Optical Image-Guided Cancer Surgery: Challenges and ... · 6/18/2013 · from these molecules (2)....

Review

Optical Image-Guided Cancer Surgery: Challenges andLimitations

Stijn Keereweer1, Pieter B.A.A. Van Driel3, Thomas J.A. Snoeks3, Jeroen D.F. Kerrebijn1,Robert J. Baatenburg de Jong1, Alexander L. Vahrmeijer4, Henricus J.C.M. Sterenborg2, andClemens W.G.M. L€owik3

AbstractOptical image-guided cancer surgery is a promising technique to adequately determine tumormargins by

tumor-specific targeting, potentially resulting in complete resection of tumor tissue with improved survival.

However, identification of the photons coming from the fluorescent contrast agent is complicated by

autofluorescence, optical tissue properties, and accurate fluorescent targeting agents and imaging systems.

All these factors have an important influence on the image that is presented to the surgeon. Considering the

clinical consequences at stake, it is a prerequisite to answer the questions that are essential for the surgeon.

What is optical image-guided surgery and how can it improve patient care? What should the oncologic

surgeon knowabout the fundamental principles of optical imaging to understandwhich conclusions canbe

drawn from the images? And how do the limitations influence clinical decision making? This article

discusses these questions and provides a clear overview of the basic principles and practical applications.

Although there are limitations to the intrinsic capacity of the technique, when practical and technical

surgical possibilities are considered, optical imaging can be a very powerful intraoperative tool in guiding

the future oncologic surgeon toward radical resection and optimal clinical results. Clin Cancer Res; 19(14);

1–10. �2013 AACR.

IntroductionFor cancer surgery with curative intentions, radical resec-

tion (i.e., removal of all cancer cells) is a sine qua non. Toachieve this, the surgeon has to adequately assess the tumorresection margin during the operation. Optical molecularimaging using near-infrared (NIR) fluorescence introducesa revolutionary new approach to address this basic chal-lenge in surgical oncology (1–3).The field of optical imaging emerged in the early 20th

century with the observation of porphyrin fluorescence incertain tumors, but a lack of fundamental knowledge andsuitable optical equipment prevented further develop-ment at the time (4). The finding that photons in the NIRrange (650–900 nm) travel through tissue much deeperthan photons in the visible light range (5) was essential forthe advancement of optical imaging toward clinical prac-tice, and further development gainedmomentum over thepast decade (1, 6). To use optical imaging for visualization

of cancer, fluorescent agents are injected that emit light inthe NIR range and are tumor specific using a variety oftargeting strategies (2). Because the human eye is notsensitive for light in the NIR region, dedicated camerasystems are required to detect the fluorescence emissionfrom these molecules (2). With real-time intraoperativefluorescence imaging of tumor margins, the techniquepromises to guide the oncologic surgeon toward optimalradical resection and clinical results.

Preclinically, optical imaging has been used in tumoridentification, image-guided resection, therapymonitoring,and detection of sentinel lymph nodes. Because tumor-specific agents were not yet approved for clinical use, thefirst clinical studies were conducted using nonspecific fluo-rescent agents that had long been approved for differentapplications and could therefore be used for sentinel lymphnodemapping (6, 7). Clinical trials of tumor imaging usingindocyanine green have been reported in hepatobiliary andcolorectal cancer (8). Recently, a milestone step was com-pleted when the first-in-human trial using tumor-specifictargeting was reported (9), marking the beginning of a newphase of optical image-guided surgery.

However, although it is very likely that the technique willdeliver an important contribution to surgical oncology,this approach has fundamental limitatioins that influencethe ability to differentiate the targeted tissue from its sur-roundings.Now that thefirst clinical trial has been reported,the group of oncologic surgeons that gets acquainted withthe flourishing field of optical image-guided surgery is

Authors'Affiliations: 1Department ofOtorhinolaryngologyHeadandNeckSurgery; 2Center of Optical Diagnostics and Therapy, Erasmus MedicalCenter, Rotterdam; 3Radiology andMolecular Imaging; and 4Department ofSurgery, Leiden University Medical Center, Leiden, the Netherlands

Corresponding Author: Stijn Keereweer, Department of Otorhinolaryn-gologyHead andNeckSurgery, ErasmusMedical Center, 's-Gravendijkwal230, 3015 CE, Rotterdam, the Netherlands. Phone: 010-70-40120; Fax:010-70-35660; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-12-3598

�2013 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org OF1

Cancer Research. on December 30, 2020. © 2013 American Association forclincancerres.aacrjournals.org Downloaded from

Published OnlineFirst May 14, 2013; DOI: 10.1158/1078-0432.CCR-12-3598

http://clincancerres.aacrjournals.org/

increasing day by day, with clinical consequences at stake.As we show in this review, it is essential for the clinician tounderstand which phenomena occur when photons travelthrough tissues and how these phenomena influence theoptical image that is acquired. We therefore believe that it isnow, more than ever, of paramount importance to providethe surgical oncologist a clear overview of the basic princi-ples of optical image-guided surgery to be able to under-stand how these limitations influence the sensitivity of thistechnique. Awareness of these limitations and focusingresearch on solving these challenges are of utmost impor-tance for proper implementation and evaluation of theclinical value of this new technology.

The Basic Principle of Fluorescence ImagingThe geometric principle of fluorescence imaging is illus-

trated in Fig. 1A. A light source is required that sends out abundle of light with a specific wavelength that is able toexcite the fluorophore (i.e., excitation photons). Becausethe targeted fluorophore is located under the tissue surface,this excitation light has to enter and travel through tissue toreach the fluorophore. Light entering the tissue is partlyinfluenced by reflection and refraction at the tissue surface.The direction of photons that travel through the tissue canbe changed due to scattering. In addition, photons can beabsorbed by various components in the tissue. Only whensuch a component is a fluorophore, which can be eitheran intrinsic tissue component (i.e., autofluorescence) or aninjected external fluorescent agent, absorption of a photonresults in a gain in energy of the fluorophore, which thenenters an excited state (Fig. 1B). The electrons remain in this

state for about 10�8 seconds depending on the molecule,which is called the lifetime of the fluorophore. After thisphase, the system returns to its ground state and emits thephoton, which could be heading into all directions (Fig.1B). The change in energy between the absorbed and theemitted photon results in a change in wavelength. This shiftof shorter wavelengths (higher energy) of the absorptionspectrum to longer wavelengths (lower energy) of theemitted fluorophores is called the Stokes shift. After exci-tation of the fluorophore, emission photons will also beinfluenced by scattering and absorption, and the photonsthat reach the tissue surface will subsequently be influencedagain by reflection and refraction at the surface. Finally, in aclinical geometry, only part of these photons can eventuallybe detected by the camera system. Although optical imagingis based on the detection of photons, Fig. 2 illustrates thatthe path that photons travel through the imaged tissue isinfluenced by optical properties of the components in thetissue (i.e., absorption and scattering). As a result, fluores-cence images will display intensities that are strongly influ-enced by varying absorption properties of the tissue and theimages may be blurred due to scattering of light. Thefollowing paragraphs explain in detail how these phenom-ena influence the acquired image.

What Happens When Light Travels throughTissue?Absorption

All tissues contain components that can absorb photonsthat travel through them. The composition of these com-ponents varies between different tissue types and organs,

© 2013 American Association for Cancer ResearchCCR Reviews

Camera

Nucleus

e

Fluorescent emission

Excitation

Air

Tissue

A B

Fem

Fex

Light source

Figure1. Geometric andbasic principle offluorescence imaging.A, light of the appropriate excitationwavelength is selectedusing a filter (Fex), which is locatedbetween a light source and the tissue. The excitation light travels through the tissue and is absorbed by a fluorophore, which subsequently emits light of adifferent wavelength. A small portion of this emitted light will exit the tissue and can be detected with a camera. A filter (Fem) is placed in front of the camera,which allows only the emitted light to pass into the camera. B, absorbed light by the fluorophore instigates an electron (e) in the ground state toward anelectronically excited state. Upon return to the ground state, the fluorophore emits a photon. The wavelength of this emitted photon is specific for thefluorophore. Fex, excitation filter; Fem, emission filter.

Keereweer et al.

Clin Cancer Res; 19(14) July 15, 2013 Clinical Cancer ResearchOF2




but also within these tissues, and may vary in time (10). Ingeneral, the most relevant absorbers of photons are water,lipids, oxyhemoglobin, and deoxyhemoglobin (Fig. 3A)(11). Blood is the main absorber in the visible region, andthe absorption of light by blood is the highest in the blue-green region. This fact can be illustrated by illuminating thehuman skin with white light. Because of the absorption ofphotons with shorter wavelengths, a reddish reflectionreturns from the skin. In this way, absorbers influence thecolor of the light that exits imaged tissues. In the NIR range,the absorbance per volume of these components is muchlower than in other parts of the spectrum, allowing fordeeper penetration of photons into the tissue (Fig. 3).

ScatteringNext to absorption of the photon, a change in photon

direction can occur, an effect known as scattering (Fig. 4).In most parts of the visible and NIR region, scatteringevents occur much more frequently than absorptionevents. Even though a forward direction of scatteringphotons is most likely to occur in tissues, the accumula-tion of multiple consecutive scatter events will result in agradual randomization of the propagation direction (i.e.,diffuse light, Fig. 4; ref. 12). Randomization of the direc-tion of light reduces the signal strength as well as theaccuracy of determination of the source localization.However, scattering can also have a positive effect onthe fluorescence signal intensity because of its effect onthe excitation light. The photons that excite the fluoro-phore follow a similar contorted light path as the emis-sion light. In highly scattering tissue, the contorted lightpath causes retention of photons in that specific area. Inother words, in a highly scattering medium, the photonsbounce around locally for a longer time, increasing thechance for a fluorescence event. Consequently, there canbe increased fluorescence intensity in tissues with morescattering compared with tissues with less scattering (13).The net effect of scattering on fluorescence imagingdepends on the exact optical properties of the tissue athand.

Influence of optical properties on light penetrationdepth

Scattering and absorption coefficients vary between loca-tions and tissues due to the different components withinthe heterogeneous tissue. Importantly, the influence ofthese phenomena on the fluorescence signal will be higherwhen fluorescent light has to emerge from larger tissuedepths and consequently has to pass more absorbers andscattering events. In the visible light region, absorption bybiologic chromophores limits the penetration depth to afewmillimeters with a maximum of 10millimeters. Because


AbsorptionScattering

Reflection

Figure 2. Light propagation through tissue. Light travelling through tissueis subject to reflection, scattering, and absorption.

CCR Reviews

6000

0.05

0.10

0.15

0.20

0.25

0.30

700 800

Ab

sorp

tion

fact

or

(cm

�1 )

900Wavelength (nm)A B

Water

DeoxyHb

OxyHb

1,000 1,100

400 nm

0.5 mm1.0 mm1.5 mm

5.0 mm

>10 mm

475 nm 510 nm 600 nm 700 nm

Figure 3. Absorption of light.A, absorption of light by variouscomponents varies over thewavelength spectrum, resultingin an optimal window forfluorescence imaging in theNIRlight region between 650 and900 nm. Figure 3A is reprintedfrom ref. 11, with permission,from John Wiley & Sons. B,penetration of light range from0.5 mm to more than 10 mmcontingent to the wavelength.OxyHb, oxygenatedhemoglobin; DeoxyHb,deoxygenated hemoglobin.

Optical Image–Guided Cancer Surgery

www.aacrjournals.org Clin Cancer Res; 19(14) July 15, 2013 OF3




of less absorption and scattering, penetration depth isincreased to more than a centimeter in the NIR region(14, 15); however, this measurement strongly depends onthe type of tissue (refs. 1, 3, 16; Fig. 5).

Reflection and refractionAs imaging during surgery always focuses on a surface,

the mismatch between the indices of refraction of tissueand air is unavoidable, resulting in a change in directionof the photons. This mismatch in index of refractioncauses reflections to occur at the surface. The excitationlight will be partially reflected from the surface, like thesun on a water surface. When the light is nearly perpen-dicular to the surface of the tissue, the surface reflection isnot more than a few percent. However, emitted fluores-cent light that is generated inside the tissue is diffuse andpartly reaches the surface under large angles, resulting incomplete reflection within the tissue. This internal diffusereflection coefficient at the surface can be in the order of50%. Consequently, the emitted fluorescent light is evenmore diffuse, and the amount of light that escapes thesurface and can be detected by the fluorescence camera isfurther decreased.

Fluorescence Imaging and AutofluorescenceThe goal of fluorescence imaging is to detect a target by

its specific fluorescence signal. Detection of the fluores-cent signal of the targeted fluorophore is hamperedbecause all cells contain various endogenous fluoro-

phores that become fluorescent when excited by UV,visible, or NIR radiation of suitable wavelength. Thisintrinsic fluorescence of the tissue induces a nonspecificbackground signal. The number of endogenous fluoro-phores varies strongly between tissue types. Moreover, theautofluorescence can change in time because of bleachingof the endogenous fluorophores and because some of theendogenous fluorophores are related to metabolism (e.g.,NADH, NAD, FAD, and FADH).

Because we can see cancerous lesions with our eyes, theoptical properties of these lesions are changed by defini-tion, and a difference in autofluorescence signal com-pared with the surroundings can be detected. This effecthas been used for autofluorescence-guided surgery (17).However, due to the aforementioned effects, "tumor-specific" autofluorescence signals can vary over time,making it an unreliable target. Furthermore, other benignvisible changes (e.g., scar formation) also result in achange in autofluorescence, limiting the specificity ofthis technique (18). Most importantly, despite positivecorrelations that have been reported in a number ofstudies that use autofluorescence to detect and removethe tumor (17), it is not clear which biologic aspects areresponsible for the change in autofluorescence signal inthe imaged lesions. Studies to assess these aspects arecompromised because of artifacts as a result of colorchanges in cancer tissue (18). The surgeon should there-fore approach changes in autofluorescence with the samelevel of uncertainty as visible clues of cancerous tissue. If

CCR Reviews

Camera

Reflection

Reflection

Forward scatter

Fluorophore

Back scatter

Refraction

© 2013 American Association for Cancer Research

Light source Figure 4. Photons changedirection multiple times whentraveling through tissue. In thecase of forward scatter, aphoton travels more or less inthe same direction before andafter the scattering event. In thecase of backscatter, the photonwill end up traveling in theopposite direction after thescattering event. Duringfluorescence imaging, thedirection of both the excitationlight and the emitted light israndomized due to scattering.Only a small portion of theexcitation light will reach afluorophore, and only thecamera will capture a smallportion of the emitted light. As aresult, it is often difficult topinpoint the exact origin of thedetected fluorescent signal.

Keereweer et al.





high correlation of autofluorescence and tumor tissue isconsistently reported, they can be used to aid the surgeonin assessing the tumor margins but should not be con-sidered as tumor-specific proof of tumor margins.As a good alternative, a target-specific agent that contains

a fluorophore is usually injected intravenously. Now thatwe understand how optical properties can influence theoptical path of the photon, we are confrontedwith a secondchallenge: How can we determine if the photon that wedetect has been excited by the targeted fluorophore, and notby its surrounding autofluorescent components?For the targeted cancer cells to be detected, the signal of

the target-specific fluorescence agent must be significantlyhigher than the nonspecific autofluorescence. Althoughautofluorescence is much lower when NIR light is used,the signal-to-background ratio (SBR) must be sufficient todistinguish the photons of the target-specific fluorescentagent from the autofluorescence signal.

What Are the Consequences of These Effects forFluorescence Imaging of Cancer?

We have now set the stage for optical imaging in tissuesand gained a clear picture of the problems thatwe facewhenwe aim to detect a minimal amount of tumor-specificfluorescence agent within this diffuse heterogeneous medi-um containing absorbers, scatter events, and endogenousfluorophores. First, the fluorescence intensity is not onlyinfluenced by the concentration and fluorescence quantumyield of the fluorescent agent but also by the tissue opticalproperties that are involved. For example, as a result ofabsorption by blood, fluorescence signals in organs withhigh blood volumes, such as the liver (19) or highly vascu-larized tumors, or in tumor cells that co-opt host vessels(20) may appear lower than surrounding less-absorbingtissue, even if they contain larger amounts of contrast agentcompared with the surrounding tissue (21). Similarly, less-absorbing lesions (e.g., cysts and lymph nodes) may seem


A

Color NIR fluorescenceReal-time color-

fluorescence merge

B C

D E F

G H I

J K L

Figure 5. Effect of optical tissueproperties during real-timeintraoperative optical imagingof liver metastases in humans.The consequences ofabsorption and scattering onthe image are essential forunderstanding the limitations ofoptical image-guided surgery:At the surface, fluorescentagents will appear as a brightand sharply delineated spot.However, the target will alwaysbe surrounded by a halo offluorescent light that wasdirected into the tissue afteremission, scattered aroundlocally, and emitted from thesurface at some distance fromthe target location. Moreover,due to absorption andscattering, an identicalfluorescent agent that islocated deeper within the tissuewill have lower signal intensityand will be imaged as anindistinct blob. An example isshown of a liver metastasis thatis delineated by fluorescencesignal (A–C). When a thick layerof greater omentum covers thearea of interest, fluorescencesignal is not detected (D–F).A second liver metastasis isindicated in a different patient(G–I). However, when the liver isflippedaround and theoppositeside is imaged, high absorptionof the liver that is saturated withblood results in an indistinctfluorescent blob (J–L).






brighter in a heterogeneous environment. The variability inabsorption and scattering between different tissues or eventissue components should be taken into account during theprocess of image-guided surgery. Second, intraoperativecamera systems will not only detect the fluorescence signalof the fluorescent agent but also the autofluorescence in thescanned region, and therefore a sufficient SBR is required. Atpresent, various strategies are being evaluated that representattempts to improve adequate identification and quantifi-cation of the targeted photons.

Strategies to Reduce the Influence of Absorptionand Scattering on the Image

With the state-of-the-art technological advancements, itremains impossible to determine all absorbers and scattereventswithin adiffuse, inhomogeneousmedium.However,calculation methods have been developed that try toestimate the perturbation causedbyoptical tissue propertiesand improve the image by partly correcting for these prop-erties (14). An intraoperative fluorescence imaging systemhas been developed that implements such a correctionscheme for light intensity variation in tissues (21).Improved accuracy was shown within phantoms and post-mortem tissues, independently of optical property variationin tissues. At a 5-fold change of absorption variation withinthe fluorescent lesions, quantification errors were reducedfrom 25% in uncorrected images to 8%using the correctionscheme (21).

A second new technology being investigated in this fieldis fluorescence differential path length spectroscopy, whichdetermines fluorophore concentration based on the fluo-rescence intensity corrected for absorption (22). This facil-itates quantitative concentration measurements even forstrong variations in either background absorption or scat-tering. However, currently this method can only be con-ducted using fiber-optic measurement at a single point. Animaging version and subsequent intraoperative applica-tions have not been developed yet. Finally, tomographicreconstruction techniques (23), Raman spectroscopy (24),and photoacoustic imaging (12, 25) may play importantroles in the future, although currently they are not suitedfor real-time imaging in the surgical theater. Raman spec-troscopy is based on inelastic scattering: the effect onthe frequency of excitation photons that changes uponinteraction with tissue, which is independent of opticaltissue properties or autofluorescence (24, 26). In surface-enhanced Raman spectroscopy (SERS), tumor-specificnanobodies are injected that are able to increase the intrin-sically very low Raman effect, thereby improving detectabil-ity. Promising preclinical results of brain tumor resectionguided by SERS have been reported (27). In photoacousticimaging (also referred to as optoacoustic imaging), thermo-elastic expansion of molecules resulting from laser pulseirradiation causes emission of acoustic waves that can bemeasured by photoacoustic spectroscopy. The resolution inphotoacoustic imaging is not limited by tissue scattering butby the attenuation of acoustic frequencies by tissue (12, 25).

This technique improves deep-tissue imaging but is lesssuitable for image-guided surgery.

Strategies to Distinguish the Target-SpecificFluorescence Signal from Autofluorescence

Twomethods are now under investigation that representattempts to separate the target-specific fluorescence signalfrom the nonspecific autofluorescence. The first is based ondifferences in the fluorescence spectrum between photonsfrom these two fluorescent sources (i.e., spectral unmixing),and the second method exploits differences in the fluores-cence lifetime of the fluorophores (i.e., lifetime imaging).

The concept of spectral unmixing is based on the "signa-ture" emission intensity that each fluorophore has at certainwavelengths, providing the fluorophore with its own spe-cific emission spectrum (28). Using spectral unmixing, thesignal is decomposed into a collection of predefined spectrathat is used todetermine the individual contributionof eachfluorophore. Using calculation models combined with thefluorophore-specific emission spectrum as reference (i.e.,linear unmixing), the contribution of each fluorophore in atotal fluorescent signal can be extracted (29). When spectralsignatures of the fluorophore of interest and the autofluor-escence are known, unmixing these specific fluorescencespectra may result in a more accurate SBR. The first-in-human trial on intraoperative tumor-specific fluorescenceimaging was conducted using a camera system that wasbased on spectral unmixing technology (9, 21). There are,however, fundamental problemswith this approach. Linearunmixing is based on the assumption that the measuredspectrum consists of the sum of the fluorescence spectra ofall the components in the tissue, that is, linear mixing. Asmentioned earlier, inhomogeneous optical properties influ-ence the path that photons travel through the imaged tissueand therefore not only put their own signatures on thefluorescence spectrum but may also do so nonlinearlyowing to the inhomogeneous nature of tissue. Furthermore,the unmixing procedure is a very complex process thatrequires all components contributing to the spectrum tobe known, as well as all of their specific "signature" spectra,which also need to be sufficiently distinctive from oneanother.

The second method, fluorescence lifetime imaging, dis-tinguishes individual fluorophores by their specific tempo-ral decay after excitation. This fluorescence decay (i.e.,lifetime) is a fluorophore-specific characteristic that is notinfluenced by the local concentration of fluorophores, theoptical path, the local excitation intensity, or the localfluorescence detection efficiency. To acquire the character-istic decay curves, a picosecond laser pulse is used forexcitation and fluorescence is measured as a function oftime. On the basis of the specific lifetime of the fluoro-phores, the target-specific fluorescence signal can be distin-guished from the nonspecific autofluorescence. In addition,the temporal response at which photons emerge from thetissue can be measured (i.e., time-domain imaging), whichis used to estimate the concentration and depth of the

Keereweer et al.





fluorescent source. Lifetime imaging requires specific con-ditions in termsof pulsed excitation anddata capturingwithcomplex data processing, which currently makes it a time-consuming technique. Although intraoperative lifetime-based techniques are being developed at a strong pace andhold promising advantages, it will take some time beforethey can be used in a real-time intraoperative setting wheremobility of the equipment and speed of data processing andinterpretation are of the essence (30).

What Is Required for a Target-SpecificFluorescence Image?So far, we have assumed that the fluorescent agent has

been able to specifically target the tissue of interest withsubsequent colocalization between the fluorescent signaland tumor cells. However, to gain a complete insight intothe different phenomena that can influence the opticalimage, it is necessary to understand the challenges oftumor-specific targeting.On the basis of the hallmarks of cancer (31), a growing

variety of tumor-specific targets are available for imaging ofcancer (2). The efficiency of target-specific agents to reachtheir intended target is definedbymany variables, includingaffinity of the agent and abundance of the target receptors orepitopes (16). To achieve target-specific fluorescence imag-ing, the contrast agent has to be delivered to the target,requires adequate contact timewith the target for binding tooccur, and has to be retained by the target while nonboundagents are cleared from the circulation (10). It has beenreported that at saturating doses of the agent, high-affinityantibody uptake is dependent on antigen expression levels.However, at subsaturating doses, the signal is generallylimited by delivery of the agent (32).For the agent tobe effectively delivered to the target,many

barriers in the human body have to be passed. Next toinhibitor proteins present in plasma that can nonspecifi-cally bind to the agent (1), walls of blood vessels provide afirst barrier for drug delivery to targeted tumor tissues. Inmost cases, abnormal neovascularization of the tumoroccurs [i.e., enhanced permeability and retention (EPR)effect] resulting in leaky tumor blood vessels (33). As aresult, macromolecular drugs can traverse the endotheliumof these leaky blood vessels and passively accumulate in theinterstitium of tumor tissues (33). However, the EPR effectis largely dependent on the size of the agent; larger agents areless efficient in crossing the endothelial barrier (34).Tumor growth beyond the size of approximately 1 mm is

dependent on oxygen and supply of nutrients and thereforerequires angiogenesis (35). However, in smaller lesions, theangiogenic switch may not have occurred and drug deliverycould be hampered due to the lack of adequate vasculariza-tion. In other cases, tumor cells may grow alongside preex-istent host vasculature, a process known as vessel co-option(20). In these cases, in which angiogenesis-directed targetingmay not yet be possible, indirect indicators of tumor growthcould beuseful for tumordetection, even in the earliest stagesof carcinogenesis. For example, NIR fluorescence agents that

detect proteases that are involved inmigration of tumor cellsand degradation of the extracellular matrix can allow forimaging of the invasive tumor front (36–38), and preneo-plastic lesions in Barrett esophagus can be identified bytargeting changed patterns of lectin binding (39).

Once extravascular, two barriers for adequate binding ofthe agent remain. First, the agent has to cross extracellularmatrix tissue surrounding the target cells (33). At this point,diffusion into the tumor is sometimes impeded by highhydrostatic pressure of many solid tumors, preventinghomogeneous infiltration of the agent (1). Second, bindingof the agent to specific epitopes of the cancer cell has tooccur, and internalization of the agent can further amplifythe fluorescence signal (23, 40, 41). Therefore, the cellularbasement membrane is the third barrier, which can bepassed using the transporter system of the cell by recep-tor-mediated endocytosis (33). Finally, clearance of non-bound agent from the circulation is required to providesufficient SBRs. This occurs through the liver (i.e., excretioninto bile and feces) and/or the kidney (i.e., excretion intourine). Both the route of clearance and the clearance rate areimportant determinants for the blood half-life time andconsequently the background signal and optimal imagingtime of contrast agents (40).

How Relevant Is All This for the Clinical Practiceof Surgical Oncology?

Optical imaging is a complicated process. To adequatelyinterpret the intraoperative fluorescent image of the tumorthat is presented, the surgeon should comprehend that thetarget-specific fluorescent agent had to get to and stay at thetumor, that part of the signal was reduced and distorted byabsorbers and scatter events, and that the signal had to besubtracted from its autofluorescent surroundings. Consid-ering all these influences that could hamper detection ofcancer cells that are located under the surface, the additionalvalue of NIR fluorescence imaging over conventional sur-gery could be questioned. After all, after amere 3 decades ofextensive research, the fundamental principles of opticalimaging still leave room for error, and it currently seems thatthis approach will never be unambiguous.

To determine the additional value of intraoperative opti-cal imaging, we need to ask ourselves what the fundamentalpurpose of this tool should be for surgical oncology. Inessence, the goal of surgery is to remove all cancer cellswhileminimizing damage to surrounding healthy tissues. Con-sidering the small size of a single cancer cell or groupof cells,the limited target epitopes available for the fluorescentagent and the optical tissue properties and autofluores-cence, detection of the last cancer cells under a surface willnot always be possible. Nevertheless, it is very likely that thecurrent technique of optical imaging, with its limitations,will offer improvement of the conventional surgical practiceto successfully treat the patient and could provide a finalsolution to the conundrum of irradical resection.

The most obvious reason is that, although there arelimitations when considering the cellular level, optical






imaging allows for amore detailed delineation of the tumormargins than the conventional practice of assessmentby palpation and visual aspects of the tumor. It was repeat-edly shown in animal models that tumor margins can beclearly demarcated by optical imaging (10, 38, 42). Con-sequently, the surgeon can resect the tumor based on theseimages with the required tumor-free margin outside thefluorescent tissue. Moreover, in some cases that have thintumor strands that are invisible to the naked eye, a tumor-free margin of up to several centimeters is required for localcontrol using conventional surgical practice. Althougheffects of scattering could result in a fluorescent halo of upto severalmillimeters under certain conditions, detection ofthese tumor strands usingoptical imagingwould still largelyreduce the need of these large tumor-free margins, resultingin improved postoperative functionality. Although yet to beproven in large clinical trials, it is anticipated that thisapproach will already lead to a decrease in the local recur-rence rate and improved patient survival and functionality.

Second, it has been suggested that penetration depthlimitations might not be relevant for surgical practicebecause the surgeon will, by definition, bring the area ofinterest closer to the surface during the surgical procedure(34, 43). When, after resection of the tumor mass, an area isfoundwith persisting fluorescence signal indicating irradicalmargins, this could be resected subsequently until no moresignal is found in a "cut the light-procedure," similar to thetechnique of Mohs surgery. Although a similar procedure,optical imaging will be much more efficient than Mohssurgery due to real-time acquisition and the fact that intrao-perative pathologic evaluation is not required. At the surface,the influence of optical properties will be minimized,improving the sensitivity even in cases of perineural orperivascular growth, as long as the photons from the contrastagent can be differentiated from autofluorescence.

Finally, although it is currently impossible to completelycorrect for an unknown variability in optical properties ofthe tissue, we should not forget that a very advanced systemis available that has the capability to identify distinct tissuecomponents (e.g., blood vessels) during the surgical pro-cess: the surgeon himor herself. The true additional value ofthis technique will become clear once the surgical expertiseto distinguish tissue components becomes incorporatedinto the interpretation of the images. For example, thesurgeon would preferably reinvestigate a suspicious regionthat has no fluorescence signal if it is covered by a bloodvessel as long as the surgeon realizes that vessels can behighly absorbent. Using the knowledge about the tissues athand, optical image-guided surgery will result in a dynamicand flexible process providing the surgeon with valuableadditional information during the entire operation. Thiswill likely require a learning curve that would follow theencouraging example of pioneer work in resection ofmalig-nant gliomas using 5-aminolevulinic acid. Preliminaryintraoperative studies in this field provide detailed reportson these learning processes (44, 45). These studies showedthat photons are highly absorbed by blood or cauterizedtissue debris after monopolar cautery resection and that

suction of these layers or rinsing the surface with saline isrequired for unperturbed assessment of tissue fluorescence.Furthermore, necrotic tumor centers accumulated littlefluorescence, which was not a problem because necrosiscould easily be distinguished under white light and there-fore did not impair fluorescence-guided tumor resection.Tissues were not falsely labeled by fluorescence from bloodcontaminating the tumor cavity, which might have beenexpected if plasma had contained substantial amounts offluorescent agent. Photobleaching after overexposure wasminimized with improved development of fluorophores.Finally, the fluorescent source could be located by manip-ulation of the tissues, as illustrated in Fig. 5. The success ofthis approach has led to a randomized controlled multi-center phase III trial showing improved progression-freesurvival in patients with malignant glioma using fluores-cence-guided surgery (46).

What Level of Accuracy Is Required for theRoutine of Surgical Practice?

Considering the intraoperative flexibility of a trainedsurgeon and the practice of exploration of the area of interestduring surgery, we believe that optical image-guided surgeryhas the potential to greatly reduce the tumor-free safetymargin. If a tumor-free margin could be minimized, post-operative functional outcome could be drastically improvedin many cases where the tumor is surrounded by importantanatomical structures. This would require tumor imagingwith optimal sensitivity and specificity of the tumor-specificagent, followed by resectionof the fluorescent tissuewithoutany excess margin of healthy tissue. It is anticipated thatsensitivity and specificity of the target-specific agent will beoptimized by simultaneously targetingmultiple tumor char-acteristics at different wavelengths (42, 47). To adequatelyassess the fluorescent border, high accuracy of the imagingtechnique is demanded. This would require high-resolutionimages that can be acquired at higher magnifications thanthe regular macroscopic field of view. Although not suitablefor intraoperative use, a system that incorporates both mac-roscopic and microscopic imaging has been tested in anexperimental setting of cancer detection (23). In the mac-roscopic field of view, the smallest lesion detectable with ahighly expressed antigen (e.g., several million antigens/cell)was approximately 600 mm, a volume that approaches pixelresolution of the macroscopic images (110 mm/pixel).Higher magnification of this tumor revealed images downto the single cell level due to higher resolution (23).

These results suggest that incorporation of zoom functioninto the intraoperative camera system would improve sur-gical accuracy by using a combination of macroscopicimaging (to survey the tissue and guide tumor resection)andmicroscopic imaging (to verify clean resectionmargins;ref. 40).However, scattering generally determines the sharp-ness of the in vivo fluorescence image, especially for deeplylocated fluorescent sources, and not the imaging equip-ment. Furthermore, microsurgical or robotic camera sys-tems are only used in some oncological surgical specialties

Keereweer et al.





(e.g., neurosurgery). Therefore, in daily routine and logis-tics of surgical practice in most other cases, the tremor ofthe surgical hand will limit removal of the tumor withaccuracy smaller than 0.5 mm. As a result, the accuracy ofremoving the fluorescent tissue will not only be limited bythe resolution of optical imaging but also by the practicallimitations of the surgical hand. As long as microscopicor robotic surgery is not a routine practice, the discussionon the level of resolution will lie within the margin oftechnical surgical possibilities. In analogy to highly accu-rate stereotactic radiotherapy using Cyberknife that hasa technical error margin of 0.5 to 2 mm (48), it is verylikely that this level of accuracy will be sufficient toadequately treat the patient with minimal unnecessaryloss of functionality.

Challenges for Clinical TranslationAlthougha rapidly increasingamountofdata is supporting

the additional value of optical image-guided surgery forcancer therapy, challenges remain in translation of the pre-clinical experimental setup into routine clinical practice(2, 6). The technique requires development of tumor-specificfluorescence agents and dedicated intraoperative camerasystems. Important progress has been made over the pastfew years in both fields, but it is essential to identify thehurdles that are still impeding successful clinical translation.An intrinsic limitation of development of fluorescence

tumor-specific agents is the fact that such drugs are used asdiagnostic tools instead of therapeutic drugs that requireadministration over a longer period of time. Developmentof diagnostic drugs is therefore subject to lower financialincentives for pharmaceutical companies. The first obvioussteps were taken by conjugation of already clinically avail-able tumor-specific agents (e.g., cetuximab) to fluorophores(2, 49). Until the technique will become available at a largescale, further development of these agents will have to becarried out by nonprofit (i.e., academic) organizations.Furthermore, dedicated imaging systems will have tobecome easily available for a large group of surgeons tostimulate adoption of the technique. Currently, the mostadvanced systems are still only available in the researchsetting of clinical trials (7, 9), whereas other systems arealready commercially available (50, 51). These economic

and implementation issues are critical for successful adop-tion of surgical optical imaging.

ConclusionsOptical imaging has the potential to revolutionize cancer

surgery by real-time fluorescence guidance in discriminat-ing between healthy and diseased tissues and identifyingvital structures. However, absorption, scattering, and auto-fluorescence are fundamental optical tissue properties thatinfluence the images and limit the ability to differentiate thetargeted tissue from its surroundings. Inpursuit of removingall cancer cells by resection of the total tumor-specificfluorescence signal, the oncologic surgeon should carefullyconsider these limitations during the process of image-guided cancer surgery. Although there are limitations tothe intrinsic capacity of the technique, when practical andtechnical surgical possibilities are considered, optical imag-ing can be a very powerful intraoperative tool in guiding thefuture oncologic surgeon toward radical resection and opti-mal clinical results.

Disclosure of Potential Conflicts of InterestH.J.C.M. Sterenborg is employed (other than primary affiliation; e.g.,

consulting) as a CEO in Luminostix bv.Nopotential conflicts of interest weredisclosed by the other authors.

Authors' ContributionsConception and design: S. Keereweer, P.B.A.A. Van Driel, T.J.A. Snoeks,R.J.B. de Jong, H.J.C.M. Sterenborg, C.W.G.M. LowikDevelopment of methodology: S. Keereweer, P.B.A.A. Van DrielAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): S. KeereweerAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): S. KeereweerWriting, review, and/or revisionof themanuscript: S.Keereweer, P.B.A.A.Van Driel, T.J.A. Snoeks, J.D.F. Kerrebijn, R.J.B. de Jong, A.L. Vahrmeijer,H.J.C.M. Sterenborg, C.W.G.M. LowikAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): P.B.A.A. Van Driel, T.J.A. Snoeks,S. KeereweerStudy supervision: A.L. Vahrmeijer

AcknowledgmentsThe authors thank R.H. Thoene for her technical support and valuable

contributions to this work.

Received November 28, 2012; revised April 16, 2013; accepted April 18,2013; published OnlineFirst May 14, 2013.

References1. Frangioni JV. New technologies for human cancer imaging. J Clin

Oncol 2008;26:4012–21.2. Keereweer S, Kerrebijn JD, van Driel PB, Xie B, Kaijzel EL, Snoeks TJ,

et al. Optical image-guided surgery–where do we stand? Mol ImagingBiol 2011;13:199–207.

3. Weissleder R, Pittet MJ. Imaging in the era of molecular oncology.Nature 2008;452:580–9.

4. Rassmussen-Taxdal DS, Ward GE, Figge FH. Fluorescence of humanlymphatic and cancer tissues following high doses of intravenoushematoporphyrin. Cancer 1955;8:78–81.

5. Patterson MS, Chance B, Wilson BC. Time resolved reflectance andtransmittance for the non-invasive measurement of tissue opticalproperties. Appl Opt 1989;28:2331–6.

6. Keereweer S, Hutteman M, Kerrebijn JD, van de Velde CJ, VahrmeijerAL, Lowik CW. Translational optical imaging in diagnosis and treat-ment of cancer. Curr Pharm Biotechnol 2012;13:498–503.

7. Mieog JS, Troyan SL, Hutteman M, Donohoe KJ, van der Vorst JR,Stockdale A, et al. Toward optimization of imaging system and lym-phatic tracer for near-infrared fluorescent sentinel lymph node map-ping in breast cancer. Ann Surg Oncol 2011;18:2483–91.

8. Schaafsma BE, Mieog JS, Hutteman M, van der Vorst JR, Kuppen PJ,Lowik CW, et al. The clinical use of indocyanine green as a near-infrared fluorescent contrast agent for image-guided oncologic sur-gery. J Surg Oncol 2011;104:323–32.

9. van Dam GM, Themelis G, Crane LM, Harlaar NJ, Pleijhuis RG, KelderW, et al. Intraoperative tumor-specific fluorescence imaging in ovarian






cancer by folate receptor-alpha targeting: first in-human results. NatMed 2011;17:1315–9.

10. Frangioni JV. In vivo near-infrared fluorescence imaging. Curr OpinChem Biol 2003;7:626–34.

11. Chance B. Near-infrared images using continuous, phase-modulated,and pulsed lightwith quantitation of blood andblood oxygenation. AnnN Y Acad Sci 1998;838:29–45.

12. Ntziachristos V. Going deeper than microscopy: the optical imagingfrontier in biology. Nat Methods 2010;7:603–14.

13. Star WM. Light dosimetry in vivo. Phys Med Biol 1997;42:763–87.14. Bogaards A, Sterenborg HJCM, Wilson BC. In vivo quantification of

fluorescent molecular markers in real-time: a review to evaluate theperformance of five existing methods. Photodiagnosis PhotodynamicTher 2007;4:170–78.

15. Farrell TJ, Patterson MS. Diffusion modeling of fluorescence in tissue.In: Mycek M, Pogue BW, editors. Handbook of biomedical fluores-cence. New York: Marcel Dekker Inc.; 2003. p. 29–60.

16. Hilderbrand SA,Weissleder R. Near-infrared fluorescence: applicationto in vivo molecular imaging. Curr Opin Chem Biol 2010;14:71–9.

17. Poh CF, Zhang L, Anderson DW, Durham JS,Williams PM, Priddy RW,et al. Fluorescence visualization detection of field alterations in tumormargins of oral cancer patients. Clin Cancer Res 2006;12:6716–22.

18. DeVeldDC,WitjesMJ,SterenborgHJ,Roodenburg JL. The statusof invivo autofluorescence spectroscopy and imaging for oral oncology.Oral Oncol 2005;41:117–31.

19. Hutteman M, Mieog JS, van der Vorst JR, Dijkstra J, Kuppen PJ, vander Laan AM, et al. Intraoperative near-infrared fluorescence imagingof colorectal metastases targeting integrin alpha(v)beta(3) expressionin a syngeneic rat model. Eur J Surg Oncol 2011;37:252–7.

20. Holash J, Maisonpierre PC, Compton D, Boland P, Alexander CR,Zagzag D, et al. Vessel cooption, regression, and growth in tumorsmediated by angiopoietins and VEGF. Science 1999;284:1994–8.

21. Themelis G, Yoo JS, Soh KS, Schulz R, Ntziachristos V. Real-timeintraoperative fluorescence imaging system using light-absorptioncorrection. J Biomed Opt 2009;14:064012.

22. Kruijt B, Kascakova S, de Bruijn HS, van der Ploeg-van den Heuvel A,Sterenborg HJ, Robinson DJ, et al. In vivo quantification of chromo-phore concentration using fluorescence differential path length spec-troscopy. J Biomed Opt 2009;14:034022.

23. Thurber GM, Figueiredo JL, Weissleder R. Detection limits of intrao-perative near infrared imaging for tumor resection. J Surg Oncol2010;102:758–64.

24. Nijssen A, Koljenovic S, Bakker Schut TC, Caspers PJ, Puppels GJ.Towards oncological application of Raman spectroscopy. J Biopho-tonics 2009;2:29–36.

25. Ntziachristos V, Ripoll J,Wang LV,Weissleder R. Looking and listeningto light: the evolution of whole-body photonic imaging. Nat Biotechnol2005;23:313–20.

26. Zavaleta CL, Kircher MF, Gambhir SS. Raman's "effect" on molecularimaging. J Nucl Med 2011;52:1839–44.

27. Kircher MF, de la Zerda A, Jokerst JV, Zavaleta CL, Kempen PJ, MittraE, et al. A brain tumor molecular imaging strategy using a new triple-modality MRI-photoacoustic–Raman nanoparticle. Nat Med 2012;18:829–34.

28. Mansfield JR. Distinguished photons: a review of in vivo spectralfluorescence imaging in small animals. Curr Pharm Biotechnol 2010;11:628–38.

29. Zimmermann T. Spectral imaging and linear unmixing in light micros-copy. Adv Biochem Eng Biotechnol 2005;95:245–65.

30. Sun Y, Hatami N, Yee M, Phipps J, Elson DS, Gorin F, et al. Fluores-cence lifetime imaging microscopy for brain tumor image-guidedsurgery. J Biomed Opt 2010;15:056022.

31. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100:57–70.

32. Thurber GM, Weissleder R. Quantitating antibody uptake in vivo:conditional dependence on antigen expression levels. Mol ImagingBiol 2011;13:623–32.

33. Fang J, Nakamura H, Maeda H. The EPR effect: unique features oftumor blood vessels for drug delivery, factors involved, and limitationsand augmentation of the effect. Adv Drug Deliv Rev 2011;63:136–51.

34. Olafsen T, Wu AM. Antibody vectors for imaging. Semin Nucl Med2010;40:167–81.

35. Naumov GN, Akslen LA, Folkman J. Role of angiogenesis in humantumor dormancy: animal models of the angiogenic switch. Cell Cycle2006;5:1779–87.

36. Blum G, von Degenfeld G, Merchant MJ, Blau HM, Bogyo M. Nonin-vasive optical imaging of cysteine protease activity using fluorescentlyquenched activity-based probes. Nat Chem Biol 2007;3:668–77.

37. Keereweer S, Mieog JS, Mol IM, Van Driel PB, Snoeks TJ, Baatenburgde Jong RJ, et al. Detection of oral squamous cell carcinoma andcervical lymph node metastasis using activatable near-infrared fluo-rescence agents. ArchOtolaryngolHeadNeckSurg 2011;137:609–15.

38. Weissleder R, Tung CH, Mahmood U, Bogdanov A Jr. In vivo imagingof tumors with protease-activated near-infrared fluorescent probes.Nat Biotechnol 1999;17:375–8.

39. Bird-Lieberman EL, Neves AA, Lao-Sirieix P, O'DonovanM, Novelli M,Lovat LB, et al. Molecular imaging using fluorescent lectins permitsrapid endoscopic identification of dysplasia in Barrett's esophagus.Nat Med 2012;18:315–21.

40. Gioux S, Choi HS, Frangioni JV. Image-guided surgery using invisiblenear-infrared light: fundamentals of clinical translation. Mol Imaging2010;9:237–55.

41. Kovar JL, Volcheck W, Sevick-Muraca E, Simpson MA, Olive DM.Characterization and performance of a near-infrared 2-deoxyglucoseoptical imaging agent for mouse cancer models. Anal Biochem 2009;384:254–62.

42. KeereweerS,Mol IM, Vahrmeijer AL, VanDriel PB,Baatenburgde JongRJ, Kerrebijn JD, et al. Dualwavelength tumor targeting for detection ofhypopharyngeal cancer using near-infrared optical imaging in ananimal model. Int J Cancer 2012;131:1633–40.

43. Pleijhuis RG, Langhout GC, Helfrich W, Themelis G, Sarantopoulos A,Crane LM, et al. Near-infrared fluorescence (NIRF) imaging in breast-conserving surgery: assessing intraoperative techniques in tissue-simulating breast phantoms. Eur J Surg Oncol 2011;37:32–9.

44. Stummer W, Stepp H, Moller G, Ehrhardt A, Leonhard M, Reulen HJ.Technical principles for protoporphyrin-IX-fluorescence guidedmicro-surgical resection of malignant glioma tissue. Acta Neurochir (Wien)1998;140:995–1000.

45. Stummer W, Stocker S, Wagner S, Stepp H, Fritsch C, Goetz C, et al.Intraoperative detection of malignant gliomas by 5-aminolevulinicacid-induced porphyrin fluorescence. Neurosurgery 1998;42:518–25; discussion 25–6.

46. StummerW, PichlmeierU,Meinel T,Wiestler OD, Zanella F, ReulenHJ.Fluorescence-guided surgery with 5-aminolevulinic acid for resectionof malignant glioma: a randomised controlled multicentre phase IIItrial. Lancet Oncol 2006;7:392–401.

47. Xie BW, Mol IM, Keereweer S, van Beek ER, Que I, Snoeks TJ, et al.Dual-wavelength imaging of tumor progression by activatable andtargeting near-infrared fluorescent probes in a bioluminescent breastcancer model. PLoS One 2012;7:e31875.

48. Kilby W, Dooley JR, Kuduvalli G, Sayeh S, Maurer CR Jr. The Cyber-KnifeRobotic Radiosurgery System in 2010. TechnolCancerResTreat2010;9:433–52.

49. Rosenthal EL, Kulbersh BD, King T, Chaudhuri TR, Zinn KR. Use offluorescent labeled anti-epidermal growth factor receptor antibody toimage head and neck squamous cell carcinoma xenografts. MolCancer Ther 2007;6:1230–8.

50. NouraS,OhueM,Seki Y, TanakaK,MotooriM, Kishi K, et al. Feasibilityof a lateral region sentinel node biopsy of lower rectal cancer guided byindocyanine green using a near-infrared camera system. Ann SurgOncol 2010;17:144–51.

51. Heath CH, Deep NL, Sweeny L, Zinn KR, Rosenthal EL. Use ofpanitumumab-IRDye800 to image microscopic head and neck cancerin an orthotopic surgical model. Ann Surg Oncol 2012;19:3879–87.

Keereweer et al.





Published OnlineFirst May 14, 2013.Clin Cancer Res Stijn Keereweer, Pieter B.A.A. Van Driel, Thomas J.A. Snoeks, et al. LimitationsOptical Image-Guided Cancer Surgery: Challenges and

Updated version

10.1158/1078-0432.CCR-12-3598doi:

Access the most recent version of this article at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://clincancerres.aacrjournals.org/content/early/2013/06/18/1078-0432.CCR-12-3598To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-12-3598

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/early/2013/06/18/1078-0432.CCR-12-3598


Optical Image-Guided Cancer Surgery: Challenges and ... · 6/18/2013 · from these molecules (2)....

Documents

Transcript of Optical Image-Guided Cancer Surgery: Challenges and ... · 6/18/2013 · from these molecules (2)....