Indi an Journal of Ex perime ntal Bio logy Vol. 4 1, May 2003 , pp. 500-510

Variability of spectra of laser-induced fluorescence of colonic mucosa: Its significance for fluorescence detection of colonic neoplasia

Barbara W Chwirot, Malgorzata Kowalska, Natalia Pl6ciennik, Mariusz Piwillski*, Zbigniew Michniewicz & Stanislaw Chwirot*

Interdi sciplinary Group of Optical Methods of Early Detec tion of Cancer, Institute of General and Molecular Biology, Nicholas Copernic us University, ul. Gagarina 9, PL 87-100 Torull, Poland

*Institute of Phys ics, Ni cholas Copernicus University, ul. Grudziqdzka 5, PL 87-100 Toruil, Po land

To determine the ex te nt o f a natural va ri ab ility of the spectra of the autonuorescence and its signifi cance for a repro­ducibility of different approaches typically used in studies on nuorescence detection of colonic les ions. Two independent se­ries of experiments have been conducted during three years in the same laboratory . Macroscopic ti ssue specimens obtained during operations of patie nts with colonic cancers were studied ill vi/ro o The tissues were excited using UV lines of C.W. He­Cd laser and pulsed nitrogen laser and the autonuorescence spectra were recorded for areas vi sually diagnosed as normal or pathologically changed mucosa. Natural variability of the auto fluorescence spectra of colonic ti ssues seems to be most im­portant fac tor limiting sensitivity and specificity of the di agnostic a lgorithms. The mean flu orescence spectra obtained for normal mucosa and its neoplastic les ions differ significantl y but the differences are difficult to observe because of the hi gh natural va ri ability among the individu al spectra. Further studies of biological basis o f the coloni c autonuoresce nce are nec­essary for a progress in the fi e ld of flu orescence detecti on of coloni c neoplastic lesions.

Keywords : Autofluorescence, Colon ic neoplasma, Colorectal cancer, Fluorescence detection, Laser-i nduced fl uorescence, Multi va riate ana lysis, Ultraviolet excitation. Variability of spectra.

Colorectal cancers are major health problem in the g lobal scale l

. It is now generally accepted that in al­most all cases the disease starts from known neoplas­tic les ions and develops for a rel atively long time of the order of several years. At the same time the can­cers are often detected only in late stages with result­ing low success rate of therapeutical treatments. Screening for the colorectal cancers relies mostly on testing for occult blood and endoscopic examinations. The occult blood tests are neither very sensitive nor spec ific. The endoscopic examinations are expensive and their result depends to a large extent on experi­ence of the endoscopist. Blind biopsies which are of­ten a screening method of choice allow in typical conditions for examination of less than 0.05 % of the colonic area of interest2

.

Since the beginning of 1990' several groups re­ported on studies on a new fluorescence technique of detecting premalignant and malignant lesions in hu­man colon 2. 10. The general idea of all such methods is based on a simple assumption that metabolic and

*For correspondence : E-mail: chwirot @biol.uni .torun.pl chwi rot @phys.uni.torun.pl

Fax: 0048-5661 - 14478

structural changes resulting from neoplastic transfor­mation are reflected in changes in intensity and emis­sion spectra of a native fluorescence (autofluores­cence) of the ti ssues of interest. These changes may result both from variations in a content of endogenous f1uorophores and from differences in architecture of ti ssues and cells. The latter may influence a propaga­tion of the exciting and fluorescence light. However, with a complexity and natural variabili ty of living systems it is very difficult to differentiate between the changes related to occurrence of a pathologic condi­tion and those resulting from the natural variations of metabolic and structural features of ti ssues of interest. During the last decade several groups described diag­nostic algorithms for analysis of the autofluorescence spectra allowing for a sensitive and specific detection of premalignant and malignant colonic les ions. On the other hand none of those method has found a way to clinical applications. It seems that the main problem is just the natural variability of the spectra of the auto­fluorescence of both the regular mucosa and its neo­plastic lesions. The present study has been aimed at investigating the range of such natural spectral varia­tions and also of a sensitivity of the fluorescence methods to the experimental conditions. The main

CHWIROT el al.: VARIABILITY OF SPECTRA OF LASER-INDUCED FLUORESCENCE OF COLONIC MUCOSA 50 I

goal of this study is to assess the significance of both the above mentioned factors for a reproducibility of diagnostic parameters calculated from the ratios of the intensity of the autofluorescence emitted in different spectral bands.

Materials and Methods The materials for study were macroscopic samples

of human colons obtained during surgical operations of patients with colonic tumours. The study was ap­proved by Commission of Ethics of Scientific Re­search at Medical School, Bydgoszcz, Poland .

Spectral measurements were carried out in two se­ries: one in 1997-1998 and the second in 1998-1999 (from hereafter respectively referred to as A and B). During the first series of measurements the autofluo­r~scence spectra were obtained for 296 cases of regu­lar mucosa, mucosal lesions of non - malignant char­acter l adenomatous polyps, colonic cancers, cancer­ous infiltrations under the mucosa and border regions separating tumours and normal mucosa, all present in the material obtained from 47 patients. In the second series 211 spectra were obtained for areas of his­tologic characteristics similar as in the series A lo­cated in the samples obtained from 28 patients. In both the series the same procedures and the same · equipment were used. However, the measurements as such were carried out in the series A and B by differ­ent persons who had in each experiment to align the optics for coupling the laser beams into the fiber and to position the probe with respect to a sample surface in a position suitable for exciting and detecting the autofl uorescence.

Immediately after the resection large samples of colonic tissues were placed in cold saline and trans­ferred to laboratory. Then the specimen was placed on a cold table (+ 4 QC). The areas of interest were local­ised visually and marked with pins of different col­ours. The autofluorescence of the selected areas was excited using either with 325 nm line of a He-Cd c.w. laser or with pulses of 337 nm radiation of a nitrogen laser. At the end of the measurements the tissue mate­rial was fixed in buffered formaline and subject to histologic examinations. Detailed histopathologic de­scriptions of the places subject to fluorescence inves­tigation were then used for differential analysis of the autofluorescence spectra.

Figure 1 shows a schematic representation of the set-up used . The exciting light was delivered to the sampled surface via 6 quartz fibres of a bifurcated cable located symmetrically around the central quartz

fibre (0.2 mm diameter) which collected the light of the autofluorescence (Ocean Optics) . During the measurements the fibre optic probe was held at a right angle at a distance of about 1 mm from the surface of the sample. At such conditions the exciting light illu­minated an area of about 1-2 mm2

. The intensity of the exciting light on the sample surface was of the order of 2 mW/mm2 in 325 nm line of c .w. He-Cd laser (Omnichrome 2056) and about I MW/mm2 for the 100 ps pulses of 337 nm radiation of nitrogen la­ser (Laser Photonics). Typical repetition rate of the nitrogen laser pulses was 10 Hz and thus the mean power delivered to tissue surface was I mW/mm2. The autofluorescence spectra were obtained lI sing S 1000 spectrometer (Ocean Optics) with a resolution of about 10 nm. A multilayer cut-off filter placed at the input of the spectrometer was used to eliminate the UV light reflected from the tissue surface. A small amount of the 337 nm light was transmitted by the filter and produced a second order maximum at 674 nm. The 325 nm line of the He-Cd laser was separated from the 442 nm line using a quartz prism. There are, however, indications that in some cases the diffused 442 nm light may have contaminated the spectra and modified a narrow part of the wide main autofluores­cence maximum. To avoid possible systematic errors the spectral band of 420-460 nm was not taken into account in the analyses concerning a variability of the autofluorescence spectra and in calculations of the diagnostic coefficients. The spectra were corrected for spectral sensitivity of the linear CCD matrix of the spectrometer and normalised to the intensity of the autofluorescence emitted in 507-540 nm band. Both

Con..,utcr flglZQ "ith NO Iloord

~~ Nitrogen ~ (337 run) or "'~

Ucliwn-Cadmium Laser ( 325 run)

O~dtable

Six Excitation Fibers Collecting Fiber

Fig. I - Schematic representation of the experimental set-up

502 IND IAN 1 EXP BIOL, MA Y 2003

the normalised in tensities and the spectral charac teris­tic of the autofluorescence emitted in th at band were similar fo r a regul ar mucosa and neoplasti c lesions of interest.

Results The autotl uorescence spectra were recorded fo r

cen tra l and border regions of the co lonic lesions and for sites within areas visuall y diagnosed as normal mucosa.

However, histologic examinations showed that on many occasions the visually normal mucosa had been in fac t changed pathologically: in the materi al in vesti ­gated in the A seri es the visual diagnos is was con­firmed by hi stopathologists in 6 1cases but in III samples of supposedly normal mucosa several patho­logic states were detected in microscopic examina­tions and among them 4 adenocarcinomas within the mucosa, 2 cases of submucous spread of adenocarci ­noma and I case of anaplast ic carcinoma beneath the mucosa. In the seri es B the visual macroscopic di ag­nosis of a normal mucosa was confirmed for 34 cases and contradicted for 86 cases. The latter included 7

0.036

0.030

~ 0_ ~ i 0020 .. ~ 0 .01$

5 0.010 :5

0.006

cases of adenocarc inoma with in the mucosa. The other pathologic conditions detected microscop ically within the macroscop icall y normal mucosa comprised most often inflammatory reaction, ulceration, Iym­phoplas ia, Iymphonodul oplas ia, oedema, hyperemi a etc. Such results clearly indicate a need fo r develop­ing new auxili ary tools for endoscopic examinations of colonic mucosa.

The mean spectra of the autoflu orescence obtained in the series A and B for the normal mucosa and fo r the pathologic lesions usi ng the 325 nm line excitation are shown in Figs 2-8 (A and B symbols label accordingly the data obtained in the two experimental series).

Similar series of the autofluorescence spectra is presented in Figs 9- 15 for the excitation with 337 nm pulses of nitrogen laser. Slightl y di ffe rent wavelength and a very high transient power of the 100 ps pulses of the nitrogen laser result in di fferen t excitation con­ditions compared to the excitation with c.w. line of He-Cd laser. Thus, one could expect di fferences in absorption probabiliti es, kinetics of the excitati on and emission of the autoflu orescence and in effec ti ve penetrati on depth of the exciting light.

0.036

0.030

~ 0.""

~ ! 0020 .. ~ 0.015

i 0.010

:5 0.006

o.()()() ..o..,-~.-...-.-~..-.--,.-~.,....-..--,--=ot-. ~ ~ ~ - ~ ~ ~ ~ ~ - ~ - ~ ~

Wavelength (rvn) Wavelength (nmJ

r ig. 2 - Average normali sed spectra o f the autonuorescenee of regul ar mucosa exc ited wi th 325 nm line of He-Cd laser (A: n= 48; B: n= 36)

0.036 · 0.036

0.030

I 0.025

i~02O .!!. ~016 f 0.010

0.006

Wavelength (rvnJ

Fig. 3 - Average normali sed spectra of the auto nuorcsecnce of 1ll<!l ig liant co(oni e tumours excited wi th 325 nm line o f He-Cd laser (A: n =55; B: n =46)

I

CHWIROT el (fl. : VA RIABILITY OF SPECTRA OF LASER- INDUCED FLUORESCENCE OF COLON IC M UCOSA 503

All the spectra shown in Figs 2-15 for the groups of the ti ssues of similar histologic characteristics demon­strate a re lati vely large scatter of the results well illus­trated by magnitudes of the standard dev iat ions. The extent of the di ffe rences was similar for the spectra collected from di fferent locations in the material ob­tained from given pati ents and for the spectra ob­tained for similar places in samples from different patients. The variabi I i ty of all the spectra is hi ghest

0.036

0.030

~ 0 025 :> ~ ~ 0.020

..!.. o.Ol ~

t 0.010

:£ 0.01:6

in the short wavelength part and thi s ca n probably be exp lained in terms of differences in a composition and archi tecture of ti ssue layers and of a different blood content of the invest igated ti ssue samples. The auto­fluorescence in the 390 nIll band is usuall y related to co llagen fibres while the minimum around 425 nm is thought to be caused by reabsorption of the fluorescence li ght by hemoglobin. It seems signifi cant that the variability of the autofluorescence spectra of the

0.035

0.030

i 0.025

~ ~ 0.020 .. i 0.015

~ 0.010 ];

0.005

0 .000.l...-~,......'--"-~~"--'-r~:;:.-,J,l..U.j.~ 360 400 _ 000 560 eoo e60 100 "'" 450 500 560 eoo e60 700

Wavelength [nml Wavelength [nml

Fig. 4 - Average norma[i sed spectra of the autofl uorescence or border regions of ma[igmnt tumours excited with 325 nm line of He-Cd laser (A: n= 13; 13: n= 10)

0 0 0.036

0036j

0.030 O.Q30

~ 0.025 I 0.025

f 0.020 ~ i 0020 .. .. f 0.01& f 0.01 5

0.010 0.01 0

:£ :£ 0.0015 0.0015

0.000 0.000 360 400 - 500 560 800 MO 700 360

Wavelength lnml

Fig. 5 - Average normali sed spectra of the autofluorescence or reg ions of cancer spread beneath the regular mucosa exc ited with 325 nm line of He-Cd laser (A: n = 12; 13 : n = 10)

0.036

0.Q30

Wavelength [nml

0.036

0.030

10 = ~ ~ 0.020 .. i 0.016

46 0.010

:£

m 460 aoo &60 eoo e60 700

Wavelength [nml

Fig. 6-- Average normalised spectra of the autofluorescence or adenomatous polyps excited with 325 11m [inc or Hc-Cd lascr (A: n = 12; 13: n = 8)

504 INDIAN J EXP RIOL, MA Y 2003

normal mucosa W2S very similar to that of the spectra of all other type~ of thG lesions. Such an observat ion suggests strongl y that the observed variali ons of th e spectral patterns are dUG at least to a similar extent both to a natural variability of the complex hiologicai systems and to a presence of pathologic conditi ons and host rcsp()n~;es to the disr,:se, espec iall y to neo­plasia.

The l1le,\Ii autofl uorescence spectra o f the normal mucosa and o f the non-tumorous les ions are very similar for the He-Cd l<lscr exci tati on and almost the

o.oa~ Inl I en 0031).1

sll)' i

~:::~ lJ(\j ,

lsi ?'O'5-j oo,~ J.oo~

o

1 o.~~~".,.......-r:

Wawleooth [nm]

same for tissues excited with the nitrogen laser. At the ~;(Imc time there are clear differences between the mean au tofluorescence spectra obtained for normal mucosa 2nd for the neoplastic lesion. of differcnt types. Thc spectra obtained for the normal mucosa and for the neoplastic lesions differ 1110 ·t significantly in tlie short wavelength region i.e. for wavelengths be lmv ca . 475 nm. The regular ti ssues arc clearly charac terised by a more pronounced pcak between 375 and 400 nm wh ile thc I11can spcctra of the lesions typical ly show slightl y higher f1uorescencc intensi tics

f-ig. 7 -- A vcrage l1()nn~liscJ ~J'cc tr~ of thc ~UlOnunrCSCCllCC of hypcrpl;t\t ic polyps exci tcd with 325 nm linc of I lc - Cd l:Jscr (A: n = 3; B: n = 2)

0.035

0.030

&be Me Wavelength [nm}

oo.", ~ J

0030

~ § 0.025

~ ~ 0.020

~ 0 .015

~ 0.010

E

,~~~~-r~~~~ 400 .(50 f\OO 560 eoo 650 700

Wavelength Inm)

Fig. R- A vcrngc normalised spcctr" of autolluorcsccncc of IIOn-lL1!llorous Icsions cxcitcd with 325 nm linc of He-Cd I,,:.cr (A : n =90: \3: n =(9)

0 0 0.035

--1 0.030 0.030

I 0.02' I 0 02.

~ ?

~·-1 ~ 0.020

I::: R rO'b 0.01 0

o~i 0.00.

0.000 0.000 , .00 .60 .00 ~ -.,-,-

360 l!OO e6CI NO 350 400 I.SO bOO Wl 600 060 700

WaveJenath [nm! Wavelength I~m}

Fig. 9- A I'crag.: l101lllali~cd spCCtra of the autonuorescellce of regular r!lllCOS:1 excitcd with 337 11m linc of nit rogcn l a~er (A: /1 = 5 I ; B: /1 == 37)

CHWIROT el al.: VARIAB ILITY OF SPECTRA OF LASER- INDUCED FLUORESCENCE OF COLO IC MUCOSA 505

for longer wavelengths above 550 - 600 nm. It also seems interest ing th at the spectra obtained using the He-Cd and nitrogen laser excitati ons are very similar in the 350 - 650 nm band investi gated in the present work.

A compari son of the data obtained in the two inde­pendent measurement series A and B prov ides an in­terest ing insigh t into a problem of reproducibility of

0.035

0.030

~ 0.025

" ~ ~ 0020

'" ~ 0.015

~ 0.010

£ 0.0011

4tIO !SOO 5tIO 800 tItIO 700

Wavelength (nml

the resul ts and the sensitivity of the autofluorescence spectra to the ex perimental cond it ions. The spectra of the autofluorescence excited us ing 325 nm He - Cd laser line in the series B show a clear indi cati on of the emi ssion peaks around 4.40 nn!. A closer analysi s re­veals the ex istence of simil arl y located but much small er peaks also in the spectra obtained in the series A. Initiall y those peaks were thought to be a charac-

0.035

4OO<tIOtIOOtI50 800

Wavelength (nm]

Fig. 10 - Average normalised spectra of the autolluorescenc<.: or mali gnant colonic tumou rs exc ited with 337 nlll line of nitrogen laser (A: n= 64; B: n = 49)

o 0.0311 0 .036

0.000 ~ J60 400 460 eoo &60 ~ 660 700

Wavelength [nml Wavelength [nml

Fig. I I - Average normalised spectra of the autolluorcscence 0(' border regions 0(' malignant colonic tutllours exciwd with 337 ntll linc of nit rog<.:n laser (A: n = 17; 13: n = 10)

0.035

0.030

:i '5 0

.­~ ~ 0020 .. i 0.0115

, 0.010

£ 0.0011

o .ooo"'-r~-r-~,....-'-~.-~,....~ 3tIO 400 ~ !SOO tI50 800 tItIO 700

Wavelength [nm]

0.036

0.030

~ 0.= " ~ ~ 0.020

'" ~ 0.Q1'

~ 0.01 0

£ 000II

.on 4tlOtIOO 5tI0600tItI01OI:

Wavelangth (nm]

Fig. 12-Average normalised spec tra of the autolluorcsccncc of rcgivns of cancer spread beneath th e: rcgular mucosa exci ted with 337 nm [it~eof llitrogen laser (A: n= 14; B: n= 10)

506 INDIAN J EXP BIOl, MA Y 2003

teri stic fea ture of the autofl uorescence of the ti ssues under in vestiga ti on but arter ca rry ing ou t the suitab le tests it was di scovered th at th ey ,vere related to a dif­fused 442 nm component of the laser beam (ex­tremely weak but strong enough to contam inate the rea l data). Perhaps similar effec ts may be responsible fo r differences in the autofl uorescence spectra re­ported 1'01' colon ic mucosa by different groups lI slng sim il ar excitation wavelengths.

o.~

0.030

:w 0025 :J

~ ~ 0 .020

:e .. i 0.016

~ 0.010

:£ 0.006

"ii

First studies on a fluorescence detec tion of neo­plas ti c les ions suggested that a transformati on to­wards mal ignancy was associated with a signifi cant qual itative change of the auto fluorescence spectra. It was suggested that di fferences in the autolluorescence of normal and neoplas tic ti ssues may resu lt from oc­currence of a new fluorescence compound or from consi derable changes in a poo l of existing endogenous f1uorophores. However, the first very promIsIng re-

O .~

0.030

'c 0.025 :J

~ ~ 0020

'" i 0.015

~ 0.010

:£ 0.'lOO

OOOO~~~~~-r--'---r-~~~-360 400 460 500 MO eoo 660 700 .000 <60 500 ~ 000 e60 700

Wavelenglh [nm] Wavelength [nm]

Fig. [3 - Average normalised speelra or ~ILilofluorescence of aticno:naIOU, polyps exc ited with 337 nm line of ni trogen laser (A: n = 12; B: n = 8)

0 0 0._ o.~

0.030 0.030

f 0 026 I 0.025

~ ~ ~ 0020 ~ 0.020

'" .. ;: 0.015 f 0.015 .~ C) 0.010 0.010 :£ :£

0.006 0.006

Fig. 14 - Average normalised spectra of autofluorescence llf hyperplastic polyps cxc iled wi th 337 nm line of nitrogen laser (A: n = 3; B: n = I )

0.036

0.030

I 0.026

~ ~ 0020

'" i 0.016

~ 0.010

:£ 0.006

o.ooo-'-r-~-r-~~-r-'---'--'-r--'T--360 400 _ 500 660 _ aoo 700

Wavelength [nm]

o .~

0.030

0.006

400 m 500 MO 000 MO 700

Wavelength [nm]

Fig. 15 - Average normalised spectra of the autofl uorescence of non-Iulllorous lesions exc itcd with 337 nm line of nitrogen laser (A : n = 100; B: n=74)

CHWIROT el al.: VARI ABILIT Y OF SPECTR A OF LASER-INDUCED FLUORESCENCE OF COLONIC MUCOSA 507

suits 11 .12 we re not confirmed by furthe r s tudies con­

ducted by other authors. It was soon clear that the dif­fe rences in the spectra were more subtle and had to be detected against a backgro und o f vari a ti o ns due to a complex ity and a vari ability o f the ti ssues o f inte rest.

Generall y two c lasses of di agnosti c a lgorithms have been deve loped and used by different g ro ups . Some autho rs use a stepw ise multi vari ate linear re­gression ana lys is (MYLR) to ana lyse the whole spec­tra and/or to localise severa l spectra l bands re le vant fo r detection of neopl as ia. The others, prefer a more simple approach based on ratioing the inte nsities o f the auto flu orescence emitted in two spectra l bands. T he latter technique brought the first c linica lly tested and commercially avail abl e apparatus fo r the fluores­cence detec tion of premalig nant and mali gnant lesio ns within bronchia l tree l 3 and a patented method fo r de­tecting several cancers in microscopic samples l4

.

More recently Coth ren et al. reported ve ry good re­sults o f cli nical tests of a fluo rescence technique fo r detecting neoplastic les io ns in human colon a lso based on ratioing the intensities of the autofluores­cence emitted in diffe rent parts o f the visible spec­trum . Also the groups using the MLRY algorithms reported ve ry good results o f feas ibility studi es3

.5 but up to now the practica l sig nificance o f those data has not been co nfirmed in full c linical tri a ls . It is poss ibl e that the multiparameter methods may be too suscepti ­ble to multiple local changes in the spectral patterns i.e. the optimum va lues o f the diagnosti c parameters may change from one calibrating set of ti ssues to an­other.

The ratios of the intensities o f the auto fluorescence emitted in the spectral bands which seemed to show most s igni ficant changes re lated to di fferent patho­logic conditions were calculated and used to compare the autofluorescence spectra obtained in the present study in the two independent experimental seri es. From hereafter these intens ity ratios will be refe rred to as the di agnostic coefficients (D j ) calculated ac­cording to fo llowing definitions: seri es A :

D I = 1385-395 / 14 10-470, D 2 = 1 385-475 /1 575-650,

seri es B:

DI = 1385-395 / I4 10-470, D 3 = (1 385-420 + 1460-470) /1 575-650'

The results presented in Tables 1-3 confirm the ex­istence o f considerable diffe rences between the spec­tra of the autofluorescence emitted by histolog ically s imilar tissue structures. The minimum and maximum values o f the D j parameters differed in some cases even by one order o f magnitude for the areas of the

mucosa hi stolog ically c lass ifi ed in the same group. At the sa me time the mean va lues o f the paramete r 0 1

de te rmined in the two experimental seri es A and B are very s imil ar as they ag ree within the li mits of one standard dev iati o n. S uch a res ult sup ports a sugges­ti on th at the main source of the scatter of the autolluo­rescence spectra recorded fo r the ti ssues of the same ty pe is rathe r the charac teri sti cs o f the bi o logical ma­te rial than the chang ing ex pe rimenta l condi tions.

Anothe r inte resting finding is the simil arity of the OJ paramete rs ca lcul ated from the spectra of the auto­fluorescence exc ited with the ultrav io let lines of the He- Cd and nitrogen lasers . The s iight di ffe rence ( 12 nm) o f the excitati on wavelength sho uld no t signi fi­cantl y influence the excitatio n o f the endogenous f1u orophores charac te rised in situ by w ide emi ss ion and excitation max ima (see fo r instance the excita tion­emi ss io n matrices in Richards- Kortum et al.\ How­eve r, bo th a kineti cs o f the exc itation and the effecti ve penetration depth o f the beams of c . w . He-Cd lase r and pul sed nitrogen laser are s ignificantl y d iffe rent. The simil arity o f the spec tra of the auto flu orescence excited by the two lasers can be c learl y seen in Fi gs 2-15 while the data presented in T abl es 1-3 show it in a more quantitati ve manner. A good reprod uc ibility fo und fo r the mean va lues o f the D j parameters in in­dependent measurements togethe r with a s imil arity o f the OJ values obta ined at diffe rent exc itati o n co ncli ­tio ns indicate that the flu orescence di agnostic a lgo­rithms based o n ratios o f spectra ll y reso lved autofl uo­rescence intensities may be transportabl e i.e . they may produce similar results w hile used in c linical condi­ti ons in di ffe rent laborato ries .

The data shown in Tabl es 1-3 illustrate a potenti al o f the autofluo rescence spec tra l analyses fo r a detec­tio n of pathologic conditi ons in human colon and for differentiating between les ions of different types. On the basis of the va lues o f th~ parameter 0 1 o ne can detect with a hi gh probability a presence of adenoma­tou s polyps. Assuming that DI < 1.0 indi cates a pres­ence of such a polyp it is possible at the 325 nm exci ­tation to detect colonic adenomas present in the pre­sent material with a sensitivity C = 100% with a posi­tive predictive value PPY =48% . The re lative ly low value of PPY is due to fa lse positive results obta ined for about 25 % cases o f normal mucosa. The ma lig­nant tumours could be detected with C = 63. 1 % and differenti ated fro m regul ar mucosa with PPY = 57% whi Ie s i mi lar paramete rs fo r non-tumo rous les ions were C = 60.2% and PPY = 59.6%. Simil ar approach applied to the spectra of the autofluorescence exc itcd

508 INDIAN J EXP BIOL, MA Y 2003

Table I - Values of the DI parameter obtai ned in two experimental series for the normal mucosa and pathologic lesions

325 nm excitation (He-Cd laser)

Tissue Series A Series B type Min . Max. Mean ± SD Min. Max. Mean±SO

m.t. 0.46 2. 12 1.I0 ± 0.35 0.50 1.81 1.07±0.3 1 e. m.t. 0.85 1.65 1.15±0.25 0.63 1.52 1.04±0.30 t.i.m. 0.10 1.74 1.03±0.40 0.58 1.26 0.93±0.20 a.p. 0.37 0.89 0.66±0.15 0.37 1.04 0.70±0.17 n.t. 0.52 2.78 1.11 ±0.33 0.59 1.78 1.07 ±0.23 reg. 0.71 1.92 1.16±0.25 0.71 1.70 1.03±0.21

337 nm excitation (n itrogen laser)

m.l. 0.38 2.43 1.I4 ±OAO 0.57 2.02 1.0210.32

e.m.1. 0.69 1.40 1.16 ±0.33 0.54 1.98 1.08±0.40 l.i .m. 0.69 1.66 0.98±0.29 0.72 1.25 0.91 ± 0. 16 a.p. 0.46 1.04 0.67±0.16 0.46 1.06 0.70±0.18 n.t. 0.62 2.99 1.01 ±0.33 0.61 2.22 1.I0 ±0.27 reg. 0.67 2.1 8 1.12±0.26 0.68 1.40 1.01 ±0.17

m.t. = malignant tumour; e.m.1. = edge of malignant tumour; t.i.m. = tumour infiltration beneath the mucosa, a.p. = adenomatous polyp, n.1. = non-tumorous lesions, reg. = regular mucosa.

Table 2 - Values of the parameters O2 and 0 ) obtained using 325 nm excitation for the normal mucosa and pathologic lesions

Tissue Or Series A Or Series B type Min . Max. Mean ±SO Min . Max. Mean ±SO

m.t. 0 .55 8.66 4. 11 ± 1.99 0.62 3.98 2.09±0.82 e.m.t. 1.05 8.05 3.63 ± 1.85 1.06 2.49 1.84±0.49 I.i.m. 1.69 6.82 4.52± 1.49 1.76 4.28 2.58±0.81 a.p. 2.43 11.95 5.91 ±3. 14 1.35 5.33 2.55± 1.03 n.t. 1.41 14.99 5.56±2.47 0.99 7.87 2.84± 1.03 reg. 1.51 14.34 6.25 ±2.67 1.33 7.35 2.92± 1.26

m.l. = malignant tumour; e.m.1. = edge of malignant tumour; l.i.ll1. = tumour infiltration beneath the mucosa, a.p. = adenomatous polyp, n.1. = non-tumorous Icsions, reg. = rcgular mucosa.

Table 3 - Values of thc parameters O2 and DJ obtained using 337 nm excitation for thc normal mucosa and pathologic lesions

Tissue 0 2-Series A 0 3 -Series B

type Min. Max. Mean ±SO Min. Max. Mean ±SO

m.1. 0.65 10.02 3.95± 1.97 0.92 3.48 2.13±0.68 e.m.1. 1.30 7.60 3.58± 1.62 0.91 2.78 1.71 ±0.56 I.i .m. 2.09 6.53 4.47 ± 1.37 1.66 3.73 2.62±0.58 a.p. 2.36 12.26 4.87±2.72 0.91 3.68 2. 15 ±0.69 n.l. 0.52 12.95 5.59±2.17 0.89 5.65 2.82±0.83 reg. 1.66 10.95 5.84±2.01 1.59 4.97 2.84±0.72

(m.1. = malignant tumour; c.m.1. = edge of malignant tumour; t.i.m. = tumour infiltration beneath the mucosa, a.p. = adcnomatous polyp, n.1. = non-tumorous lesions, reg. = rcgular mucosa.

by pulses of the 337 nm radiation yielded for the ade­nomatous polyps C=9L6% with PPY=42.3%, for the malignant tumours C=57.6% and PPY=58.2% and for the non-tumorous lesions C = 51.9% and PPY = 57.7%. Thus, if similar results can be obtained du ring in vivo studies a simple fluorescence algorithm

based on a determination of the D I value can be a use­ful auxiliary technique for a more efficient endoscopic detection of adenomatous polyps.

On the other hand, the malignant tumours can be efficiently detected using another diagnostic pat'ame­ter D2 . The distribution of the values of that parame-

CHWIROT el at.: VARIABILITY OF SPECTRA OF LASER-INDUCED FLUORESCENCE OF COLONIC MUCOSA 509

ter is highly asymmetric and sensitive and quite spe­cific a detection of the malignant les ions can be achieved with the threshold value much higher than the mean va lue of the parameter. Assuming that the mal ignant lesions were ident.ified as those character­ised by D2 < 6.5 it was possible for the 325 nm excita­tion to detect correctly 89.3% of tumours in the mate­ri:ll under study with PPY = 69.4%. The algori thm worked even better for the 337 nm excitation and yielded.

C = 92.9% with PPY = 70.2%.

The D3 parameter did not allow for a good detec­tion of any of the lesions investigated in the study.

It should be noted that the diagnostic parameters discussed above should be considered examples illus­trating the potential of the method. Further analysis of the spectra is under way now and we have found re­cently that the adenornatous polyps can be detected with a better efficiency lI sing a new parameter D 4 = h85.3'15 / 1460-475. For t he He-Cd laser exci tation with a condition that DI < 0.25 we obtained for the polyps C = 88.2% and PPY = 83.3% while a condition of a D j < 0.3 resulted in C = 94.1 % and PPY = 72.7%. The same approach used for the analysis of the data from the material excited with the nitrogen laser yielded respectively C =: 94.1 %, PPY = 84.2% and C = 100%, PPY = 68%.

Discussion The present results demonstrate clearly that the

variability of the spectra of the autofluorescence of both regular colonic mucosa and of its pathologic le­sions is to a large extent a result of the natural vari­Jb ility of the tissues and cells which is typ ical for bio­logical systems. It is also possible that on several oc­casions the spectra obtained for areas histologically classified as neopiastic may in fact have been ob­tai ned for ti ssues of different character located in their vicinity. Such situations may happen because the dys­plastic and malignant les ions may at in itial stages be restricted to small areas of the mucosa and of the ade­nomatous polyps which may not be included wi thin the small region illuminated by the exciting light. However, the mean autofluorescence spectra obtained in the two independent seli es of measurements con­ducted over a period of three years are for all the ti s­sue types very similar. The val;abi lity of the spectra measured by magnitudes of the deviations is very similar too.

It is also interesting that the spectra of the autofluo­rescence excited using ultraviolet li nes of the He- Cd

and nitrogen lasers do not show sign ificant differ­ences in the 350 nm - 650 nm band investigated in thi s work. The observed differences are of the same order (or even smaller) as the differences between the spectra reported for the colonic tissues by different groups using the same excitations wavelengths ( sec for instance Schomacker et at. 5 and Eker et al. lo

) and significant differences can be seen while comparing the spectra obtained by the groups usillg the nitrogen laser (Schomacker et al . .5 , Eker et at. 10) and He-Cd laser exci tation (Kapadia et al.\ II such differences point to the importance of all teps of experimental protocols for a comparability of the data and of the diagnostic algorithms based on the fluo rescence measurements. Several au thors indicated the depend­ence of the spectra on factors like tissue handling, geometry of the excitation and detec.tion of the auto­nuorescenc.e and the intensity of the exciting light.

The relatively high variability of the autonuores­cence spectra found in the present work and also indi ­cated by the discrepancies between the earlier results reported by different groups may explain the difficul­ties with clinical implementation of the diagnostic algorithms based on the MLR V approach. Such algo­rithms typically involve simultaneous analysis of the intensities of the autofluorescence emitted in many narrow spectral bands and values of re levant coeffi ­cients are calculated from the results obtained for a so-called training set of the tissue samples of known histologic characteristics. As such the MLRY algo­rithms are sensi ti ve to change. of the relative intensi­ties of the autofluorescence that may occur locally in the spectra. Such local changes are less s ign ificant for the coefficien ts calculated as ratios of the intensi ties emitted in broad spectral bands7 or for the meihods based on comparing the intensities of the autofluores­cence emi tted in selected wide spectral bands by d if­ferent regions of the. mucosa and measured simultane­olls ly in digital images of the au tofl uorescenceR

•9

•

In conclusion , at the present tage of development of the fluorescence cietection of colonic neopl astic lesions it is the natura! variability of the spectra and of the intensity of the autofluorescence that is the most important factor li miting the sensitivity and the speci­ficity of diagnostic algori thms. Better understanding of a bio logical basis of the autOfluorescence, espe­cially of the origin of differences between the mean nuorescence spectra of the normal colonic mucosa and of its pathologic lesions seems to be a necessary condition for achieving a real progress in the fie ld . Only then it will be possible to select opt imum excita-

510 IND IAN J EXP BIOL, MA Y 2003

tion and detection conditions allowi ng for a sensitive and speci fic observation of the spectral features char­acteristic for the lesions of interest despite the back­ground of natural and experimental variability of the measured spectra.

Acknowledgement This work was supported by a grant from the State

Committee for Scientific Research (KBN no. 4P05B 055 12). The authors thank colleagues from hosp itals in Torun and the staff of histopathological units at those hosp itals for providing histological evaluations of les ions subjected to fluorescence studies.

References I Winawer S J, Fletcher R H, Miller L, Godlee F, Stola r M H,

Mu lrow C D, Wo lf S H, Gl ick S N, Gan iats T G , Bond J H, Rosen L, Zapka J G , Ol sen L S, G iarde ll o F M , Sisk J E, van Antwerp R, Brown-Dav is. Marc ini ak D A & Mayer R J , Co­loreetal cancer screening: C linica l gu idelines and rati onal c, Gastroellterology, 11 2 ( 1997) 594.

2 Baneljec 13 , Miedema 13 & C handrasck har H R, Emi ss ion spectra of coloni c ti ssue and e ndogenous flu orophores, Am J Med Sci, 3 16 ( 1998) 220.

3 Kapadia C R, C utru zzola F W, O'B ricn K M, Stetz M L, Enriquez R & Decke lbaum L I, Laser- induced flu orescence spectroscopy of human coloni c mucosa detection of adeno­matous transfonnation, GastroellteroloMY, 99 ( 1990) 150

4 Ri chards- Kortum R, Rava R P, Petras R, E, Fitzmaurice M, Sivak M & Feld M S, Spectroscopic diagnos is o f co lonic dysplas ia, Photoehelll PllOtOiJiol, 53 ( 199 1) 777.

5 Schomacker K T , Fri so li J K, Compton C C , Flo tte T J, Richtc r J M, Nishioka N S & Deutsch T F, Ultrav io le t laser­induced fluorescence o f colonic ti ssue: basic biology and diagnostic potential , lLIsers Surg Med, 12 ( 1992) 63.

6 Bottirol, G , Croce A C, Locate lli D, Marchesini R, Pig nol i E, Tomati s S , Cuzzoni C, Di Pa lma S , Dalfante M & Spinelli P. Natura l fluorescence o f norma l and neoplas tic human colon: a compre hensive "ex vivo" study, lLI.I'ers Surg Med , 16 (1995) 48.

7 Cothre n R M, Sivak M V, Van Dam J, Pe tras R E, Fitzmauri ce M, Crawford JM , Wu J, Brennan J F, Rava R P, Manoharan R & Fe ld M S, Detec tion of dysplasia at colonoseopy using lase r- induced fluorescence: A blinded s tudy, Gastroilltest Elldose, 44 ( 1996) 168.

8 Wang T D, Van Dam J, C rawford JM , Pre is inger E A, Wang Y & Feld M S , Fluorescence endoscopic imaging of human colonic adeno mas, Gastroellterology, I I I ( 1996) I 182.

9 Chwirot B W, Michniewicz Z, Kowalska M & Nu ssbeutel J , Detec tion of coloni c malignant les io ns by d igita l imag ing of UV laser-induced autofluorescence, Photochelll PholOiJiol, 69 ( 1999) 336.

10 Ecker C , Montan S , Jaramillo E, Ko izum i K, Rubio C , A ndersson- Engels S, Sva nberg K, Svanbe rg S & Slezak P, Clinical spect ra l charac terisatio n o f colo nic mucosal les ions

usi ng autofl uorescence and o-am inolevulini c acid sensiti sation , Gut, 44 ( 1999) 5 I I.

I I Yuan long Y, Yanming Y, Fuming L, Yu fen L & Paozhong. M, C haracteristic autofl uorescence for cancer diagnosis and

its orig in, Lasers Surg Med, 7 ( 1987) 528.

12 Sha Glassman W , Liu C H, Tang G C , Lubicz S & Alfano R, Ultraviole t excited fluorescence spectra from no n-mali gnant and mali gnant ti ssues o f the gy necological trac t, Lasers Life Sci , 5 ( 1992) 49.

13 Pale ie 13 , Lam S, Hung J & MacAul ay C, Detection and loca lisation o f early cancer by imaging techniques , Chest, 99 (1991) 742.

14 Alfano R R, Pradhan A, Tang G C, Das B B & Yoo K M, Optica l spectroscopy o ffers no vel diagnost ic approaches for the medica l proFession. In Laser non-surg ical medicine: New cha llenges fo r an o ld app lica tion, edi ted by Goldmann L (Tech nomic La ncaser-Basel) 199 1.