are hard to be corrected for by a simple subtraction, such as the dead space volume. As a result of the above variances, data from breath analysis usually show wide variations. More effort has to be made in this area to provide adequate standardized procedure for breath sampling, analysis and background correction, to allow reliable comparison between different published results.

Conclusion
A Raman-gas cell based on HC-PCF for breath analysis has been designed and developed in the BC Cancer Research Centre in Vancouver, BC. This breath analyzer prototype based on photonics technology is potentially simpler, more cost-effective and faster than GC-MS. This Raman-gas cell contains two parts: optical and tubing system. The optical section contains a laser, optical fibres and other optical components. HC-800-01 is the HC-PCF used in this project and it serves as a light guide and gas cell. A tubing system was also constructed for the gas supply. In this project, Raman spectra from reference gases (oxygen gas, hydrogen gas, carbon dioxide gas) have been obtained. Human exhaled breath sample tests were also run to produce Raman spectrum for analysis. The experimental eO 2 :eCO 2 concentration ratio was found to be almost 4:1. The detection limit of the system was found to be ~ 15 ppmv obtained from the averaged spectra of CO 2 in the ambient air. This is more than 100-fold improvement over the recent published results by a Raman-gas cell based on hollow capillary fibre. To be useful in testing molecules in the pptv to ppbv concentration range, the device sensitivity has to be improved in certain aspects such as: optimizing the interaction length of the fibre and possibly using a pre-concentration method to enhance signal-to-noise ratio.
Breath analysis is promising as a simple, painless and real-time testing method in the medical field. Exhaled VOCs are proven to be an indication of certain disease states, such as lung cancer. The photonics technique based breath analyzer can offer advantages in terms of compactness, affordability and multiplicity. This Raman-gas cell has great potential to be an inexpensive, non-invasive and point-of-care lung cancer screening tool to benefit our society.