+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
Photodiagnosis and Photodynamic Therapy (2014) xxx, xxx—xxx
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/pdpdt
Protoporphyrin-IX fluorescence guided surgical resection in high-grade gliomas: The potential impact of human colour perception Max Petterssen b, Sarah Eljamel c, Sam Eljamel MBBCh, MD a,∗ a
Department of Neurosurgery, The University of Dundee, UK1 Linkopings University, Scotland, UK c Lothian Radiology Training Programme, Scotland, UK b
KEYWORDS Colourblind; Colour-perception; Fluorescence-scale; Glioblastoma; Surgical resection
Summary Introduction: Protoporphyrin-IX (Pp-IX) fluorescence had been used frequently in recent years to guide microsurgical resection of high-grade gliomas (HGG), particularly following the publication of a randomized controlled trial demonstrating its advantages. However, Pp-IX fluorescence is dependent upon the surgeons’ eyes’ perception of red fluorescent colour. This study was designed to evaluate human eye fluorescence perception and establish a fluorescence scale. Materials and methods: 20 of 108 pre-recorded images from intraoperative fluorescence of HGG were used to construct an 8-panel visual analogue fluorescence scale. The scale was validated by testing 56 participants with normal colour vision and three red-green colour-blind participants. For intra-rater agreement ten participants were tested twice and for inter-observer reliability the whole cohort were tested. Results: The intra- and inter-observer reliability of the scale in normal colour vision participants was excellent. The scale was less reliable in the violet-blue panels of the scale. Colour-blind participants were not able to distinguish between red fluorescence and blue-violet colours. Conclusion: The 8-panel fluorescence scale is valid in differentiating red, pink and blue colours in a fluorescence surgical field among participants with normal colour perception and potentially useful to standardize fluorescence-guided surgery. However, colourblind surgeons should not use fluorescence-guided surgery. © 2014 Published by Elsevier B.V.
∗ Corresponding author at: Department of Neurosurgery, Ninewells Hospital and Medical School, Dundee DD1 9SY, UK. Tel.: +44 1382 425712; fax: +44 1382 496202. E-mail addresses: [email protected], [email protected] (S. Eljamel). 1 www.neurosurgery-mse.com.
http://dx.doi.org/10.1016/j.pdpdt.2014.05.002 1572-1000/© 2014 Published by Elsevier B.V.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
2
M. Petterssen et al.
Introduction
Scale validation
Protoporphyrin-IX (Pp-IX) is a naturally occurring fluorophore that absorbs blue light at 404 nm and fluoresces red light with a peak at 635 nm. Pp-IX is a byproduct of 5-aminolevulinic acid (5-ALA) metabolism in the mitochondria, where normal cells transform Pp-IX into heme [3,18,26]. Furthermore oral 5-ALA at 20 mg/kg body weight given orally induces Pp-IX-fluorescence in high grade gliomas (HGG) improving the rate of surgical resection, and time to tumour progression [6—9,21,23,25]. A multicenter randomized controlled trial had demonstrated Pp-IX fluorescence improved completeness of HGG resection (64% versus 36%) and progression free survival (41% versus 21%) [23]. Several other studies demonstrated gross total resection (GTR) of HGG enhancing lesion was an important prognostic factor [1,5,14,16,24]. However, Pp-IX fluorescence is dependent upon red-colour perception by the surgeons’ eyes. The aim of this study was to evaluate Pp-IX fluorescence colour perception in a cohort of doctors and establish a fluorescence colour scale to standardize fluorescence detection.
The first phase of the study was a feasibility phase to establish how easy was it to carry out this test. Ten healthcare professionals participated in this phase of the study, seven males and 3 females. Each individual passed a colour vision test before participating in the study. 75 intraoperative fluorescence images were selected for this phase. The images depicted variable degrees of fluorescence. Each participant viewed each image separately on Mac computer (MacBook Pro 7.1, Mac OSX 10.6.7, with a 13.3 in. screen set for 1200 × 800 pixels resolution and Colour-LCD profile). The Mac computer display screen was split horizontally to display the scale shown in Fig. 1 without the RGB-colour percentages at the top of the screen and the intraoperative fluorescence image at the bottom of the Mac screen. A total of 465 image areas in 75 images demonstrating variable degrees of fluorescence and the background colours were circled in the images. Each participant was asked to compare the shades of the 465 selected areas to the scale and decide which panel (1—8) had the best correlation to colour and intensity. The participants were able to move their heads freely during these experiments. They were also allowed to take breaks whenever they wanted. There was no time limit to complete the experiments. All tests were performed with one person at a time, to prevent peer influence. A second phase was performed one week after the end of the first phase. Because of the length of time it took to complete the first phase, a smaller set of 10 images with 34 selected areas was used. The whole spectrum of fluorescence was still represented in these images. The order of the pictures was randomly rearranged; the tests were otherwise performed in the same way as in the first phase. In this phase fifty-nine participants with normal colour vision from the healthcare staff were involved (22 males and 37 females) and three additional participants with red-green colour blindness were tested. The 59 participants included the ten participants from the first phase. All participants were tested for colour blindness prior to the tests and only three out of 59 were red-green colour-blind.
Materials and methods Intraoperative Pp-IX fluorescence image database was created from 2009 using recorded images obtained from a Pentera surgical microscope calibrated to detect HGG fluorescence (Carl Zeiss OPMI® PenteroTM , Germany). This surgical microscope was designed to switch between white and blue light at the press of a button, the blue light range used was 373—440 nm. The reflected/emitted light was subjected to a longpass filter to maximize red light detection but allowing sufficient background light back to the surgeon to be able to continue operating under the blue light. The images were also recorded simultaneously on the hard disc of the microscope by the integrated multichip-digital video camera. To prevent distortion of the colours during recording and between cases, the video camera was calibrated before each procedure by using a quality assurance phantom provided by the vender (Zeiss) as part of the system to make sure that the recorded images were of the same intensity, quality and colours. To prevent image deterioration over time the microscope as a unit was subjected to regular quality assurance testing by the vender (Zeiss) to ensure light intensity and all components were in good working order. Initially a random sample of 108 images was used in this study, from which 20 images were selected to construct a fluorescence detection scale. The selection of these images to construct the scale was based on the fact these images were images where all shades of fluorescence could be detected in the same image by two observers (MP&SE). From these 20 images, two hundred and thirty-three colour samples were extracted. The colour samples were arranged in an indexed database. An image-handling program, GiMP 2.6 for Mac (ICC profile: sRGB IEC61966-2.1), combined with the Mac Color sync tool 4.6.2 (for colour preservation) was used. The samples were analyzed for content of red, blue and green expressed as RGB % using Colormeter.app 3.7.2. Then the colour samples were arranged based on their colour content in order to construct the fluorescence scale.
Statistical analysis Because it is likely that only one surgeon will make the decision regarding fluorescence during surgery, differences between colour perception and agreement were evaluated by the test and retest correlation in the 10 participants (Intra-rater agreement), which was analyzed using intraclass correlation (ICC). The two-way random model for absolute agreement was chosen because the images and raters were samples from different populations and the agreement between actual scale values is of interest in this study. The ICC is a descriptive statistic that can be used when quantitative measurements are made on units that are organized into groups. It describes how strongly units in the same group resemble each other. While it is viewed as a type of correlation, unlike most other correlation measures it operates on data structured as groups, rather than data structured as paired observations [13]. To assess agreement between different participants (ICC) of the fluorescence intensity,
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
Impact of human colour perception on Pp-IX fluorescence
Figure 1
Fluorescence visual analogue scale (1—8).
single measures of ICC were used. The inter-rater agreement was calculated with ICC of the whole database including the first 10 participants’ results. Inter-rater bias was analyzed on both the populations (10 and 59 participants) from the first and second phases, using repeated measures analysis of variance. Agreement between participants was measured by calculating a coefficient of variation (CV), where the lower the CV the more agreement and the higher the CV the less agreement between participants. The coefficient of variation (CV) is defined as the ratio of the standard deviation to the mean : Cv =
3
It shows the extent of variability in relation to mean of the population [21]. Values of ICC were interpreted as follows: ≥0.75 excellent, 0.4—0.75 fair to good and 0.13) and less likely to agree on violet-blue colour perception (p < 0.05). The three participants who were red-green colour-blind were unable to detect any fluorescence including the very strong red fluorescence.
Discussion The importance of gross total resection (GTR) of HGG and the use of 5-ALA induced fluorescence to achieve this goal had been highlighted in the introduction of this paper. However, it is clear that human-eye colour perception is a significant factor and plays a very important role during the resection. Hence it might partly explain why the rate of gross total resection (GTR) using Pp-IX fluorescence varies between studies, e.g. in the randomized controlled trial (RCT) the GTR rate in the study group was 64% [23], while in another RCT the rate of GTR in the study arm was 75% [8], and the GTR was 73.2% in a more recent third study [1]. Therefore constructing and validating a fluorescence scale will go along way in the standardization of fluorescence detection. To our knowledge this is the first fluorescence scale for HGG that had been constructed and validated with an intra-rater and inter-rater agreement calculations. The intra-rater agreement was based on scores from ten participants that were tested twice. Overall the intra-rater agreement was excellent. A possible weakness of this study was the limited number of pre-recorded images
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
4
M. Petterssen et al.
Figure 2
Recorded intraoperative HGG fluorescence from which the scale was constructed.
used for intra-rater correlation. However, this is still the largest sample of shades tested (340 test areas) than in other studies [15,17]. Additionally, the intra-rater analyses were free from significant bias. The inter-rater agreement was estimated on different datasets. The overall results were considered excellent (ICC > 0.75). Scores from the first phase had the lowest ICC value of the datasets (ICC = 0.76). The large number of images and time-consuming validation-tests made in the first phase round could explain the lower ICC value. Even if there were only 10 participants in this phase, the number of images each person had to go through and the time spent on the validation was substantial. This could have lead to the participants getting tired and less attentive the longer the test went on. The second phase, with a larger population sample size and a smaller number of images, had an overall ICC-value of 0.86 with better agreement. The degree of bias in the scale was analyzed in order to find out whether the tests had been properly performed and to what degree the results could be relied upon. If the bias had been too great, the scale would have been unusable despite a seemingly good validity. The limits of tolerable bias were set to ±1 point of the scale. More deviation than this would have probably affected the reliability and the clinical usefulness of the scale. The intra-rater analysis had no bias, but there were significant biases in the inter-rater analysis (as great as 0.98 points in the 56 person dataset of the second phase). Nevertheless, all bias was still within acceptable limits and would have nothing but negligible effects in clinical practice.
The correlation of the scale to fluorescence intensity gave rise to numerous questions. As explained above, the scale has excellent intra- and inter-rater correlation as well as acceptable levels of bias. However, in the violet-blue areas of the scale there were significant disagreements. In practice this should not matter as both violet and blue colour are considered none fluorescent and background colour, which does not need to be resected. The basis of this study has been visual colour analysis of 5-ALA-induced fluorescence, an area that is not very well documented in the literature. However, there were studies regarding colour correlation in other fields of industry. The first area was in optometrics, e.g. the study by Laborie et al. in 2010 [15]. It aimed at determining acceptable colour deviation in computer displays used in cars. Even if the final purpose is quite different from our study, the goal of determining colour appearances and acceptable deviations is quite similar. The researchers also managed to construct a scale to determine colour deviation. Unfortunately, this was a categorical scale (ranging from ‘very unsatisfying correlation’ to ‘very satisfying correlation’) instead of a visual analogue scale and it was not applicable in our experiment. However, this study supports our idea that a scale for colour acceptability actually can be constructed and validated. Another area where colour analysis has been well researched is in dentistry [4,17], to establish and improve the cosmetic appearance of artificial teeth. Since there were several well-documented shade-guides in use [2,19], later methodological studies have instead focused on
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
Impact of human colour perception on Pp-IX fluorescence factors such as the effect of changed illumination [19] and the reliability of colorimetric equipment [22]. C ¸ apa and collaborators [4] published a study in 2010 that analyzed the degree of correlation when observers determined the colour of artificial teeth. Colours were compared to a shade guide, and the population was analyzed to see how many persons chose the correct colour from the scale. The correlations were good among experienced dentists, and less good among students and laymen. As in our study, there was inter-observer agreement regarding shades, even though C ¸ apa detected weaker agreements. The participants in C ¸ apa’s et al. study knew the shades used in the experiment beforehand, which made it easier to differentiate between correct and incorrect answers. In our study, focus lied on the validity of the scale. We had no way to point out answers as correct or incorrect, and deviations of one panel were considered an acceptable level of bias. Nevertheless, C ¸ apa’s study implies that there is a possibility of interobserver agreement in colour perception when using a fixed scale. Our study had highlighted that using a fluorescence scale which can be injected in the head up display of the surgical microscope might be useful to differentiate what is fluoresced and what did not particularly in the red fluorescent zone where fluorescence guided surgical resection should be used. However, our scale was less reliable with significant inter-observer disagreement in violet-blue areas and it was useless in colourblind participants. Therefore it might be a good idea to explore an automated computer based technology to detect the fluorescence. In histopathology, such computer programs exist. In a recent study by Wang et al. [27] aimed to describe the colours of Prussian blue staining as numerical values and develop a high flexibility model for automatic, computer based colour assessment. Furthermore, this study used a colour wheel model to compare the visual colour assessment to the computer-detection system. For comparison, the numerical values of colours have been necessary for our study’s scale construction and an automatic colour detection model has been discussed as a possible future goal. In addition, the use of a colour reference model (a colour wheel and a visual analogue scale, respectively) has been central. Wang et al. found the colours could be quantified as numbers, and the quantification correlated to the visual appearance. This point is important, since the numerical representation of colour values is crucial to the construction of a scale like the one used in our study. Another important point made by Wang et al. is that the use of CIELAB colour space is crucial if a method is to be exported from one machine to another, since RGB-values are dependent on the hardware used [27]. If our scale is to be analyzed further, the RGB-values will need to be recalculated to CIELAB coordinates. Nevertheless, the use of RGB had not affected the validity of our study, because a colour preservation programme was used and images were analyzed by one piece of hardware only. The results of our scale validation in this study suggest that the designed fluorescence scale is effective and potentially useful to surgeons who have normal colour vision in clinical practice. 5-ALA-induced fluorescence has a very high sensitivity (99.6%) and selectivity (81.6%) in glioblastoma tissue, making it an excellent method to distinguish tumour from normal brain [22]. Maximal safe resection of glioblastoma (more than 98% of the enhancing tumour) is very
5 important prognostic factor for tumour free survival and overall survival [1,8,11]. Tumour removal of 98% or higher, extends the life expectancy by at least 4.2 months compared to patients that underwent less extensive resection [11]. If the procedure depended only on the specificity of 5-ALA-induced fluorescence, there would be few problems. However, colour perception varies greatly with the environmental conditions of time and place [4,10,17,19] and with the person observing the colour [15]. The environment in the operating theatre is quite consistent, but there might be big inter- and intra observer differences between different surgeons’ perceptions of colours. This is a normal variation in the population [15], but also a source of error when using fluorescence guided surgery. Colour blindness can also contribute to differences in colour perception, as none of the three colour-blind participants in our cohort was able to detect any fluorescence. The fact remains, the results of all operations for glioblastoma are dependent on the surgeon’s ability to distinguish tumorous tissue from normal tissue, and differences in colour perception are a possible source of error. A more objective analysis of fluorescence using spectroscopy- and optical biopsy systems has been tried. These involve the use of probes to measure the wavelengths emitted from the fluorescent tissue [8,12,20]. However, these methods are not perfect. Light absorption and scattering by the fluorescent tissues may cause errors in measurements [12]. The analytical models used to calculate wavelengths and compensate for disturbances in the measurements may also cause the results to vary depending on the model used [19]. Another point of interest is the effects of video recording and its impact on the results of our study. As our study used recorded images to construct the scale and validate it, image deterioration and distortion was minimized by the stringent quality assurance measures used to calibrate the multichip video camera before each procedure and the frequent calibration of the surgical microscope. Furthermore, though the majority of HGG resections are performed using the microscope directly, endoscopic minimally invasive procedures using an endoscope and video screen are gaining support in many fields and our scale would be of great value during these procedures. Although our visual analogue scale for fluorescence may not measure the exact wavelengths, it would be a valuable supportive tool for surgeons with normal colour vision performing fluorescence guided surgery. It could be used as an easy accessible point of reference, and could possibly be integrated in the equipment used for fluorescence-guided surgery to make the resections as standard and exact as possible.
Conclusions Within the limits of this study, the fluorescence scale has an intra-rater and inter-rater reliability as well as acceptable levels of bias among surgeons with normal colour perception. However, there were high deviation from the median and coefficients of variance at violet-blue end of the scale. The scale and fluorescence guided surgery should not be used by colour-blind surgeons. With further research, development and tests in clinical use, our scale may become a useful tool to improve the results of fluorescence-guided surgery.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
6
M. Petterssen et al.
Acknowledgement Many thanks to Dr. Lynda Cochrane (Division of Clinical and Population Sciences and Education, Dundee University) for her invaluable work with the statistical analysis.
References [1] Aldave G, Tejada S, Pay E, Marigil M, Bejarano B, Idoate MA, et al. Prognostic value of residual fluorescent tissue in glioblastoma patients after gross total resection in 5-aminolevulinic acid-guided surgery. Neurosurgery 2013;72(6):915—21. [2] Analoui M, Papkosta E, Cochran M, Matis B. Designing visually optimal shade guides. J Prosthet Dent 2004;92(4):371—6. [3] Battersby AR, Fookes CJR, Matcham GWJ, McDonald E. Biosynthesis of the pigments of life: formation of the macrocycle. Nature 1980;285:17—21. [4] Capa N, Malkondu O, Kazazoglu E, Calikkocaoglu S. Evaluating factors that affect the shade-matching ability of dentists, dental staff members and laypeople. J Am Dent Assoc 2010;141(1):71—6. [5] Chaichana K, Parker S, Olivi A, Qui˜ nones-Hinojosa A. A proposed classification system that projects outcomes based on preoperative variables for adult patients with glioblastoma multiforme. J Neurosurg 2010;112(5):997—1004. [6] Eljamel MS. Fluorescence image guided surgery of brain tumours: explained step by step. Photodiag Photodyn Ther 2008;5:260—3. [7] Eljamel S. Photodynamic applications in brain tumors: a comprehensive review of the literature. Photodiag Photodyn Ther 2010;7(June (2)):76—85. [8] Eljamel MS, Goodman C, Moseley H. ALA and photofrin(R) fluorescence-guided resection and repetitive PDT in glioblastoma multiforme: a single centre Phase III randomised controlled trial. Lasers Med Sci 2008;23:361—7. [9] Eljamel S, Petersen M, Valentine R, Buist R, Goodman C, Moseley H, et al. Comparison of intraoperative fluorescence and MRI image guided neuronavigation in malignant brain tumors, a prospective controlled study. Photodiag Photodyn Ther 2013, http://dx.doi.org/10.1016/j.pdpdt.2013.03.006. [10] Fairchild MD. Color appearance models and complex visual stimuli. J Dent 2010;38(Suppl. 2):e25—33. [11] Hentschel SJ, Sawaya R. Optimizing outcomes with maximal surgical resection of malignant gliomas. Cancer Control 2003;10(2):109—14. [12] Kim A, Khurana M, Moriyama Y, Wilson BC. Quantification of in vivo fluorescence decoupled from the effects of tissue optical properties using fiber-optic spectroscopy measurements. J Biomed Opt 2010;15(6):067006. [13] Koch, Gary G. Intraclass correlation coefficient. In: Samuel Kotz, Norman LJohnson, editors. Encyclopedia of statistical sciences, Vol. 4. New York: John Wiley & Sons; 1982. p. 213—7.
[14] Kuhnt D, Ganslandt O, Schlaffer SM, Buchfelder M, Nimsky C. Quantification of glioma removal by intraoperative highfield magnetic resonance imaging: an update. Neurosurgery 2011;69(4):852—62. [15] Laborie B, Vienot F, Langlois S. Methodology for constructing a colour-difference acceptability scale. Ophthalmic Physiol Opt 2010;30(5):568—77. [16] Lacroix M, Abi-Said D, Fourney DR, Gokaslan ZL, Shi W, DeMonte F, et al. A multivariate analysis of 416 patients with glioblastoma multiforme: prognosis, extent of resection, and survival. J Neurosurg 2001;95(2):190—8. [17] Lagouvardos PE, Fougia AG, Diamantopoulou SA, Polyzois GL. Repeatability and interdevice reliability of two portable color selection devices in matching and measuring tooth color. J Prosthet Dent 2009;101(1):40—5. [18] Layer G, Reichelt J, Jahn D. Structure and function of enzymes in heme biosynthesis. Protein Sci 2010;19:1137—61. [19] Lee YK, Yu B, Lim JI, Lim HN. Perceived color shift of a shade guide according to the change of illuminant. J Prosthet Dent 2011;105(2):91—9. [20] Muller MG, Georgakoudi I, Zhang Q, Wu J, Feld MS. Intrinsic fluorescence spectroscopy in turbid media: disentangling effects of scattering and absorption. Appl Opt 2001;40(25): 4633—46. [21] Reed JF, Lynn F, Meade BD. Use of coefficient of in assessing variability of quantitative variation assays. Clin Diagn Lab Immunol 2002;9(6):1235—9, http://dx.doi.org/10.1128/CDLI.9.6.1235-1239.2002. [22] Stummer W, Novotny A, Stepp H, Goetz C, Bise K, Reulen HJ. Fluorescence-guided resection of glioblastoma multiforme by using 5-aminolevulinic acid-induced porphyrins: a prospective study in 52 consecutive patients. J Neurosurg 2000;93:1003—13. [23] Stummer W, Pichlmeier U, Meinel T, Wiestler OD, Zanella F, Reulen HJ. Fluorescence-guided surgery with 5-aminolevulinic acid for resection of malignant glioma: a randomised controlled multicentre phase III trial. Lancet Oncol 2006;7:392—401. [24] Stummer W, Reulen HJ, Meinel T, Pichlmeier U, Schumacher W, Tonn JC, et al. Extent of resection and survival in glioblastoma multiforme: identification of and adjustment for bias. Neurosurgery 2008;62(3):564—76. [25] Stummer W, Stepp H, Moller G, Ehrhardt A, Leonhard M, Reulen HJ. Technical principles for protoporphyrin-IX-fluorescence guided microsurgical resection of malignant glioma tissue. Acta Neurochir (Wien) 1998;140:995—1000. [26] Uehlinger P, Zellweger M, Wagnieres G, Juillerat-Jeanneret L, van den Bergh H, Lange N. 5-Aminolevulinic acid and its derivatives: physical chemical properties and protoporphyrin IX formation in cultured cells. J Photochem Photobiol B 2000;54(1):72—80. [27] Wang D, Shi L, Wang YX, Man GC, Heng PA, Griffith JF, et al. Color quantification for evaluation of stained tissues. Cytometry A 2011;79(4):311—6.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
ARTICLE IN PRESS
Photodiagnosis and Photodynamic Therapy (2014) xxx, xxx—xxx
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/pdpdt
Protoporphyrin-IX fluorescence guided surgical resection in high-grade gliomas: The potential impact of human colour perception Max Petterssen b, Sarah Eljamel c, Sam Eljamel MBBCh, MD a,∗ a
Department of Neurosurgery, The University of Dundee, UK1 Linkopings University, Scotland, UK c Lothian Radiology Training Programme, Scotland, UK b
KEYWORDS Colourblind; Colour-perception; Fluorescence-scale; Glioblastoma; Surgical resection
Summary Introduction: Protoporphyrin-IX (Pp-IX) fluorescence had been used frequently in recent years to guide microsurgical resection of high-grade gliomas (HGG), particularly following the publication of a randomized controlled trial demonstrating its advantages. However, Pp-IX fluorescence is dependent upon the surgeons’ eyes’ perception of red fluorescent colour. This study was designed to evaluate human eye fluorescence perception and establish a fluorescence scale. Materials and methods: 20 of 108 pre-recorded images from intraoperative fluorescence of HGG were used to construct an 8-panel visual analogue fluorescence scale. The scale was validated by testing 56 participants with normal colour vision and three red-green colour-blind participants. For intra-rater agreement ten participants were tested twice and for inter-observer reliability the whole cohort were tested. Results: The intra- and inter-observer reliability of the scale in normal colour vision participants was excellent. The scale was less reliable in the violet-blue panels of the scale. Colour-blind participants were not able to distinguish between red fluorescence and blue-violet colours. Conclusion: The 8-panel fluorescence scale is valid in differentiating red, pink and blue colours in a fluorescence surgical field among participants with normal colour perception and potentially useful to standardize fluorescence-guided surgery. However, colourblind surgeons should not use fluorescence-guided surgery. © 2014 Published by Elsevier B.V.
∗ Corresponding author at: Department of Neurosurgery, Ninewells Hospital and Medical School, Dundee DD1 9SY, UK. Tel.: +44 1382 425712; fax: +44 1382 496202. E-mail addresses: [email protected], [email protected] (S. Eljamel). 1 www.neurosurgery-mse.com.
http://dx.doi.org/10.1016/j.pdpdt.2014.05.002 1572-1000/© 2014 Published by Elsevier B.V.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
2
M. Petterssen et al.
Introduction
Scale validation
Protoporphyrin-IX (Pp-IX) is a naturally occurring fluorophore that absorbs blue light at 404 nm and fluoresces red light with a peak at 635 nm. Pp-IX is a byproduct of 5-aminolevulinic acid (5-ALA) metabolism in the mitochondria, where normal cells transform Pp-IX into heme [3,18,26]. Furthermore oral 5-ALA at 20 mg/kg body weight given orally induces Pp-IX-fluorescence in high grade gliomas (HGG) improving the rate of surgical resection, and time to tumour progression [6—9,21,23,25]. A multicenter randomized controlled trial had demonstrated Pp-IX fluorescence improved completeness of HGG resection (64% versus 36%) and progression free survival (41% versus 21%) [23]. Several other studies demonstrated gross total resection (GTR) of HGG enhancing lesion was an important prognostic factor [1,5,14,16,24]. However, Pp-IX fluorescence is dependent upon red-colour perception by the surgeons’ eyes. The aim of this study was to evaluate Pp-IX fluorescence colour perception in a cohort of doctors and establish a fluorescence colour scale to standardize fluorescence detection.
The first phase of the study was a feasibility phase to establish how easy was it to carry out this test. Ten healthcare professionals participated in this phase of the study, seven males and 3 females. Each individual passed a colour vision test before participating in the study. 75 intraoperative fluorescence images were selected for this phase. The images depicted variable degrees of fluorescence. Each participant viewed each image separately on Mac computer (MacBook Pro 7.1, Mac OSX 10.6.7, with a 13.3 in. screen set for 1200 × 800 pixels resolution and Colour-LCD profile). The Mac computer display screen was split horizontally to display the scale shown in Fig. 1 without the RGB-colour percentages at the top of the screen and the intraoperative fluorescence image at the bottom of the Mac screen. A total of 465 image areas in 75 images demonstrating variable degrees of fluorescence and the background colours were circled in the images. Each participant was asked to compare the shades of the 465 selected areas to the scale and decide which panel (1—8) had the best correlation to colour and intensity. The participants were able to move their heads freely during these experiments. They were also allowed to take breaks whenever they wanted. There was no time limit to complete the experiments. All tests were performed with one person at a time, to prevent peer influence. A second phase was performed one week after the end of the first phase. Because of the length of time it took to complete the first phase, a smaller set of 10 images with 34 selected areas was used. The whole spectrum of fluorescence was still represented in these images. The order of the pictures was randomly rearranged; the tests were otherwise performed in the same way as in the first phase. In this phase fifty-nine participants with normal colour vision from the healthcare staff were involved (22 males and 37 females) and three additional participants with red-green colour blindness were tested. The 59 participants included the ten participants from the first phase. All participants were tested for colour blindness prior to the tests and only three out of 59 were red-green colour-blind.
Materials and methods Intraoperative Pp-IX fluorescence image database was created from 2009 using recorded images obtained from a Pentera surgical microscope calibrated to detect HGG fluorescence (Carl Zeiss OPMI® PenteroTM , Germany). This surgical microscope was designed to switch between white and blue light at the press of a button, the blue light range used was 373—440 nm. The reflected/emitted light was subjected to a longpass filter to maximize red light detection but allowing sufficient background light back to the surgeon to be able to continue operating under the blue light. The images were also recorded simultaneously on the hard disc of the microscope by the integrated multichip-digital video camera. To prevent distortion of the colours during recording and between cases, the video camera was calibrated before each procedure by using a quality assurance phantom provided by the vender (Zeiss) as part of the system to make sure that the recorded images were of the same intensity, quality and colours. To prevent image deterioration over time the microscope as a unit was subjected to regular quality assurance testing by the vender (Zeiss) to ensure light intensity and all components were in good working order. Initially a random sample of 108 images was used in this study, from which 20 images were selected to construct a fluorescence detection scale. The selection of these images to construct the scale was based on the fact these images were images where all shades of fluorescence could be detected in the same image by two observers (MP&SE). From these 20 images, two hundred and thirty-three colour samples were extracted. The colour samples were arranged in an indexed database. An image-handling program, GiMP 2.6 for Mac (ICC profile: sRGB IEC61966-2.1), combined with the Mac Color sync tool 4.6.2 (for colour preservation) was used. The samples were analyzed for content of red, blue and green expressed as RGB % using Colormeter.app 3.7.2. Then the colour samples were arranged based on their colour content in order to construct the fluorescence scale.
Statistical analysis Because it is likely that only one surgeon will make the decision regarding fluorescence during surgery, differences between colour perception and agreement were evaluated by the test and retest correlation in the 10 participants (Intra-rater agreement), which was analyzed using intraclass correlation (ICC). The two-way random model for absolute agreement was chosen because the images and raters were samples from different populations and the agreement between actual scale values is of interest in this study. The ICC is a descriptive statistic that can be used when quantitative measurements are made on units that are organized into groups. It describes how strongly units in the same group resemble each other. While it is viewed as a type of correlation, unlike most other correlation measures it operates on data structured as groups, rather than data structured as paired observations [13]. To assess agreement between different participants (ICC) of the fluorescence intensity,
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
Impact of human colour perception on Pp-IX fluorescence
Figure 1
Fluorescence visual analogue scale (1—8).
single measures of ICC were used. The inter-rater agreement was calculated with ICC of the whole database including the first 10 participants’ results. Inter-rater bias was analyzed on both the populations (10 and 59 participants) from the first and second phases, using repeated measures analysis of variance. Agreement between participants was measured by calculating a coefficient of variation (CV), where the lower the CV the more agreement and the higher the CV the less agreement between participants. The coefficient of variation (CV) is defined as the ratio of the standard deviation to the mean : Cv =
3
It shows the extent of variability in relation to mean of the population [21]. Values of ICC were interpreted as follows: ≥0.75 excellent, 0.4—0.75 fair to good and 0.13) and less likely to agree on violet-blue colour perception (p < 0.05). The three participants who were red-green colour-blind were unable to detect any fluorescence including the very strong red fluorescence.
Discussion The importance of gross total resection (GTR) of HGG and the use of 5-ALA induced fluorescence to achieve this goal had been highlighted in the introduction of this paper. However, it is clear that human-eye colour perception is a significant factor and plays a very important role during the resection. Hence it might partly explain why the rate of gross total resection (GTR) using Pp-IX fluorescence varies between studies, e.g. in the randomized controlled trial (RCT) the GTR rate in the study group was 64% [23], while in another RCT the rate of GTR in the study arm was 75% [8], and the GTR was 73.2% in a more recent third study [1]. Therefore constructing and validating a fluorescence scale will go along way in the standardization of fluorescence detection. To our knowledge this is the first fluorescence scale for HGG that had been constructed and validated with an intra-rater and inter-rater agreement calculations. The intra-rater agreement was based on scores from ten participants that were tested twice. Overall the intra-rater agreement was excellent. A possible weakness of this study was the limited number of pre-recorded images
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
4
M. Petterssen et al.
Figure 2
Recorded intraoperative HGG fluorescence from which the scale was constructed.
used for intra-rater correlation. However, this is still the largest sample of shades tested (340 test areas) than in other studies [15,17]. Additionally, the intra-rater analyses were free from significant bias. The inter-rater agreement was estimated on different datasets. The overall results were considered excellent (ICC > 0.75). Scores from the first phase had the lowest ICC value of the datasets (ICC = 0.76). The large number of images and time-consuming validation-tests made in the first phase round could explain the lower ICC value. Even if there were only 10 participants in this phase, the number of images each person had to go through and the time spent on the validation was substantial. This could have lead to the participants getting tired and less attentive the longer the test went on. The second phase, with a larger population sample size and a smaller number of images, had an overall ICC-value of 0.86 with better agreement. The degree of bias in the scale was analyzed in order to find out whether the tests had been properly performed and to what degree the results could be relied upon. If the bias had been too great, the scale would have been unusable despite a seemingly good validity. The limits of tolerable bias were set to ±1 point of the scale. More deviation than this would have probably affected the reliability and the clinical usefulness of the scale. The intra-rater analysis had no bias, but there were significant biases in the inter-rater analysis (as great as 0.98 points in the 56 person dataset of the second phase). Nevertheless, all bias was still within acceptable limits and would have nothing but negligible effects in clinical practice.
The correlation of the scale to fluorescence intensity gave rise to numerous questions. As explained above, the scale has excellent intra- and inter-rater correlation as well as acceptable levels of bias. However, in the violet-blue areas of the scale there were significant disagreements. In practice this should not matter as both violet and blue colour are considered none fluorescent and background colour, which does not need to be resected. The basis of this study has been visual colour analysis of 5-ALA-induced fluorescence, an area that is not very well documented in the literature. However, there were studies regarding colour correlation in other fields of industry. The first area was in optometrics, e.g. the study by Laborie et al. in 2010 [15]. It aimed at determining acceptable colour deviation in computer displays used in cars. Even if the final purpose is quite different from our study, the goal of determining colour appearances and acceptable deviations is quite similar. The researchers also managed to construct a scale to determine colour deviation. Unfortunately, this was a categorical scale (ranging from ‘very unsatisfying correlation’ to ‘very satisfying correlation’) instead of a visual analogue scale and it was not applicable in our experiment. However, this study supports our idea that a scale for colour acceptability actually can be constructed and validated. Another area where colour analysis has been well researched is in dentistry [4,17], to establish and improve the cosmetic appearance of artificial teeth. Since there were several well-documented shade-guides in use [2,19], later methodological studies have instead focused on
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
Impact of human colour perception on Pp-IX fluorescence factors such as the effect of changed illumination [19] and the reliability of colorimetric equipment [22]. C ¸ apa and collaborators [4] published a study in 2010 that analyzed the degree of correlation when observers determined the colour of artificial teeth. Colours were compared to a shade guide, and the population was analyzed to see how many persons chose the correct colour from the scale. The correlations were good among experienced dentists, and less good among students and laymen. As in our study, there was inter-observer agreement regarding shades, even though C ¸ apa detected weaker agreements. The participants in C ¸ apa’s et al. study knew the shades used in the experiment beforehand, which made it easier to differentiate between correct and incorrect answers. In our study, focus lied on the validity of the scale. We had no way to point out answers as correct or incorrect, and deviations of one panel were considered an acceptable level of bias. Nevertheless, C ¸ apa’s study implies that there is a possibility of interobserver agreement in colour perception when using a fixed scale. Our study had highlighted that using a fluorescence scale which can be injected in the head up display of the surgical microscope might be useful to differentiate what is fluoresced and what did not particularly in the red fluorescent zone where fluorescence guided surgical resection should be used. However, our scale was less reliable with significant inter-observer disagreement in violet-blue areas and it was useless in colourblind participants. Therefore it might be a good idea to explore an automated computer based technology to detect the fluorescence. In histopathology, such computer programs exist. In a recent study by Wang et al. [27] aimed to describe the colours of Prussian blue staining as numerical values and develop a high flexibility model for automatic, computer based colour assessment. Furthermore, this study used a colour wheel model to compare the visual colour assessment to the computer-detection system. For comparison, the numerical values of colours have been necessary for our study’s scale construction and an automatic colour detection model has been discussed as a possible future goal. In addition, the use of a colour reference model (a colour wheel and a visual analogue scale, respectively) has been central. Wang et al. found the colours could be quantified as numbers, and the quantification correlated to the visual appearance. This point is important, since the numerical representation of colour values is crucial to the construction of a scale like the one used in our study. Another important point made by Wang et al. is that the use of CIELAB colour space is crucial if a method is to be exported from one machine to another, since RGB-values are dependent on the hardware used [27]. If our scale is to be analyzed further, the RGB-values will need to be recalculated to CIELAB coordinates. Nevertheless, the use of RGB had not affected the validity of our study, because a colour preservation programme was used and images were analyzed by one piece of hardware only. The results of our scale validation in this study suggest that the designed fluorescence scale is effective and potentially useful to surgeons who have normal colour vision in clinical practice. 5-ALA-induced fluorescence has a very high sensitivity (99.6%) and selectivity (81.6%) in glioblastoma tissue, making it an excellent method to distinguish tumour from normal brain [22]. Maximal safe resection of glioblastoma (more than 98% of the enhancing tumour) is very
5 important prognostic factor for tumour free survival and overall survival [1,8,11]. Tumour removal of 98% or higher, extends the life expectancy by at least 4.2 months compared to patients that underwent less extensive resection [11]. If the procedure depended only on the specificity of 5-ALA-induced fluorescence, there would be few problems. However, colour perception varies greatly with the environmental conditions of time and place [4,10,17,19] and with the person observing the colour [15]. The environment in the operating theatre is quite consistent, but there might be big inter- and intra observer differences between different surgeons’ perceptions of colours. This is a normal variation in the population [15], but also a source of error when using fluorescence guided surgery. Colour blindness can also contribute to differences in colour perception, as none of the three colour-blind participants in our cohort was able to detect any fluorescence. The fact remains, the results of all operations for glioblastoma are dependent on the surgeon’s ability to distinguish tumorous tissue from normal tissue, and differences in colour perception are a possible source of error. A more objective analysis of fluorescence using spectroscopy- and optical biopsy systems has been tried. These involve the use of probes to measure the wavelengths emitted from the fluorescent tissue [8,12,20]. However, these methods are not perfect. Light absorption and scattering by the fluorescent tissues may cause errors in measurements [12]. The analytical models used to calculate wavelengths and compensate for disturbances in the measurements may also cause the results to vary depending on the model used [19]. Another point of interest is the effects of video recording and its impact on the results of our study. As our study used recorded images to construct the scale and validate it, image deterioration and distortion was minimized by the stringent quality assurance measures used to calibrate the multichip video camera before each procedure and the frequent calibration of the surgical microscope. Furthermore, though the majority of HGG resections are performed using the microscope directly, endoscopic minimally invasive procedures using an endoscope and video screen are gaining support in many fields and our scale would be of great value during these procedures. Although our visual analogue scale for fluorescence may not measure the exact wavelengths, it would be a valuable supportive tool for surgeons with normal colour vision performing fluorescence guided surgery. It could be used as an easy accessible point of reference, and could possibly be integrated in the equipment used for fluorescence-guided surgery to make the resections as standard and exact as possible.
Conclusions Within the limits of this study, the fluorescence scale has an intra-rater and inter-rater reliability as well as acceptable levels of bias among surgeons with normal colour perception. However, there were high deviation from the median and coefficients of variance at violet-blue end of the scale. The scale and fluorescence guided surgery should not be used by colour-blind surgeons. With further research, development and tests in clinical use, our scale may become a useful tool to improve the results of fluorescence-guided surgery.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
+Model PDPDT-565; No. of Pages 6
ARTICLE IN PRESS
6
M. Petterssen et al.
Acknowledgement Many thanks to Dr. Lynda Cochrane (Division of Clinical and Population Sciences and Education, Dundee University) for her invaluable work with the statistical analysis.
References [1] Aldave G, Tejada S, Pay E, Marigil M, Bejarano B, Idoate MA, et al. Prognostic value of residual fluorescent tissue in glioblastoma patients after gross total resection in 5-aminolevulinic acid-guided surgery. Neurosurgery 2013;72(6):915—21. [2] Analoui M, Papkosta E, Cochran M, Matis B. Designing visually optimal shade guides. J Prosthet Dent 2004;92(4):371—6. [3] Battersby AR, Fookes CJR, Matcham GWJ, McDonald E. Biosynthesis of the pigments of life: formation of the macrocycle. Nature 1980;285:17—21. [4] Capa N, Malkondu O, Kazazoglu E, Calikkocaoglu S. Evaluating factors that affect the shade-matching ability of dentists, dental staff members and laypeople. J Am Dent Assoc 2010;141(1):71—6. [5] Chaichana K, Parker S, Olivi A, Qui˜ nones-Hinojosa A. A proposed classification system that projects outcomes based on preoperative variables for adult patients with glioblastoma multiforme. J Neurosurg 2010;112(5):997—1004. [6] Eljamel MS. Fluorescence image guided surgery of brain tumours: explained step by step. Photodiag Photodyn Ther 2008;5:260—3. [7] Eljamel S. Photodynamic applications in brain tumors: a comprehensive review of the literature. Photodiag Photodyn Ther 2010;7(June (2)):76—85. [8] Eljamel MS, Goodman C, Moseley H. ALA and photofrin(R) fluorescence-guided resection and repetitive PDT in glioblastoma multiforme: a single centre Phase III randomised controlled trial. Lasers Med Sci 2008;23:361—7. [9] Eljamel S, Petersen M, Valentine R, Buist R, Goodman C, Moseley H, et al. Comparison of intraoperative fluorescence and MRI image guided neuronavigation in malignant brain tumors, a prospective controlled study. Photodiag Photodyn Ther 2013, http://dx.doi.org/10.1016/j.pdpdt.2013.03.006. [10] Fairchild MD. Color appearance models and complex visual stimuli. J Dent 2010;38(Suppl. 2):e25—33. [11] Hentschel SJ, Sawaya R. Optimizing outcomes with maximal surgical resection of malignant gliomas. Cancer Control 2003;10(2):109—14. [12] Kim A, Khurana M, Moriyama Y, Wilson BC. Quantification of in vivo fluorescence decoupled from the effects of tissue optical properties using fiber-optic spectroscopy measurements. J Biomed Opt 2010;15(6):067006. [13] Koch, Gary G. Intraclass correlation coefficient. In: Samuel Kotz, Norman LJohnson, editors. Encyclopedia of statistical sciences, Vol. 4. New York: John Wiley & Sons; 1982. p. 213—7.
[14] Kuhnt D, Ganslandt O, Schlaffer SM, Buchfelder M, Nimsky C. Quantification of glioma removal by intraoperative highfield magnetic resonance imaging: an update. Neurosurgery 2011;69(4):852—62. [15] Laborie B, Vienot F, Langlois S. Methodology for constructing a colour-difference acceptability scale. Ophthalmic Physiol Opt 2010;30(5):568—77. [16] Lacroix M, Abi-Said D, Fourney DR, Gokaslan ZL, Shi W, DeMonte F, et al. A multivariate analysis of 416 patients with glioblastoma multiforme: prognosis, extent of resection, and survival. J Neurosurg 2001;95(2):190—8. [17] Lagouvardos PE, Fougia AG, Diamantopoulou SA, Polyzois GL. Repeatability and interdevice reliability of two portable color selection devices in matching and measuring tooth color. J Prosthet Dent 2009;101(1):40—5. [18] Layer G, Reichelt J, Jahn D. Structure and function of enzymes in heme biosynthesis. Protein Sci 2010;19:1137—61. [19] Lee YK, Yu B, Lim JI, Lim HN. Perceived color shift of a shade guide according to the change of illuminant. J Prosthet Dent 2011;105(2):91—9. [20] Muller MG, Georgakoudi I, Zhang Q, Wu J, Feld MS. Intrinsic fluorescence spectroscopy in turbid media: disentangling effects of scattering and absorption. Appl Opt 2001;40(25): 4633—46. [21] Reed JF, Lynn F, Meade BD. Use of coefficient of in assessing variability of quantitative variation assays. Clin Diagn Lab Immunol 2002;9(6):1235—9, http://dx.doi.org/10.1128/CDLI.9.6.1235-1239.2002. [22] Stummer W, Novotny A, Stepp H, Goetz C, Bise K, Reulen HJ. Fluorescence-guided resection of glioblastoma multiforme by using 5-aminolevulinic acid-induced porphyrins: a prospective study in 52 consecutive patients. J Neurosurg 2000;93:1003—13. [23] Stummer W, Pichlmeier U, Meinel T, Wiestler OD, Zanella F, Reulen HJ. Fluorescence-guided surgery with 5-aminolevulinic acid for resection of malignant glioma: a randomised controlled multicentre phase III trial. Lancet Oncol 2006;7:392—401. [24] Stummer W, Reulen HJ, Meinel T, Pichlmeier U, Schumacher W, Tonn JC, et al. Extent of resection and survival in glioblastoma multiforme: identification of and adjustment for bias. Neurosurgery 2008;62(3):564—76. [25] Stummer W, Stepp H, Moller G, Ehrhardt A, Leonhard M, Reulen HJ. Technical principles for protoporphyrin-IX-fluorescence guided microsurgical resection of malignant glioma tissue. Acta Neurochir (Wien) 1998;140:995—1000. [26] Uehlinger P, Zellweger M, Wagnieres G, Juillerat-Jeanneret L, van den Bergh H, Lange N. 5-Aminolevulinic acid and its derivatives: physical chemical properties and protoporphyrin IX formation in cultured cells. J Photochem Photobiol B 2000;54(1):72—80. [27] Wang D, Shi L, Wang YX, Man GC, Heng PA, Griffith JF, et al. Color quantification for evaluation of stained tissues. Cytometry A 2011;79(4):311—6.
Please cite this article in press as: Petterssen M, et al. Protoporphyrin-IX fluorescence guided surgical resection in highgrade gliomas: The potential impact of human colour perception. Photodiagnosis and Photodynamic Therapy (2014), http://dx.doi.org/10.1016/j.pdpdt.2014.05.002
