2.1.

Experimental Setup

Two experimental setups were used to acquire the spectra from trp powder and tissues. One setup used a CD-Scan (Mediscience Technology Corp, Cherry Hill, New Jersey). The CD-Scan is a modified LS50B (Perkin Elmer, Shelton, Connecticut) fluorescence spectrophotometer. The CD-Scan uses a pulsed xenon excitation lamp $({\tau }_p=10\mu \mathrm{s})$ and gated photomultiplier tube (PMT) detectors. The delay between the lamp pulse and the detector gate is computer selectable, enabling collection of either fluorescence (zero delay between lamp pulse and detector gate) or phosphorescence (delays greater than the lamp pulse width). The illumination area of the CD-Scan is a function of the excitation monochromator slit width. For the phosphorescence measurements, the excitation monochrometer resolution was set to $10\mathrm{nm}$ , giving an excitation area of approximately $5\times 10\mathrm{mm}$ . The scan rate was $120\mathrm{nm}∕\min$ . The photodetector gate delay can be varied in $1.0\text{-}\mathrm{ms}$ steps.

The second experimental setup shown in Fig. 1 was used to generate ratio maps from the breast tissue specimens. The excitation source was an AlGaN alloy UV LED with emission centered at $300\mathrm{nm}$ with $400\mu \mathrm{W}$ of average power. A narrow-band (NB) interference filter ( $10\mathrm{nm}$ FWHM) was used to block the longer wavelength fluorescence emitted by the LED. The emission was collected through a $200\text{-}\mu \mathrm{m}$ optical fiber inserted inside a metal jacket. The collection region was approximately $0.5\mathrm{mm}$ in diameter. The metal jacket is similar in size to a stereotactic needle such as the type used for fine needle aspiration (FNA) procedures on human breast and the collection fiber can easily be integrated into an FNA system. The digital output of the CCD is connected to a PC via a USB (universal serial bus) connection. The integration time of the CCD is computer controllable and was set to $2\mathrm{s}$ so that both fluorescence and phosphorescence were detected simultaneously. The spectrograph exhibited relatively more efficiency in the visible region than in the UV, compared to the CD-Scan. Thus, after normalizing the spectra to the trp fluorescence peak $(\sim 350\mathrm{nm})$ , the trp phosphorescence normalized “counts” were higher with the spectrograph setup than with the CD-Scan. Specimens were mounted on a two-axis translation stage, enabling spectra to be acquired at different locations on the samples.

Fig. 1

Experimental setup for mapping tissue fluorescence and phosphorescence.

2.2.

Tissue Specimens

Fresh surgical specimens of cancerous and normal breast tissues were acquired under Institutional Review Board approval, from a tissue bank within $24\mathrm{h}$ of surgery. The normal specimens, acquired from subjects undergoing breast reduction surgery, had clearly defined regions of normal glandular and adipose tissue. The malignant specimens were acquired from patients undergoing lumpectomies. These specimens were from the margins of the tumor and were a mix of malignant and normal tissues. Specimens were stored at $4°\mathrm{C}$ and were not subjected to any additional processing prior to measuring the spectra. Specimens were cut into pieces ranging from $\sim 0.5\text{to}\sim 1{\mathrm{cm}}^2$ and placed in a $1\text{-}\times 1\text{-}\mathrm{cm}$ quartz cuvette, prior to mounting on the translation stage. Using the spectrograph setup, the spectra were acquired in a grid pattern with a $0.5\text{-}\text{to}1.0\text{-}\mathrm{mm}$ spacing so as to cover most of the specimen. Each spectrum was integrated for $2\mathrm{s}$ . Intensity ratio maps were generated from the individual spectra. Five cancerous and six normal breast tissue specimens were scanned for a total of $\sim 3000$ locations on the tissue specimens. After completion of a sequence of scans with the spectrograph apparatus, the fluorescence and phosphorescence were measured with the CD-Scan. In fluorescence mode, the photodetector gate was synchronized with the excitation lamp pulse (zero time delay). The much higher efficiency of fluorescence compared to phosphorescence meant that the signal was dominated by the fluorescence. In phosphorescence mode, the photodetector gate delay was varied from $0.5\text{to}6.5\mathrm{ms}$ with respect to the xenon lamp pulse.

3.1.

Trp Powder

Figure 2 shows the room-temperature phosphorescence spectra of D-L trp powder excited at 282, 300, 380, and $400\mathrm{nm}$ and acquired with the CD-Scan. Each spectrum shown in Fig. 2 is the sum of seven spectra, acquired with delays of $1\text{to}7\mathrm{ms}$ in $1\text{-}\mathrm{ms}$ steps. The gate width was $1\mathrm{ms}$ , thus each spectrum shown is the integrated intensity from $t=0.5\text{to}7.5\mathrm{ms}$ . Although it is well known that the trp excitation maximum is at $282\mathrm{nm}$ , and that trp does not have significant absorption at wavelengths longer than $300\mathrm{nm}$ , the phosphorescence intensity with $282\text{-}\mathrm{nm}$ excitation was weaker than with $400\text{-}\mathrm{nm}$ excitation. For the longer wavelength excitations (380 and $400\text{-}\mathrm{nm}$ ), the phosphorescence was blue shifted compared to shorter wavelength excitation (282 and $300\mathrm{nm}$ ). For 380- and $400\text{-}\mathrm{nm}$ excitation, trp exhibits two phosphorescence peaks at 480 and $525\mathrm{nm}$ . For excitation at 282 and $300\mathrm{nm}$ , only the phosphorescence peak at $525\mathrm{nm}$ was observed. Both the mechanism responsible for the shift in phosphorescence and the reason for the more intense phosphorescence at $400\text{-}\mathrm{nm}$ excitation is not known. It may be due to dimers or trimers present in the trp powder.

Fig. 2

Phosphorescence spectra of D-L tryptophan powder for different excitation wavelengths. Spectra are integrated from 0.5- to $7.5\text{-}\mathrm{ms}$ delay with respect to excitation lamp pulse.

Figures 3a and 3b show a 3-D representation of trp phosphorescence plotted as a function of wavelength and photodetector gate delay for ${\lambda }_{\mathrm{ex}}=300$ and $400\mathrm{nm}$ , respectively. The emission intensity at $525\mathrm{nm}$ is plotted in Figs. 3c and 3d as a function of gate delay for ${\lambda }_{\mathrm{ex}}=300$ and $400\mathrm{nm}$ , respectively. For ${\lambda }_{\mathrm{ex}}=300$ , the lifetime is $0.8\mathrm{ms}$ , and for ${\lambda }_{\mathrm{ex}}=400$ , the lifetime is $1.0\mathrm{ms}$ .

Fig. 3

Phosphorescence intensity of D-L tryptophan powder excited with (A) 300 and (B) $400\mathrm{nm}$ as a function of wavelength and gate delay. Intensity at $525\mathrm{nm}$ as a function of gate delay for (C) 300- and (D) $400\text{-}\mathrm{nm}$ excitation.

3.2.

Phosphorescence from ex vivo Human Breast Tissues

Although the phosphorescence intensity in trp powder was greater for excitation at $400\mathrm{nm}$ than for excitation at $300\mathrm{nm}$ , tissues contain several fluorophores (NADH, collagen, and flavins) whose emission wavelengths can overlap with the trp phosphorescence. Therefore, the breast tissue specimens were excited at $300\mathrm{nm}$ to reduce the contribution from these fluorophores.

The main types of tissue types found in human breast are glandular and adipose. Figure 4 shows the phosphorescence from ex vivo human normal and malignant breast tissues acquired with the CD-Scan. The plots shown in Fig. 4 are each the integration of seven spectra acquired with gate delays of $1\text{to}7\mathrm{ms}$ with $1\text{-}\mathrm{ms}$ intervals. The normal tissues exhibited phosphorescence emission from $440\text{to}500\mathrm{nm}$ . Fluorescence can be also observed at $350\mathrm{nm}$ —most likely due to a long “tail” on the lamp emission. The normal tissues showed greater fluorescence and phosphorescence than the malignant tissues, which had almost no detectable phosphorescence.

Fig. 4

Phosphorescence from normal and malignant glandular breast tissues. Signals were integrated for detector gate delays of $1\text{to}7\mathrm{ms}$ .

Typical spectra, excited at $300\mathrm{nm}$ , from cancerous and normal glandular and adipose breast tissues are presented in Fig. 5 . These spectra were acquired with the spectrograph system. The spectra have been normalized to unit maxima. The $2\text{-}\mathrm{s}$ integration time enabled collection of both fluorescence and phosphorescence, and both types of emissions can be observed in Fig. 5. The CCD detector sensitivity is weaker in the UV than the visible, thus, the trp “counts” for fluorescence and phosphorescence do not differ as much as they would have if corrected for spectral response of the detector. Referring to Fig. 5, it can be observed that the adipose tissue with a fluorescence peak at $320\mathrm{nm}$ has a smaller Stokes shift than either the normal glandular or cancerous tissue. This may be a result of lower water content in adipose tissue. A broad trp phosphorescence band from $420\text{to}540\mathrm{nm}$ is observed in both the normal glandular and adipose tissues, which is not present in the cancerous tissues. The presence of trp phosphorescence in normal tissue but not in cancerous tissues suggests that the ratio of trp phosphorescence to fluorescence may be able to distinguish malignant breast tissue from normal glandular or adipose tissue. This difference can be quantified by computing the ratio of emission intensity at $345\mathrm{nm}$ (trp fluorescence) to the intensity at $500\mathrm{nm}$ (trp phosphorescence).

Fig. 5

Combined fluorescence and phosphorescence spectra of normal adipose and glandular breast and cancerous breast tissue. Excitation at $300\mathrm{nm}$ .

A map of the $I_{345}∕I_{500}$ ratio and positions of fluorescence peaks for a normal glandular specimen are shown in Figs. 6a and 6b , respectively. Spectra were acquired at 50 locations on a $\sim 4\text{-}\times 4\text{-}\mathrm{mm}$ specimen. The average value of the $I_{345}∕I_{500}$ ratio is $1.6\pm 0.98$ and the average fluorescence peak wavelength is $349\pm 14\mathrm{nm}$ . Figures 7a and 7b plot the $I_{345}∕I_{500}$ ratio and fluorescence peak position for an adipose specimen, respectively. This specimen was also scanned at a total of 50 locations. The $I_{345}∕I_{500}$ ratio average value is $8.5\pm 1.14$ and the fluorescence peak is at $331\pm 1\mathrm{nm}$ . The larger standard deviation in the Stokes shift of the glandular tissues reflects the greater variation in local protein environment in the glandular tissue, a result of its more complicated structure compared to fat tissue. The $I_{345}∕I_{500}$ ratio for the adipose specimen is a factor of 5 larger than for the glandular tissue, indicating the relatively weaker phosphorescence from fat. The Stokes shift for the adipose tissue is $18\mathrm{nm}$ less than for the glandular tissue, indicating that trp residue in fat tissues have less solvent (water) exposure than in glandular tissues.

Fig. 6

Maps of (a) $I_{345}∕I_{500}$ ratio and (b) fluorescence peak position for normal gandular breast specimen.

Fig. 7

Maps of (a) $I_{345}∕I_{500}$ ratio and (b) fluorescence peak position for adipose breast specimen.

The $I_{345}∕I_{500}$ ratio was calculated at 431 locations on malignant breast tissues and 455 locations on normal glandular and adipose tissues. These ratio values are plotted in Fig. 8 . For the normal specimens (glandular and adipose) the $I_{345}∕I_{500}$ ratio is $6.6\pm 3.2$ and for malignant specimens, the $I_{345}∕I_{500}$ ratio is $27.7\pm 12.6$ . Using a cutoff ratio value of 12, 421 of the 431 malignant and 446 of the 455 nonmalignant locations are correctly identified for a sensitivity of 97.6% and a specificity of 98%.

Fig. 8

Plot of $I_{345}∕I_{500}$ ratio for normal and cancerous breast tissues.

Figure 9a shows a map of the $I_{345}∕I_{500}$ ratio from a malignant breast tissue specimen. The specimen consists of a malignant tumor located on the right side of the figure, adipose tissue in the upper left, and regions of normal glandular tissue in the middle left. The malignant regions distinctly show a higher $I_{345}∕I_{500}$ ratio [darker regions in Fig. 9a] than the adipose or normal glandular tissue. Figure 9b plots the location of the emission peak from the same specimen. The areas of adipose tissue had a smaller Stokes shift, i.e., shorter wavelength fluorescence [lighter regions in Fig. 9b] than glandular tissue.

Fig. 9

Maps of (a) $I_{345}∕I_{500}$ ratio from typical specimen and (b) location of fluorescence peak from same specimen. Right side of specimen is tumor, upper left is adipose tissue, and middle left is normal glandular.