Technical field

The present application relates to lock-in detection free Raman imaging method and apparatus, particularly for determining a stimulated Raman scattering (SRS) signal.

Background

A conventional stimulated Raman scattering (SRS) microscopic system is known to use a lock-in amplifier to extract a Raman scattering signal from imaging sensor data of a photodetector in accordance with a reference modulation signal. However, such a microscopic system typically has a low imaging speed due to the time constant of the lock-in amplifier as well as the pixel dwell time of the photodetector. Furthermore, the use of a high power laser beam (100 mW/pm ² ) may cause photodamage to living cells and is therefore unsuitable for high speed bioimaging.

It is desirable to provide a lock-in detection free Raman imaging method of determining a stimulated Raman scattering (SRS) signal, a field-programmable gate array (FPGA) device configured to perform the method, a lock-in detection free Raman imaging apparatus and a computer readable medium for causing a computing device to perform the method, which address at least one of the drawbacks of the prior art and/or to provide the public with a useful choice. Summary

According to one aspect, there is provided a lock-in detection free Raman imaging method of determining a stimulated Raman scattering (SRS) signal, comprising: receiving imaging sensor data representative of a line-scan SRS signal with first and second scattering states; calculating a first scattering value based on a difference between a first intensity value of the imaging sensor data corresponding to the first scattering state and a second intensity value of the imaging sensor data corresponding to the second scattering state; and determining the line-scan SRS signal based on the calculated first scattering value.

The described embodiment is particularly advantageous. By calculating the first scattering value based on the first and second intensity values, no lock-in detection operation needs to be performed. The method allows a lower excitation power to be used and is thus suitable for bio-imaging even at a high imaging speed. Preferably, the lock-in detection free Raman imaging method may further comprise receiving corresponding lines of scattered light from a target which has been excited line by line by respective laser beams formed from two laser sources, each line representing a plurality of pixels of the target; and generating the imaging sensor data based on the corresponding lines of the scattered light. With line scanning, a dwell time may be increased to achieve a better signal to noise ratio.

Preferably, the first scattering value may be calculated by subtracting the second intensity value from the first intensity value. The subtraction operation is easier and more efficient to perform in comparison with a lock-in operation.

The first intensity value may be temporally adjacent to the second intensity value. It is preferred that the method may further comprise calculating a second scattering value based on a difference between a third intensity value of the imaging sensor data corresponding to the first scattering state and a fourth intensity value of the imaging sensor data corresponding to the second scattering state, wherein the line-scan SRS signal is determined based on a result of averaging the calculated first scattering value and the calculated second scattering value. The line-scan SRS signal thus calculated has an improved signal-to-noise ratio. The first and second scattering states may correspond to first and second modulation states of a modulated Stokes beam, respectively. The imaging sensor data is preferably received with reference to a frequency of alternation of the line-scan SRS signal between the first and second scattering states. For example, a detector array for parallel detection in association with line excitation with cylindrical lens module may be provided to achieve high speed SRS imaging. That is, parallel detection can be realised using the frequency of alternation where the imaging sensor is a detector array, such as a linear detector. Data acquisition and/or processing can be synchronized and programmed via FPGA clocking for SRS imaging.

The imaging sensor data may be associated with at least one pixel line number.

According to another aspect, there is provided a field-programmable gate array (FPGA) device configured to perform the method. The method is simple and does not require complex hardware for implementation, which makes the method particularly suitable for implementation using an FPGA device.

According to yet another aspect, there is provided a lock-in detection free Raman imaging apparatus for determining a stimulated Raman scattering (SRS) signal. The apparatus comprises a receiving module arranged to receive imaging sensor data representative of a line-scan SRS signal with first and second scattering states; a processing module arranged to calculate a first scattering value based on a difference between a first intensity value of the imaging sensor data corresponding to the first scattering state and a second intensity value of the imaging sensor data corresponding to the second scattering state; and to determine the line-scan SRS signal based on the calculated first scattering value. The apparatus may include: a pump laser source and a Stokes laser source adapted to generate respective laser beams to excite a target; an optical subsystem adapted to guide the laser beams to the target; an FPGA device comprising the receiving module, and the processing module; and an imaging sensor adapted to received scattered light from the target to generate the imaging sensor data representative of the SRS signal in the scattered light.

In one specific example, the imaging sensor of the apparatus may be arranged to receive corresponding lines of the scattered light from the target, each line representing a plurality of pixels of the target; and the FPGA device may also be arranged to generate the imaging sensor data based on the corresponding lines of the scattered light,

The imaging sensor may be a linear detector array, which may have a high imaging speed. Further, the imaging sensor may be a complementary metal- oxide-semiconductor (CMOS) imaging sensor. The typically large full well capacity may eliminate the need for frame accumulation.

The apparatus may be adapted for use in bioimaging of living cancer cells. For example, the laser beam generated by the pump laser source may have a power density of substantially 0.3 mW/pm ² , the laser beam generated by the Stokes laser source may have a power density of substantially 30 mW/pm ² , and the imaging sensor may have a pixel dwell time of substantially 100 s. Such a configuration, as discussed in the described embodiment, may be useful for high-speed bioimaging with reduced or eliminated risk of photodamage.

It is envisaged that features relating to one aspect may be applicable to the other aspects. Brief Description of the drawings

Example embodiments will now be described hereinafter with reference to the accompanying drawings, wherein like parts are denoted by like reference numerals. Among the drawings:

Figure 1 shows a schematic diagram of a preferred embodiment of a stimulated Raman scattering (SRS) microscopic system of the present invention in association with a computing device;

Figure 2 is a block diagram of a field-programmable gate array (FPGA) device of the microscopic system of Figure 1 ;

Figure 3A is a waveform representation of a modulated Stokes beam generated by the microscopic system of Figure 1 ;

Figure 3B is a waveform representation of a line-scan SRS signal of a sample scanned by the microscopic system of Figure 1 ;

Figure 4 shows a flowchart of a method performed by the FPGA device of the microscopic system of Figure 1 ;

Figure 5(a) is a line chart showing dependence of a relationship between a line-scan SRS signal and a pump power density in an experiment conducted using the microscopic system of Figure 1 ;

Figure 5(b) is a line chart showing a relationship between an overall noise and the pump power density in the experiment of Figure 5(a);

Figure 5(c) is a line chart showing a relationship between a signal-to- noise ratio and the pump power density in the experiment of Figure 5(a);

Figure 5(d) is a line chart showing a relationship between the overall noise and a dwell time in the experiment of Figure 5(a);

Figure 6 is a line chart showing SRS signal intensity versus DMSO concentration;

Figure 7(a) shows a cropped portion of an image of polystyrene (PS) and poly methyl methacrylate (PMMA) beads, obtained using the microscopic system of Figure 1 ;

Figure 7(b) shows a line chart of spatial resolution along lateral and axial directions of a PS bead; Figure 8 shows six corresponding cropped portions of images showing the Brownian motion of PMMA beads, obtained using the microscopic system of Figure 1 at intervals of 200 ms; and

Figure 9 shows six images of living gastric cancer cells obtained using the microscopic system of Figure 1 at intervals of two hours.

Detailed Description

Figure 1 shows a schematic diagram of a preferred embodiment of a stimulated Raman scattering (SRS) microscopic system 100 of the present invention. The system 100 includes an excitation laser source 105, an optical parametric oscillator (OPO) 1 10, an acoustic-optic modulator (AOM) 1 15, a delay mirror group 120, a dichroic mirror (DM) 125, a telescopic lens group 130, a cylindrical lens 135, a polarising beam splitter (PBS) 140, a quarter-wavelength plate (QWP) 145, a microscopic objective (MO) 150, a filter (F) 155, a linear detector (LD) 160, and a field-programmable gate array (FPGA) device 165.

The excitation laser source 105 is operable to output an intermediate beam with a wavelength of 532 nm and a Stokes beam with a wavelength of 1064 nm. The intermediate beam is reflected by two mirrors (M) 106 to reach the OPO 1 10. The Stokes beam is reflected by two mirrors (M) 107 to reach the AOM 1 15.

The Stokes beam is modulated by the AOM 1 15 with a drive signal of 10 kHz to regularly alternate between a first modulation state and a second modulation state. In this embodiment, when the modulated Stokes beam is in the first modulation state, an intensity level of the Stokes beam is reduced; and when the modulated Stokes beam is in the second modulation state, the intensity level of the Stokes beam is not reduced. The AOM 1 15 thus serves as a Stokes laser source. The OPO 1 10 is operable to tune or adjust the wavelength of the intermediate beam to fall within the range of 670 nm to 980nm, and to output the tuned intermediate beam as a pump beam towards the dichroic mirror 125. The wavelength tuning range covers both fingerprint (800 cm ^"1 to 1800 cm ^"1 ) and high-wavenumber (2800 cm ^"1 to 3600 cm ^"1 ) spectral regions. The OPO 1 10 thus serves as a pump laser source. The mirror group 120 serves to provide an additional length of propagation path for the modulated Stokes beam to thereby temporally overlap the modulated Stokes beam with the pump beam. Following the mirror group 120, the modulated Stokes beam is reflected by a mirror 121 towards the dichroic mirror 125.

The dichroic mirror 125 serves to spatially overlap the modulated Stokes beam with the pump beam by allowing passage of the pump beam and reflecting the modulated Stokes beam. The modulated Stokes beam is reflected by the dichroic mirror 125 to traverse coaxially with the pump beam towards the telescopic lens group 130.

The telescopic lens group 130 includes two lenses (L1 and L2) and serve to expand or magnify the overlapped beams to fill a back aperture of the microscopic objective 150.

The cylindrical lens 135, with a focal length of 150 mm, cooperates with the microscopic objective 150 to form the expanded beams into a line shape of approximately 100 pm. The polarizing beam splitter 140 is configured to permit passage of the beams from the cylindrical lens 135 through the beam splitter 140 towards the quarter-wavelength plate 145. The quarter-wavelength plate 145 serves to convert the beams from p-polarisation to circular polarisation. The polarisation-converted beams are then reflected by a Galvano mirror (GM) 146 controlled by the FPGA device 165 towards a sample via the microscopic objective 150.

In response to the pump and Stokes beams, the sample (also referred to herein as "target") arranged at a sample plane (X) of the system 100 scatter or emit light including an SRS signal. In this embodiment, the scattered light including the SRS signal is detected in an epi-direction, and the SRS signal is a line-scan SRS signal. The light scattered by the sample traverses through the microscopic objective 150 to be reflected by the Galvano mirror 146. The reflected light then undergoes a polarisation conversion from circular polarisation to s-polarisation at the quarter-wavelength plate 145. Next, the polarisation-converted light is reflected by the polarisation beam splitter 140 towards the linear detector 160 via a lens element (L3) 141 and the filter 155.

The filter 155, functioning as an optical bandpass filter, allows passage of spectral components of the SRS signal of the polarisation-converted light corresponding in frequency to the pump beam, and blocks passage of other spectral components of the of the polarisation-converted light. For SRS imaging in the high-wavenumber spectral region, the filter 155 can be configured to permit passage of spectral components with wavelengths ranging from 700 nm to 860 nm whilst blocking passage of other spectral components. For SRS imaging in the fingerprint spectral region, the filter 155 can be configured to permit passage of spectral components with wavelengths ranging from 850 nm to 950 nm whilst blocking passage of other spectral components.

The linear detector 160 is an imaging sensor including a single line of 512 sensor pixels conjugated with a linear region of the sample plane for sensing the SRS signal. Each sensor pixel is 50 pm by 1000 pm in dimensions. Imaging sensor data generated by the linear detector 160 and representative of the SRS signal is provided directly to the FPGA device 165 for processing. In this embodiment, the imaging sensor is a complementary metal-oxide- semiconductor (CMOS) imaging sensor with a large full-well capacity (discussed in further detail below). Shown in Figure 2 is a block diagram of the FPGA device 165 in association with the AOM 1 15, the linear detector 160 and a scanning stage controller 175 of the Galvano mirror 146. The FPGA device 165 includes a timing logic block

166 and a subtraction logic block 167.

The timing logic block 166 is operable to provide the drive signal of 10 kHz to the AOM 1 15 via a DAC 1 16, another drive signal of 10 kHz to the scanning stage controller 175 via another DAC 176, and a drive signal of 20 kHz directly to the linear detector 160. The drive signal provided to the AOM 1 15 for modulation of the Stokes beam is in the form of a square wave.

Figure 3A shows a signal diagram for waveform representation of the modulated Stokes beam. It is worth noting that, in this example scenario, the intensity level of the Stokes beam is reduced to zero when the modulated beam is in the first modulation state. Figure 3B shows a signal diagram for waveform representation of the SRS signal represented by the imaging sensor data in relation to one of the sensor pixels of one image and in relation to the signal diagram of Figure 3A. The SRS signal is shown to alternate between first and second scattering states corresponding to the first and second modulation states of the Stokes beam, respectively. Operation of the subtraction logic block

167 with respect to the imaging sensor data received from the linear detector 160 via an ADC 161 is described below with reference to Figure 3B.

Referring to Figure 4, the FPGA device 165 is configured to perform a preferred embodiment of a lock-in detection free Raman imaging method 200 of determining an SRS signal, according to the present embodiment. For example, the FPGA device 165 may be configured or programmed using a computer readable medium with instructions for causing the FPGA device 165 (or other suitable integrated circuits) to perform the method 200. The method 200 includes steps 210 to 250. The FPGA device 165 includes a receiving module (not shown) and at step 210, the FPGA device 165 is arranged to receive, from the linear detector 160, the imaging sensor data representative of the SRS signal with the first and second scattering states. The received imaging sensor data includes pixel data generated by the 512 sensor pixels for multiple pixel lines and is therefore associated with a plurality of pixel line numbers.

In step 220, the FPGA device 165 calculates a first scattering value based on a difference between the first intensity value of the imaging sensor data corresponding to the first scattering state and the second intensity value of the imaging sensor data corresponding to the second scattering state. In this embodiment, this calculation is performed for each of the 512 sensor pixels for each pixel line based on the corresponding pixel data. In such a manner, 512 first scattering values are calculated for each pixel line of each image.

Referring again to Figure 3B, for the calculation in relation to each sensor pixel, the subtraction logic block 167 (or generally a processing module) of the FPGA device 165 is operable to perform the calculation with reference to the first adjacent pair of intensity values by subtracting the second intensity value l ₂ from the first intensity value . An alternative arrangement involving more than one adjacent pair of the intensity values is discussed below.

In step 230, the FPGA device 165 determines the SRS signal for each pixel based on the calculated first scattering value of the respective pixel, and outputs the determined SRS signal together with the corresponding pixel line number to the computing device 170 for forming a corresponding linear region of an image of the sample.

Next, a determination is made at step 240 by (or for) the FPGA device 165 as to whether another linear region remains to be scanned by the linear detector 160. If a result of the determination is affirmative, the FPGA device 165 proceeds to step 250 to control the Galvano mirror 146, via the DAC 176 and the scanning stage controller 175, to orient at the another linear region to be scanned. Following step 250, the FPGA device 165 repeats steps 210 to 230 with respect to the another linear region. If the result of determination at step 240 is negative, the process ends.

The SRS microscopic system 100 has numerous advantages as discussed below.

The SRS signal for each sensor pixel can be determined by performing the subtraction operation on the intensity values extracted from the corresponding imaging sensor data. Since the FPGA device 165, which provides the drive signal to the AOM 1 15, is configured with the modulation frequency of the Stokes beam, extraction of the intensity values from the corresponding imaging sensor data by the FPGA device 165 can be accurately timed with reference to the modulation frequency. These operations are performed in the time domain. In contrast with the frequency domain operation of the lock-in amplifier used by the prior art arrangement, the FPGA 165 is easier to implement and the method 200 performed by the FPGA 165 easier to perform. The FPGA device 165 can efficiently process the imaging sensor data received from the sensor pixels of the linear detector 160 in parallel via a communication interface of a typical bandwidth of approximately 12.5 Gb/s. This high bandwidth allows a line rate of approximately 1 .5 MHz (calculated based on 512 pixels at 16 bits accuracy). Instead of having the imaging sensor data transferred to and processed by the computing device 170, the calculation is efficiently performed by the FPGA device 165 by virtue of the method 200. The SRS signal determined based on the calculated scattering value of each pixel and the corresponding pixel line number are then provided to the computing device 170 to form an image. In contrast with the prior art arrangement where raw imaging sensor data is communicated to a computing device, communication of the determined SRS signal to the computing device 170 requires a much lower bandwidth and is thus more efficient. The CMOS-based linear detector 160 is advantageous in terms of shot noise. For the measurement of N photons, the corresponding shot noise is N and thus the signal-to-noise ratio (SNR) is VN . The SRS signal, mathematically representable by ΔΙ _Ρ /Ι _Ρ , is in the order of 10 ^"4 to 10 ^"6 for biological tissue and cells. In the example of Figure 3B, ΔΙ _Ρ indicates a subtraction of the second intensity value l ₂ from the first intensity value (ΔΙ _Ρ = -l ₂ ) and / _P is equal to the first intensity value . This means that at least 10 ⁸ photons need to be collected to extract and determine the SRS signal from the large DC laser background. A conventional charge-coupled device (CCD) has a full well capacity in the order of 10 ² ke ^" and thus necessitates the accumulation of at least 10 ³ frames for extraction of the SRS signal. The conventional CCD, albeit compatible with the system 100 and the method 200, is thus less preferred for rapid SRS imaging applications. In contrast, with a full well capacity of approximately 60 Me ^" , the CMOS-based linear detector 160 can be used to collect or capture the SRS signal in the order of 10 ^"4 without the need for frame accumulation.

Three noise components exist in the SRS signal:

v ' n oise = Iv th ² erm al + ^T v ' s ² hot + ^T V ' II ² f ^ ^ where V _n0 ise represents an overall noise affecting the SRS signal, Vthermai represents a thermal noise, V _sho t represents a shot noise, and ^represents an 1/f noise.

The thermal noise is dependent on an operation temperature of the linear detector 160, and satisfies:

_ 2k _B TR _L

v Thermal ^— J _Δί

where k _B represents the Boltzmann constant, T represents the operating temperature of the detector 160 (approximately 295 K), R _L represents an equivalent load resistance, and At represents a signal collection time. The shot noise \/ _sho t can be mathematically expressed as: where i _p represents the current induced by the pump beam (equivalent to the first intensity value in the example of Figure 3B), and i _d represents a dark current. The dark current is negligible because a high power pump beam is used.

The 1/f noise V _1/f can be expressed as:

where o _R | _N represents a relative intensity noise of laser (or the pump beam) with the unit of decibels relative to the carrier/frequency (dBc/Hz). The induced current i _p in the SRS microscopic system 100 can be calculated as follows:

= h - ^w , (5) where f _L is the line rate (20 kHz) and W _e - is the full well capacity (60 Me-) of the linear detector 160.

Figures 5(a), 5(b) and 5(c) are line charts showing dependence of the SRS signal, the overall noise and the SNR on the pump power density, respectively, for polystyrene beads at 3050 cm ^"1 in an experiment with a Stokes power density of 20 mW/pm ² Figure 5(d) is a line chart showing dependence of the overall noise on the line dwell time in the same experiment.

With reference to Figure 5(a), the SRS signal exhibits a linear dependence on the pump power density when the pump power density is below or equal to 0.3 mW/pm ² , and is saturated when the pump power density exceeds 0.3 mW/pm ² due to saturation of the linear detector 160. With reference to Figure 5(b), the overall noise exhibits a similar trend, showing a linear increase when the pump power density is below or equal to 0.3 mW/pm ² and showing saturation when the pump power density exceeds 0.3 mW/pm ² This is because the shot noise (Equation 3) and the 1/f noise (Equation 4) are determined by the induced current i _p (Equation 5), which is limited by the full well capacity of the linear detector 160.

By fitting the linear region of noise versus pump power density using Equations 1 to 5, the relative influence of each noise type (i.e, the shot noise, the 1/f noise, and the thermal noise) on the SRS signal can be calculated and presented as a ratio. For instance, a pump power density of 0.3 mW/pm ² results in a ratio of the noise types of 4.0:2.1 :1 .5 { V _sho t ^' - Vi/f. Vthermai), the shot noise being dominant.

With reference to Figure 5(c), the SNR increases with the pump power density and reaches saturation at approximately 145 where the pump power density is 0.3 mW/pm ² .

With reference to Figure 5(d), with a pump power density of 0.3 mW/ m ² , the overall noise is shown to be inversely proportional to the square root of the line dwell time.

As the shot noise is the dominant noise source in SRS imaging, the SNR of SRS signal can be expressed as:

where l _pump and Istokes are the power densities of the pump and Stokes beams, respectively. At is the pixel dwell time, which is the line dwell time in line-scan SRS imaging, k is a constant.

A conventional point-scan microscopic system can achieve an SNR _S RS of 0.1 1 k with a pump beam power density of 50 mW/pm ² , a Stokes beam power density of 50 mW/pm ² , and a pixel dwell time of 100 ns. In contrast, the line-scan SRS microscopic system 100 can achieve the same SNR _S RS of approximately 0.1 1 k with a lower pump beam power density of 0.3 mW/pm ² , a lower Stokes beam power density of 20 mW/pm ² , and a longer pixel dwell time of 100 ps, where the sample is 100 χ 0.5 pm ² in dimensions. That is, in comparison with the conventional microscopic system, the SRS microscopic system 100 can achieve a similar SNR with an excitation power density that is fivefold smaller.

The detection limit of the system 100 is measured to be approximately 2.7 M for dimethyl sulfoxide (DMSO) using a pump beam power density l _pump of 0.3 mW/pm ² and a Stokes beam power density Istokes of 30 mW/pm ² with a line dwell time of 100 με (see Figure 6 for a line chart of intensity (of the SRS signal) versus concentration of DMSO). The detection limit of the known conventional system is measured to be 141 mM for DMSO using a pixel dwell time of 60 με (i.e., imaging speed of approximately 15.7 sec/frame). If the system 100 is configured with a similar imaging speed, a resultant SRS image may be averaged up to 314 times, which improves the SNR by approximately 18-fold, achieving a detection sensitivity of 150 mM for DMSO.

In an experiment, the system 100 is used to obtain SRS images of a mixture of polystyrene (PS) beads of 0.5 pm and poly methyl methacrylate (PMMA) beads of 1 .0 pm in water. Figure 7(a) shows a cropped portion of one of the obtained images of 512 pixels by 512 pixels, with a field of view (FOV) of 9.4 m *9.4 pm. The cropped portion is 48 pixels by 48 pixels in dimensions. The cropped portion shows eight smaller dots representing PS beads (3050 cm ^"1 of aromatic stretching), and three larger dots representing poly methyl methacrylate (PMMA) beads (v _s (C-H) of O-CH ₃ at 2957 cm ^"1 ). When viewed in colour, the smaller dots are in red and the larger dots are in green. Each SRS signal for each pixel of the image is determined based on a single calculated scattering value. That is to say, the SRS signal is not based on an averaged scattering value. The image is acquired at an imaging speed of 20 fps (i.e., a line dwell time of 100 μδ). The cropped portion shows a clear distribution of PS and PMMA beads with a high chemical specificity and without nonresonant background. The spatial resolution of the developed line-scan SRS imaging technique is calculated to be approximately 500 nm (lateral) and 1 .5 pm (axial) from the SRS intensity profile along the lateral and axial directions of a PS bead (500 nm in size) after deconvolution (see Figure 7(b), which is similar to the theoretical calculations of the point-scan SRS imaging technique). Figure 8 shows a series of corresponding cropped portions of images obtained using the system 100 at 2957 cm ^"1 of 1 .0 pm PMMA beads acquired at intervals of 200 ms from 0 second to 1 second (i.e., 0 ms, 200 ms, 400 ms, 600 ms, 800 ms, and 1000 ms) during the Brownian motion of the beads in water, with a pump beam excitation power l _pump of 0.3 mW/pm ² and a Stokes beam exicitation power Istokes of 30 mW/pm ² . A scale bar of 5 pm is shown in each cropped portion. The scattering value calculated for each pixel of each image is not averaged. The frame rate is 20 fps. Each cropped portion has a field of view of 25 pm x 25 pm (i.e., 128 χ 128 pixels). The system 100 performing the method 200 is suitable for capturing the Brownian motion processes of PMMA beads resulting from their collisions with water molecules.

Figure 9 shows a series of images (512 χ 512 pixels, 20 fps) with a field of view of 100 m x 100 pm, obtained at intervals of two hours over a period of 10 hours, using the system 100 at 2940 cm ^"1 (CH ₃ stretching of proteins), of living gastric cancer cells (MKN28) during epithelial-mesenchymal transition (EMT) processes induced by transforming growth factor β1 (TGFpi ). A scale bar 25 pm is shown in each image. The pump beam excitation power density l _pump is 0.3 mW/pm ² , the Stokes beam excitation power density Istokes is 30 mW/pm ² , and the line dwell time is 100 με.

The images show that after 2 hours of TGFpi treatment, the gastric cancer cells change to spindle-like in cell morphology, and the cell shrinkage and fibroblast- like morphological changes become more visible after 8 hours of TGFpi treatment. With the low excitation power density and rapid line-scan SRS imaging, no photodamage is observed in the cells and the function of the cells is not adversely influenced during the 10 hour observation period.

For the experiment of Figure 9, the maximum power of the pump beam is restricted to 15 mW in light of the full well capacity of the linear detector 160. According to Equation 6, assuming that ltotai =l _P um _P + Istokes, the SNR of the SRS signal can be expressed as:

(7)

The maximum of SNR can thus be obtained when the differentiation of Equation 5 is equal to zero. That is:

0.5k -

Hence, the maximum SNR is achieved when the intensity ratio of pump beam to Stokes beam is 1 :2. With a total excitation power of 1015 mW (i.e., l _pump = 15 mW; Istokes =1000 mW), the resultant SNR is approximately 0.12k, which is 3.2- fold lower than the maximum SNR value of 0.35k (Equation 7). Linear photodetectors with a larger full well capacity and a higher line rate may be used to improve performance, achieving a higher SNR.

Other alternative embodiments are described below.

In an alternative embodiment, with reference to Figure 3B, a second scattering value may be further calculated in step 220 by subtracting a fourth intensity value l ₄ from a third intensity value l ₃ , a third scattering value may be further calculated by subtracting a sixth intensity value l ₆ from a fifth intensity value l ₅ , and a fourth scattering value may be further calculated by subtracting an eighth intensity value l ₈ from a seventh intensity value l ₇ . Next, in step 230, the SRS signal may be determined further based on the second, third and fourth scattering values. For example, the SRS signal may be calculated based on a result of averaging the first to fourth scattering values calculated for a particular pixel of a particular image.

In other embodiments, the intensity level of the modulated Stokes beam in the second modulation state may be reduced and different from the intensity of the Stokes beam in the first modulation state.