Virchows Arch DOI 10.1007/s00428-010-1008-3
ORIGINAL ARTICLE
JPEG2000 for automated quantification of immunohistochemically stained cell nuclei: a comparative study with standard JPEG format Marylène Lejeune & Carlos López & Ramón Bosch & Anna Korzyńska & Maria-Teresa Salvadó & Marcial García-Rojo & Urszula Neuman & Łukasz Witkowski & Jordi Baucells & Joaquín Jaén
Received: 13 August 2010 / Revised: 14 October 2010 / Accepted: 2 November 2010 # Springer-Verlag 2010
Abstract The Joint Photographic Experts Group (JPEG) standard format is one of the most widely used in image compression technologies. More recently, JPEG2000 format has emerged as a state-of-the-art technology that provides substantial improvements in picture quality at higher compression ratios. However, there has been no attempt to date to determine which of the two compression formats produces less variability in the automated evaluation of immunohistochemically stained digital images in agreement with their compression rates and complexity Marylène Lejeune and Carlos López have contributed equally to this work. M. Lejeune (*) : C. López Molecular Biology and Research Section, Hospital de Tortosa Verge de la Cinta, IISPV, URV, c/Esplanetes 14, 43500, Tortosa, Spain e-mail:
[email protected] R. Bosch : M.-T. Salvadó : J. Jaén Department of Pathology, Hospital de Tortosa Verge de la Cinta, IISPV, URV, Tortosa, Spain
degrees. The evaluation of Ki67 and FOXP3 immunohistochemical nuclear markers was performed in a total of 329 digital images: 47 were captured in uncompressed Tagged Image File Format (TIFF), 141 were converted to three JPEG compressed formats (47 each with 1:3, 1:23 and 1:46 compression) and 141 were converted to three JPEG2000 compressed formats (47 each with 1:3, 1:23 and 1:46 compression). The count differences between images in TIFF versus JPEG formats were compared with those obtained between images in TIFF versus JPEG2000 formats at the three levels of compression. It was found that, using JPEG2000 compression, the results of the stained nuclei count are close enough to the results obtained with uncompressed images, especially in highly complex images at minimum and medium compression. Otherwise, in images of low complexity, JPEG and JPEG2000 had similar count efficiency to that of the original TIFF images at all compression levels. These data suggest that JPEG2000 could give rise to an efficient means of storage, reducing file size and storage capacity, without compromise on the immunohistochemical analytical quality.
A. Korzyńska : U. Neuman : Ł. Witkowski
Keywords Automated image analysis . immunohistochemistry . JPEG . JPEG2000
Laboratory of Processing Systems of Microscopic Image Information, Nalecz Institute of Biocybernetics and Biomedical Engineering, Polish Academy of Sciences, Warsaw, Poland
Introduction
M. García-Rojo Department of Pathology, Hospital General de Ciudad Real, Ciudad Real, Spain J. Baucells Department of Informatics, Hospital de Tortosa Verge de la Cinta, IISPV, URV, Tortosa, Spain
Cell quantification and morphological measurement in the field of pathology have several clinical and research applications [1–7]. Non-automated immunohistochemical quantification has an inherent variability that produces various results, leading to intra- and inter-observer variability [8–11]. Intra-observer variability is defined as the
Virchows Arch
differences between repeated evaluations made by the same observer, and inter-observer variability arises from those between evaluations made by different observers. The use of digital image computer analysis to quantify and measure the different morphological parameters of cells or tissues reduces these variabilities [12]. Image quality is one of the most important features in digital image analysis and several factors, such as tissue processing, illumination and camera quality, are very important in ensuring good-definition images. The large number of digital images created in scientific studies and current clinical practice require a suitable image capture format since high-resolution images stored without compression needs very-high-volume devices. Image compression has been used in several fields of medicine to reduce the file size and data transmission time [13–17]. Although compression improves the ability to handle large numbers of photographs, the use of different compression techniques could produce variations in digital images that affect their medical interpretation. There are two types of image compression: lossless and lossy. Lossless compression enables the original file to be reconstructed without loss of data, but with the disadvantage of poor file-size reduction. Lossy compression allows greater file size reduction but with the disadvantage that some details may be lost. Two of the most widely used lossy formats in medicine are JPEG and JPEG2000 compression. The first uses the discrete cosine transform (DCT) technique and the latter the discrete wavelet transform (DWT) method. Despite all the advantages of JPEG compression schemes based on DCT—simplicity, satisfactory performance for human eyes and availability of special-purpose hardware for implementation—shortcomings exist. Since the input image needs to be “blocked”, correlation across the block boundaries is not eliminated, causing noticeable artefacts well visible in high compression rates. Wavelet-based coding has emerged as a state-of-the-art technology within the field of image compression and appears to be more robust than other compression techniques with respect to transmission and decoding errors as well as facilitates the progressive transmission of images. Different studies that have compared the DCT technique to wavelet transformation in different kinds of radiological images with different compression ratios provided contradictory results [18–21]. In immunohistochemical digital images, there is no agreement about whether JPEG2000 is better than JPEG or if differences are indistinguishable at lower or higher compression ratios with the two formats. Most of the studies in the field of pathology that use digital images of immunohistochemically stained tissue sections stored these in the uncompressed TIFF format [12, 22], but some studies have also employed JPEG compression [5, 23]. Our group has developed an automated
computer-assisted method to quantify digital images with immunohistochemically stained nuclear markers that have been tested and validated resting digital image stored in TIFF format [12]. When this method was applied to images after JPEG compression, we found a distortion in the images that reduces the nuclear quantification efficiency with an increase in the compression rates and image complexity degrees [24]. To the best of our knowledge, no previous studies have evaluated the effects of JPEG2000 compression in computer-assisted immunohistochemical nuclear quantification. The method developed in our laboratory is based on colour and morphological segmentation methods and used to compare the results of images stored in TIFF and JPEG formats. The aim of this study is to evaluate the differences in images stored in TIFF and JPEG2000 formats using our method of automated quantification of stained nuclei. We also compare the count differences between JPEG and JPEG2000 with the same or similar degree of compression in order to determine which of the two techniques of compression is better for image storage taking into account the fidelity of further analysis.
Materials and methods Sample selection and immunohistochemistry The samples used in this study were selected from the archives of the Department of Pathology of the Hospital de Tortosa Verge de la Cinta, Catalonia, Spain. Sample pretreatment and staining procedures were standardized as previously described [25]. Briefly, 3-μm-thick sections were dried, deparaffinized in xylene, rehydrated in graded ethanol and washed in water and PBS. Staining was performed with monoclonal antibodies directed against the nuclear proteins Ki67 (clone MIB-1, Dako, Carpinteria, CA, USA) and FOXP3 (clone FOXP3-236A/E7, CNIO, Spain) using standard protocols as previously described [25]. This study receives an institutional review board approval. Image capture and file compression The photographs of each sample were taken with a Leica DFC320 digital camera at a resolution of 3.3 Mpixels CCD with a Bayer Array RGB filter for brilliant pictures and with exposure times from 320 μs to 20 s. The camera was connected to a Leica DM LB2 microscope (Leica Microsystems Wetzlar GmbH, Wetzlar, Germany) with ×40 magnification and 0.65 numeric aperture under the same range of illumination. The camera was controlled with the Leica IM50 v.4.0 software installed on a Compaq Professional Workstation computer (3.6 GHz Pentium IV CPU,
Virchows Arch
1.00 GB RAM). Forty-seven images marked with both immunohistochemical markers were saved as TIFF files of around 9.27 Mbytes. These images were selected according to the variety in the concentration and distribution of the stained nucleus. All images were converted to JPEG/JFIF (JPEG File Interchange Format), a format with the ACDSee 9.0 program that uses an adjustable compression factor, subsampling (colour compression scheme) and Huffman code optimization processes. The adjustable compression factor uses a 0–100 scale of compression (high quality/low compression—low quality/high compression). The ratio of the “chroma subsampling” was adjusted automatically at the standard of 4:2:2 (or 2:1 horizontal) for all compression rates. At the final process, TIFF format images were converted to JPEG images with the lowest compression and the highest quality that the software allows (1:3 compression, file size approximately 3.25 Mbytes), intermediate-quality images (1:23 compression, 400 kbytes) and images with maximum compression rate and poor quality (1:46 compression, 200 kbytes) [24]. We used the same software to convert the original TIFF files to JPEG2000 with the same degrees of compression as the JPEG files and similar file sizes. Nuclear quantification All of the images were quantified by a method developed in our centre and that had been previously tested and validated [12]. In this method, the images are processed in two steps. In the first step, the contrast of the image is modified in order to reduce the background and minimize variations in staining intensity. A colour and iterative morphological segmentation is applied to these modified images to select the positive-stained objects and all the information associated to these objects such as their number, the area and the roundness values of each of them. The second step of the method involves data management of all the extracted information using several algorithms in order to obtain the final number of nuclei in each image, including those of the clusters. We used this automatic procedure to analyse 329 images, 47 from original uncompressed TIFF digital images, 141 from JPEG compressed images (47 each with 1:3, 1:23 and 1:46 JPEG compression) and 141 from JPEG2000 compressed images (47 each with 1:3, 1:23 and 1:46 JPEG2000 compression). Statistical analysis Statistical analysis was carried out with SPSS 11.0. Bland– Altman and Kaplan–Meier analyses, with their corresponding graphical output, were used to determine the nuclear count differences between the original TIFF images and both types of compressed images and to
measure in order to identify which of the two compression methods produced less variability in comparison to the results obtained with TIFF-format images. Bland–Altman graphs illustrate the difference of the TIFF/JPEG and TIFF/ JPEG2000 count results obtained with respect to the mean of each paired count. Kaplan–Meier curves portray the conditional probability of observing count differences between the results obtained with the TIFF and compressed methods. The paired results were obtained for TIFF images versus each group of compressed JPEG images with 1:3, 1:23 and 1:46 compression and TIFF images versus each group of compressed JPEG2000 images with 1:3, 1:23 and 1:46 compression. Furthermore, nuclear count differences between JPEG and JPEG2000 were compared for the same compression degree.
Results Nuclear counts based on images stored in TIFF formats are taken as reference in comparison with those obtained in different compression rates of JPEG and JPEG2000 image formats. Overall, Fig. 1 shows that the differences in nuclear counts between TIFF and the two compressed formats were lowest in JPEG2000 at the low (1:3) and medium (1:23) compression levels. This indicates that, at these compression rates, the JPEG2000 compression method does not damage the objects and the results of stained nuclei counts are more accurate than for the images compressed in JPEG format. The differences in the number of stained nuclei increase in both compressed formats as the compression level rises. The count differences between TIFF and JPEG 1:46 compression and between TIFF and JPEG2000 1:46 were very similar, indicating that the two methods had similar count variability at maximum compression. As previously demonstrated, image complexity, relative to the number of positive-stained nuclei and the presence of clusters, may affect the automated nuclear quantification and lead to count variability in images compressed in JPEG format [12]. In the present study, the images were divided as before into a low-complexity group (Fig. 2a, b; images with fewer than 100 nuclei, no clusters or small-area clusters of less than 25 μm2) and a high-complexity group (Fig. 2c, d; images with more than 100 nuclei or large clusters of more than 25 μm2). The current results obtained with low-complexity images showed that the count differences in TIFF vs. JPEG compressed images and the count differences in TIFF vs. JPEG2000 compressed images were very similar at all compression levels. Figure 3 shows these small differences obtained in images with small-area clusters and Fig. 4 shows the results obtained from images of samples with fewer than 100 nuclei.
Virchows Arch
Fig. 1 Overall results obtained by automatic quantification at different compression degrees. Kaplan–Meier curves (upper row) compare the probability of difference between TIFF/JPEG (black) and
TIFF/JPEG2000 compression (grey). Bland–Altman graphs (lower row) compare count differences between TIFF/JPEG (black) and TIFF/JPEG2000 compression (grey)
In high-complexity images, the differences in the number of stained nuclei quantified in large-area cluster JPEG2000 images are lower than in corresponding JPEG images (Fig. 5). These differences increase along with the compression level, and at maximum compression the two techniques yield similar count differences. Similar results were obtained from images with more than 100 nuclei, showing that the estimation of the number of stained nuclei in these images is also more accurate in the JPEG2000 format (Fig. 6). However, in all kinds of high-complexity images, maximum level of JPEG2000 compression (1:46) damages the objects so that a similar count variability with JPEG can be observed.
lymphoproliferative disorders, breast cancer and other carcinomas [26–31]. The staining pattern of proliferating cells has been also demonstrated to be important for the determination of the nuances and pitfalls that may exist in the interpretation of the Ki67 pattern [26]. The marker of forkhead box protein 3 (FOXP3+) regulatory T cells has been extensively used for research purpose. The presence and the pattern of distribution of these cells, identified by immunohistochemistry, have been demonstrated to be associated with the clinical outcome of patients with different neoplasms. These regulatory T cells were able to reduce tumour-driven immune suppression with clinical benefit in haematological malignancies like follicular and Hodgkin’s lymphomas [6, 32–35], whereas in epithelial carcinomas these regulatory T cells correlated inversely with the clinical outcome [36]. Moreover, it has been demonstrated that the presence of these regulatory T cells, as an element of the microenvironment of some lymphomas, predicts these patients’ response to the treatment [37]. The choice of an adequate file format in digital image analysis is fundamental in medicine for correct analysis and diagnostic evaluations [19]. The performances of JPEG and JPEG2000 formats have been extensively evaluated in radiological images using subjective and objective metrics. Although some papers mentioned the benefits of JPEG2000 [13, 20, 38–42], others said that the results are indistin-
Discussion For this study, the choice of the two antibodies came from the importance of their accurate immunohistochemical evaluation and quantification. Ki67 is a nuclear marker used for assessing the proliferative index in a neoplastic cell population (all stages in the cell cycle except G0). The establishment of the growth fraction of neoplasms by quantitatively identifying Ki67-positive cells by immunohistochemistry is useful for the diagnosis (differential or not) of a great number of neoplastic diseases such as
Virchows Arch Fig. 2 Representative digital images positively grouped in function of nuclei density complexity. The first column (a and c) corresponds to FOXP3 marker in lymph node tissues and the second column (b and d) corresponds to Ki67 marker in breast cancer tissues. a Immunostained images classified in the lowcomplexity group with fewer than 100 nuclei and no clusters. b Immunostained images classified in the low-complexity group with fewer than 100 nuclei and small-area clusters of less than 25 μm2. c Immunostained images classified in the high-complexity group with more than 100 nuclei and smallarea clusters of less than 25 μm2. d Immunostained images classified in the highcomplexity group with more than 100 nuclei and large-area clusters larger than 25 μm2
guishable [21, 43]. Some papers even reported that JPEG is superior to JPEG2000 for medical images using “mean opinion score”, especially in low-compression-ratio areas [18, 44, 45]. Nevertheless, such studies are difficult to compare because different algorithms, image types and evaluation paradigms have been used. In the field of cell quantification in digital pathology, most of the studies use images in an uncompressed TIFF format [10, 22, 46, 47]. This format preserves all photographic information without loss of image quality. Nevertheless, it is clear that the need to reduce the size of files and especially the storage space for these images has stimulated the requirement for different formats of compression. The effects and changes produced by the compression in automated nuclear quantification in digital images have rarely been studied in Feulgen stain [48] or in immunohistochemically stained nuclei [24]. As previously observed, the image complexity and the level of JPEG image compression influenced the automated nuclear quantification results [24, 48] whereby count variability increases as the compression level rises. The present study shows that the JPEG2000 compression method, at low (1:3) and medium (1:23) compression levels, does not disturb the information about immunohistochemically stained nuclei in comparison to the JPEG
compression method. Although the JPEG2000 1:3 and JPEG2000 1:23 count differences are less than in JPEG 1:3 and JPEG 1:23, respectively, the variability is similar for both formats at maximum compression. The effectiveness of JPEG2000 compression at minimum and medium compression levels is especially visible in highly complex images. Images with lower complexity could be analysed with uncompressed TIFF or JPEG and JPEG2000 formats without loss of accuracy. The distortion observed in highly complex images (large-area clusters or more than 100 nuclei) appears to be due to the nearness neighbourhood of the cells and the fact that the boundaries of the cells become blurred during JPEG compression, causing difficulties in cell edge discrimination. Under these conditions, positive pixels that are altered by the compression may fall outside the range of previously selected positive colour values and some negative pixels may fall within the positive range of values [24]. This variability would reduce the efficiency of the segmentation algorithms used in the computer-assisted image analysis procedure, leading to serious and unacceptable inaccuracy in the final nuclear count. The compression gains of JPEG2000 over JPEG are attributed to the use of DWT and a more sophisticated entropy-encoding scheme such as the decomposition of the image into a multiple-
Virchows Arch
Fig. 3 Results obtained by automatic quantification in images with small-area clusters at different compression levels. Kaplan–Meier curves (upper row) compare the probability of difference between
TIFF/JPEG (black) and TIFF/JPEG2000 compression (grey). Bland– Altman graphs (lower row) compare count differences between TIFF/ JPEG (black) and TIFF/JPEG2000 compression (grey)
Fig. 4 Results obtained by automatic quantification in images with fewer than 100 nuclei at different compression levels. Kaplan–Meier curves (upper row) compare the probability of difference between
TIFF/JPEG (black) and TIFF/JPEG2000 compression (grey). Bland– Altman graphs (lower row) compare count differences between TIFF/ JPEG (black) and TIFF/JPEG2000 compression (grey)
Virchows Arch
Fig. 5 Results obtained by automatic quantification in images with large-area clusters at different compression levels. Kaplan–Meier curves (upper row) compare the probability of difference between
TIFF/JPEG (black) and TIFF/JPEG2000 compression (grey). Bland– Altman graphs (lower row) compare count differences between TIFF/ JPEG (black) and TIFF/JPEG2000 compression (grey)
Fig. 6 Results obtained by automatic quantification in images with more than 100 nuclei at different compression levels. Kaplan–Meier curves (upper row) compare the probability of difference between
TIFF/JPEG (black) and TIFF/JPEG2000 compression (grey). Bland– Altman graphs (lower row) compare count differences between TIFF/ JPEG (black) and TIFF/JPEG2000 compression (grey)
Virchows Arch
resolution representation during the course of compression. In these circumstances, to reduce the number of bits of a picture below a certain point, it is advisable that the resolution of the input image be reduced with the first JPEG standard (from 4:4:4 to 4:2:2) before encoding it. This is not necessary when using JPEG2000 because this is done automatically through the format’s multiresolution decomposition structure. So, at high bit rates, where artefacts become nearly imperceptible, JPEG2000 has a small advantage over JPEG. At lower bit rates, JPEG2000 has a much more significant advantage over certain modes of JPEG in that the artefacts are less visible and there is no blocking. This may explain the greater effectiveness of nuclear quantification in JPEG2000 images when a colour (DAB) and morphological (area and roundness) segmentation model with high complexity is used at minimum and medium compression levels compared with the same images in the corresponding JPEG formats. The variations found in this work are only applicable for this kind of segmentation method based on colour and the morphological parameters cited above. Further studies are required to assess the effects of image compression with various segmentation techniques. JPEG2000 compression is a better technique than JPEG compression at low and medium levels of compression for highly complex immunohistochemically stained images. Although high-capacity data storage devices are available these days, image compression will continue to be necessary due to the large number of photographs generated in research studies by diagnostic techniques or as a result of the implementation of telepathology in pathology departments. The study of the adequate use of the different compression techniques and their effects on cell quantification could give rise to an efficient means of storage that has no loss of analytical quality and with a reduction in the time required for data transmission. Acknowledgments This work was supported by grants FIS 08/0796 (Dr. Jaén) from the Ministerio de Ciencia e Innovación (FEDER Fonds, Spain), Fundación Mutua Madrileña (FMM Dr. Bosch 2008, Spain) and the COST Action IC0604 (COST-STSM-IC0604-4046). The authors thank María del Mar Barbera, Bárbara Tomás, Vanesa Gestí, Ainhoa Montserrat, Ana Suñé, Verònica Echevarría and Marc Iniesta for their skilful technical assistance and Anna Carot, Montse Sebastià and Rosa Cabrera for their excellent secretarial work. Conflicts of interest The authors declare that they have no conflicts of interest.
References 1. Cheretis C, Angelidou E, Dietrich F, Politi E, Kiaris H, Koutselini H (2008) Prognostic value of computer-assisted morphological and morphometrical analysis for detecting the recurrence tendency of basal cell carcinoma. Med Sci Monit 14:MT13–MT19
2. Makarov DV, Marlow C, Epstein JI, Miller MC, Landis P, Partin AW, Carter HB, Veltri RW (2008) Using nuclear morphometry to predict the need for treatment among men with low grade, low stage prostate cancer enrolled in a program of expectant management with curative intent. Prostate 68:183–189 3. Alvaro T, Lejeune M, Garcia JF, Salvado MT, Lopez C, Bosch R, Jaen J, Escriva P, Pons LE (2008) Tumor-infiltrated immune response correlates with alterations in the apoptotic and cell cycle pathways in Hodgkin and Reed–Sternberg cells. Clin Cancer Res 14:685–691 4. Zhang K, Prichard JW, Yoder S, De J, Lin F (2007) Utility of SKP2 and MIB-1 in grading follicular lymphoma using quantitative imaging analysis. Hum Pathol 38:878–882 5. Joshi AS, Sharangpani GM, Porter K, Keyhani S, Morrison C, Basu AS, Gholap GA, Gholap AS, Barsky SH (2007) Semiautomated imaging system to quantitate Her-2/neu membrane receptor immunoreactivity in human breast cancer. Cytom A 71:273–285 6. Alvaro T, Lejeune M, Salvado MT, Lopez C, Jaen J, Bosch R, Pons LE (2006) Immunohistochemical patterns of reactive microenvironment are associated with clinicobiologic behavior in follicular lymphoma patients. J Clin Oncol 24:5350–5357 7. Markiewicz T, Osowski S, Patera J, Kozlowski W (2006) Image processing for accurate cell recognition and count on histologic slides. Anal Quant Cytol Histol 28:281–291 8. Johnsson A, Olsson C, Anderson H, Cavallin-Stahl E (1994) Evaluation of a method for quantitative immunohistochemical analysis of cisplatin–DNA adducts in tissues from nude mice. Cytometry 17:142–150 9. Mosedale DE, Metcalfe JC, Grainger DJ (1996) Optimization of immunofluorescence methods by quantitative image analysis. J Histochem Cytochem 44:1043–1050 10. Johansson AC, Visse E, Widegren B, Sjogren HO, Siesjo P (2001) Computerized image analysis as a tool to quantify infiltrating leukocytes: a comparison between high- and low-magnification images. J Histochem Cytochem 49:1073–1079 11. Lejeune M, Jaen J, Pons L, Lopez C, Salvado MT, Bosch R, Garcia M, Escriva P, Baucells J, Cugat X, Alvaro T (2008) Quantification of diverse subcellular immunohistochemical markers with clinicobiological relevancies: validation of a new computer-assisted image analysis procedure. J Anat 212:868–878 12. Lopez C, Lejeune M, Salvado MT, Escriva P, Bosch R, Pons LE, Alvaro T, Roig J, Cugat X, Baucells J, Jaen J (2008) Automated quantification of nuclear immunohistochemical markers with different complexity. Histochem Cell Biol 129:379–387 13. Shiao YH, Chen TJ, Chuang KS, Lin CH, Chuang CC (2007) Quality of compressed medical images. J Digit Imaging 20:149– 159 14. Tuominen VJ, Isola J (2007) The application of JPEG2000 in virtual microscopy. J Digit Imaging 22:250–258 15. Penedo M, Lado MJ, Tahoces PG, Souto M, Vidal JJ (2006) Effects of JPEG2000 data compression on an automated system for detecting clustered microcalcifications in digital mammograms. IEEE Trans Inf Technol Biomed 10:354–361 16. Belair ML, Fansi AK, Descovich D, Leblanc AR, Harasymowycz P (2005) The effect of compression on clinical diagnosis of glaucoma based on non-analyzed confocal scanning laser ophthalmoscopy images. Ophthalmic Surg Lasers Imaging 36:323– 326 17. Ohgiya Y, Gokan T, Nobusawa H, Hirose M, Seino N, Fujisawa H, Baba M, Nagai K, Tanno K, Takeyama N, Munechika H (2003) Acute cerebral infarction: effect of JPEG compression on detection at CT. Radiology 227:124–127 18. CM EF, Winkler S (2004) JPEG vs. JPEG2000: an objective comparison of image encoding quality. Proc SPIE Appl Digit Image Proc 5558:300–308
Virchows Arch 19. Erickson BJ (2002) Irreversible compression of medical images. J Digit Imaging 15:5–14 20. Janhom A, van der Stelt PF, Sanderink GC (2002) A comparison of two compression algorithms and the detection of caries. Dentomaxillofac Radiol 31:257–263 21. Slone RM, Foos DH, Whiting BR, Muka E, Rubin DA, Pilgram TK, Kohm KS, Young SS, Ho P, Hendrickson DD (2000) Assessment of visually lossless irreversible image compression: comparison of three methods by using an image-comparison workstation. Radiology 215:543–553 22. Matkowskyj KA, Cox R, Jensen RT, Benya RV (2003) Quantitative immunohistochemistry by measuring cumulative signal strength accurately measures receptor number. J Histochem Cytochem 51:205–214 23. Sharangpani GM, Joshi AS, Porter K, Deshpande AS, Keyhani S, Naik GA, Gholap AS, Barsky SH (2007) Semi-automated imaging system to quantitate estrogen and progesterone receptor immunoreactivity in human breast cancer. J Microsc 226:244–255 24. Lopez C, Lejeune M, Escria P, Bosch R, Salvadoacute MT, Pons LE, Baucells J, Cugat X, Alvaro T, Jaen J (2008) Effects of image compression on automatic count of immunohistochemically stained nuclei in digital images. J Am Med Inform Assoc 15:794–798 25. Alvaro-Naranjo T, Lejeune M, Salvado-Usach MT, Bosch-Princep R, Reverter-Branchat G, Jaen-Martinez J, Pons-Ferre LE (2005) Tumor-infiltrating cells as a prognostic factor in Hodgkin’s lymphoma: a quantitative tissue microarray study in a large retrospective cohort of 267 patients. Leuk Lymphoma 46:1581– 1591 26. Bryant RJ, Banks PM, O’Malley DP (2006) Ki67 staining pattern as a diagnostic tool in the evaluation of lymphoproliferative disorders. Histopathology 48:505–515 27. Yerushalmi R, Woods R, Ravdin PM, Hayes MM, Gelmon KA (2010) Ki67 in breast cancer: prognostic and predictive potential. Lancet Oncol 11:174–183 28. Montebugnoli L, Badiali G, Marchetti C, Cervellati F, Farnedi A, Foschini MP (2009) Prognostic value of Ki67 from clinically and histologically ‘normal’ distant mucosa in patients surgically treated for oral squamous cell carcinoma: a prospective study. Int J Oral Maxillofac Surg 38:1165–1172 29. Ooi SM, Vivian JB, Cohen RJ (2009) The use of the Ki-67 marker in the pathological diagnosis of the epithelioid variant of renal angiomyolipoma. Int Urol Nephrol 41:559–565 30. Aslan DL, Gulbahce HE, Pambuccian SE, Manivel JC, Jessurun J (2005) Ki-67 immunoreactivity in the differential diagnosis of pulmonary neuroendocrine neoplasms in specimens with extensive crush artifact. Am J Clin Pathol 123:874–878 31. Rioux-Leclercq N, Leray E, Patard JJ, Lobel B, Guille F, Jouan F, Bellaud P, Epstein JI (2005) The utility of Ki-67 expression in the differential diagnosis of prostatic intraepithelial neoplasia and ductal adenocarcinoma. Hum Pathol 36:531–535 32. Carreras J, Lopez-Guillermo A, Fox BC, Colomo L, Martinez A, Roncador G, Montserrat E, Campo E, Banham AH (2006) High numbers of tumor-infiltrating FOXP3-positive regulatory T cells are associated with improved overall survival in follicular lymphoma. Blood 108:2957–2964 33. Glas AM, Knoops L, Delahaye L, Kersten MJ, Kibbelaar RE, Wessels LA, van Laar R, van Krieken JH, Baars JW, Raemaekers J, Kluin PM, van’t Veer LJ, de Jong D (2007) Gene-expression and immunohistochemical study of specific T-cell subsets and accessory cell types in the transformation and prognosis of follicular lymphoma. J Clin Oncol 25:390–398
34. Lee AM, Clear AJ, Calaminici M, Davies AJ, Jordan S, MacDougall F, Matthews J, Norton AJ, Gribben JG, Lister TA, Goff LK (2006) Number of CD4+ cells and location of forkhead box protein P3-positive cells in diagnostic follicular lymphoma tissue microarrays correlates with outcome. J Clin Oncol 24:5052–5059 35. Alvaro T, Lejeune M, Salvado MT, Bosch R, Garcia JF, Jaen J, Banham AH, Roncador G, Montalban C, Piris MA (2005) Outcome in Hodgkin’s lymphoma can be predicted from the presence of accompanying cytotoxic and regulatory T cells. Clin Cancer Res 11:1467–1473 36. Curiel TJ (2007) Tregs and rethinking cancer immunotherapy. J Clin Invest 117:1167–1174 37. de Jong D, Koster A, Hagenbeek A, Raemaekers J, Veldhuizen D, Heisterkamp S, de Boer JP, van Glabbeke M (2009) Impact of the tumor microenvironment on prognosis in follicular lymphoma is dependent on specific treatment protocols. Haematologica 94:70– 77 38. Brennecke R, Burgel U, Rippin G, Post F, Rupprecht HJ, Meyer J (2001) Comparison of image compression viability for lossy and lossless JPEG and wavelet data reduction in coronary angiography. Int J Cardiovasc Imaging 17:1–12 39. Eikelboom RH, Yogesan K, Barry CJ, Constable IJ, Tay-Kearney ML, Jitskaia L, House PH (2000) Methods and limits of digital image compression of retinal images for telemedicine. Investig Ophthalmol Vis Sci 41:1916–1924 40. Hui OBR (2002) Medical image compression using JPEG2000 and JPEG: a comparison study. J Mech Med Biol 2:313–328 41. Iyriboz TA, Zukoski MJ, Hopper KD, Stagg PL (1999) A comparison of wavelet and joint photographic experts group lossy compression methods applied to medical images. J Digit Imaging 12:14–17 42. Ricke J, Maass P, Lopez Hanninen E, Liebig T, Amthauer H, Stroszczynski C, Schauer W, Boskamp T, Wolf M (1998) Wavelet versus JPEG (Joint Photographic Expert Group) and fractal compression. Impact on the detection of low-contrast details in computed radiographs. Investig Radiol 33:456–463 43. Li F, Sone S, Takashima S, Kiyono K, Yang ZG, Hasegawa M, Kawakami S, Saito A, Hanamura K, Asakura K (2001) Effects of JPEG and wavelet compression of spiral low-dose CT images on detection of small lung cancers. Acta Radiol 42:156–160 44. Kalyanpur A, Neklesa VP, Taylor CR, Daftary AR, Brink JA (2000) Evaluation of JPEG and wavelet compression of body CT images for direct digital teleradiologic transmission. Radiology 217:772–779 45. Siddiqui KM, Johnson JP, Reiner BI, EL S (2005) Discrete cosine transform JPEG compression vs. 2D JPEG2000 compression: JNDmetrix visual discrimination model image quality analysis. Proc. SPIE 5748:202–207 46. Matkowskyj KA, Schonfeld D, Benya RV (2000) Quantitative immunohistochemistry by measuring cumulative signal strength using commercially available software Photoshop and MATLAB. J Histochem Cytochem 48:303–312 47. Singh SS, Kim D, Mohler JL (2005) Java Web start based software for automated quantitative nuclear analysis of prostate cancer and benign prostate hyperplasia. Biomed Eng Online 4:31 48. Atalag K, Sincan M, Celasun B, Karaagaoglu E (2004) Effects of lossy image compression on quantitative image analysis of cell nuclei. Anal Quant Cytol Histol 26:22–27
