2.1.

Sample Processing

Exactly 5 ml of venous blood from each of fourteen healthy women (age range 25 to 60 years) was collected in a violet sterile vial that contained the anticoagulant ethylenediamine tetra-acetic acid (EDTA). The vial was gently rocked five times to adequately mix the EDTA with the whole blood, and the sample was centrifuged at a rotational speed (rpm) of 3000 for 15 min. Clear, pale, greenish-yellow plasma supernatant was obtained by this centrifugation. A total of 1.5 ml of supernatant plasma (containing mostly proteins and enzymes) and buffy coat (containing mostly white blood cells) were removed and discarded leaving the formed elements (FE) undisturbed.

Then, exactly 1 ml of the thick FE from the bottom layer, which contained mostly red blood cells (RBCs), was removed to a sterile vial and mixed with 2 ml of analytical grade acetone. Proper care was taken to ensure that the FE did not develop lumps. After thorough mixing to enable the acetone extract fluorophores within and around the cells, the sample was centrifuged again (3000 rpm for 15 min). The resulting supernatant was subjected to steady-state fluorescence emission spectra (FES) and time-resolved spectra.

The same protocol was used to process blood samples from the cervical cancer patients ($N=13$, age range from 25 to 60) at the Government Hospital in Royapettah, Chennai. The classification of cervical cancer subjects was made by the histopathological examination as early stages (Stage I and Stage II). All of the above said patients were confirmed cases of cancer, waiting for the onset of treatment. They had been informed about the research investigation and consent from each had been obtained. Suitable permission had been given from the Internal Review Board.

2.2.

Spectral Measurements

Steady-state FES for the acetone extract of samples (control and cervical cancer patients) were obtained by PerkinElmer Spectrofluorometer LS 55. Each sample, taken in a quartz cuvette of $1\mathrm{cm}\times 1\mathrm{cm}\times 4\mathrm{cm}$, was excited at 400 nm, and the fluorescence signal from the sample was collected (transversely) and scanned from 425 to 700 nm. This was the spectral domain analysis.

The same set of samples was used to obtain transient, excited-state laser-induced fluorescence spectra, employing the TCSPC technique. This was the time domain analysis. Our attempt was to establish correspondence between these two types of spectral analyses.

The fluorescence decay measurements were carried out with a micro-channel plate photomultiplier tube (MCP-PMT) as detector and a picosecond laser as the excitation source. The laser used was mode locked Ti-Sapphire Tsunami laser (Spectra Physics, 3960) capable of generating $\mathrm{ps}/\mathrm{fs}$ laser pulses at 800 nm. Such pulses were passed through the pulse picker (Spectra Physics, 3980 2S) to generate 35 ps 4 MHz pulses from the 82 MHz train of pulses. Following this, the second-harmonic laser (400 nm) output was obtained from the flexible harmonic generator (Spectra Physics, Santa Clara, California). The vertically polarized 400 nm laser beam was used to excite the sample.

The fluorescence emission at the magic angle (54.7 deg) was acquired by an MCP-PMT (Hamamatsu R 3809U, Tokyo, Japan) after passing through the monochromator and processed through a constant fraction discriminator, a time to amplitude converter, and multichannel analyzer modules. The instrument response function for this system was 50 ps. Since we are interested in studying the porphyrin lifetime of blood components at 630 nm, a band-pass filter was used before the PMT to obtain the desired native fluorescence and to exclude undesired fluorescence from other sources.

The measured fluorescence decay was the convolution of the true fluorescence decay, the excitation function, and the instrument response function. The fluorescence kinetic parameters (lifetime, amplitude, etc.) were obtained by deconvoluting the excitation function and the instrument response function from the measured fluorescence decay function. The data analysis was carried out by the software provided by the fluorocube of Horiba Jobin Yuvon (formerly IBH product) IBH (DAS-6), which was based on the reconvolution technique using iterative nonlinear least-squares methods.¹¹

The fluorescence decay curves obtained from the sample were fitted to single, double, and triple exponential decay functions. The three component approaches improved the fitting, and the quality of fit was judged by the reduced ${\chi }^2$ value. The three exponential functions used to fit the data were

The fractional contribution of each component to the fluorescence intensity could be defined as

The average lifetime for the three exponential decay was

2.3.

Statistical Analysis

Statistical analyses were performed to evaluate the diagnostic potential of the collected time-resolved spectra of blood components of normal and cancer patients using ANOVA discriminant analysis commonly available in the SPSS +11 software. This technique assigns a case to group K, if its score on the function for that group is greater than its score on the function for any other group. This function called the Fisher classification function will give $n$ classification function coefficients and ($n-1$) canonical discriminant functions.¹¹ To check the classification results, we used “leave-one-out cross-validation” method. The algorithm of “leave-one-out cross-validation” involves classifying each case into a group according to the classification functions computed from all the data except the case being classified.¹² Such an analysis gave a measure of reliability of collected time-resolved data.

3.1.

Steady-State FES

Figure 1(a) shows the typical FES of acetone extract of cellular components in the normal control. This spectrum has three important fluorescence peaks: one at 500 nm (due to Flavin adenine dinucleotide), next at 585 nm, and third at 630 nm (due to basic and neutral forms of porphyrin). Figure 1(b) shows the typical FES of acetone extract of cellular components of a cancer patient. This too has three peaks similar to that of Fig. 1(a), but porphyrin peaks are out of proportion. If we define ratio parameters $R_1=I_{635}/I_{585}$, intensities at 635 and 585 nm, it is 1.1 for the control group, but 2.4 for a cancer patient. That is, the normal porphyrin is twice elevated for the patient. This is true for many other types of cancer, too.^8–12

Fig. 1

(a) Steady-state fluorescence emission spectra (FES) of acetone extract of formed elements (FE) of normal control. (b) Steady-state FES of acetone extract of FE of cervical cancer patients. Both are typical spectra.

3.2.

Fluorescence Lifetime Spectra

The native fluorescence lifetime spectra of porphyrin found in the blood samples of normal patients and cervical cancer patients were acquired at 630 nm emission under 400 nm excitation. Figure 2(a) shows the typical time-resolved spectra for normal samples. A time decay curve, such as the one shown in Fig. 2(a) may be fitted with single, double, or triple exponential decay functions. For us, it was the three component functions that gave the best fit, and the quality of such a fit could be judged by the residuals lying between $-1.4$ and $+1.4$ as shown in Fig. 2(b). A similar set of fluorescent decay curves and residuals for a typical cancer sample are shown in Figs. 3(a) and 3(b), respectively.

Fig. 2

(a) Fluorescence decay curve of FE of normal control and (b) with fitted function and residuals.

Fig. 3

(a) Fluorescence decay curve of FE of cervical cancer patients and (b) fitted functions with residuals.

The three decay times and their amplitudes for normal and cancer samples are shown in Table 1.

Table 1

Mean values of lifetimes and their amplitudes of formed element.

4.1.

Statistical Results

The statistical analysis was performed as described in Sec. 2 to evaluate the diagnostic potential of the time-resolved spectra of blood components using porphyrin as a biomarker. One of the multivariate analyses called linear stepwise discriminant analysis was used to distinguish the diseased blood components from normal blood components. Out of twenty-seven samples, known cases of normal and cervical cancer (each four in number) were selected for preliminary measurement of lifetimes; these averaged out values are in Table 1. With these values as classification tools, three component lifetimes obtained from each sample were given as inputs for the subsequent discriminant analysis for classification into two separate sets.

4.2.

Discriminant Analysis

Discriminant analysis gives the discriminant function, called first canonical discriminant function, that is used to classify the diseased FE from the normal control FE using discriminant scores. Figure 4 shows the scatter plot between the sample number and the discriminant score obtained from the first canonical discriminant function. It shows that normal and cervical cancer patients are well separated.

Fig. 4

Plot between the sample number and values of discriminant function for normal and cervical cancer subjects.

Table 2 shows the classification results of the first discriminant analysis. The classification accuracy for classifying normal subjects and cervical diseased subjects is 88.9% for the original case and 81.5% for the cross-validation case. The discriminant analysis results show that cervical malignant subjects are discriminated with a sensitivity of 92.3% and a specificity of 85.7% for the original case and sensitivity of 76.9% and specificity of 85.7% for the cross-validated case. In other words, 88.9% of the original grouped cases were correctly classified, and 81.5% of cross-validated grouped cases were correctly classified.

Table 2

Classification results of discriminant analysis.