Evaluation of the Influence of Image Compression to the Automatic Discrimination of Histological Images of Skin Lesions

Summary

Objectives: Telemedicine offers the possibility to get the opinion of an expert within a short time. To reduce data transfer via a network image compression is recommended. The disadvantage of error-free compression is a low compression rate. On the other hand an increased compression rate affects the information content of the image. In this study we evaluated the influence of the compression rate on the automatic discrimination of histological skin lesions in dermatopathology.

Methods: To be independent from subjective reviewing by a dermatopathologist we used tissue counter analysis (TCA) for automatic discrimination of skin tissue. TCA is based on the partition of the image into square elements where the features are calculated out of each square element. 40 cases of benign common nevi and 40 cases of malignant melanoma were used as the study set. First TCA was applied to the uncompressed images to check the discrimination power of the method. Then in the next steps the method was applied to the images with successively higher compression rates. For image compression the wavelet compression was used, where the compression rate was determined by neglecting wavelet coefficients with a magnitude below a given threshold. The number of remaining wavelet coefficients was used as criteria for the compression rate.

Results: This study shows that TCA allows automated discrimination even at higher compression rates where only 6-18% of the wavelet coefficients are used for image reconstruction. The recognition rate at higher compression is better for malignant melanoma than for benign common nevi.

Conclusion: The power of automated discrimination is not essential affected by wavelet image compression.

Keywords

Medical image processing - tissue counter analysis - image compression - teledermatology

• References

• 1 Hicks LL, Boles KE, Hudson S, Kling B, Tracy J, Mitchell J, Webb W. Patient satisfaction with teledermatology services. J Telemed Telecare 2003; 9 (01) 42-5.
• 2 Cipolat C, Bader U, Rufli T, Burg G. Teledermatology in Switzerland. Curr Probl Dermatol 2003; 32: 257-60.
• 3 Schmid-Grendelmeier P, Doe P, Pakenham-Walsh N. Teledermatology in sub-Saharan Africa. Curr Probl Dermatol 2003; 32: 233-46.
• 4 Puniena J, Punys V, Punys J. Ultrasound and angio image compression by cosine and wavelet transforms. International Journal of Medical Informatics 2001; 64: 473-81.
• 5 Terae S, Miyasaka K, Kudoh K, Nambu T, Shimizu T, Kaneko K, Yoshikawa H, Kishimoto R, Omatsu T, Fujita N. Wavelet compression on detection of brain lesions with magnetic resonance imaging. Journal of Digital Imaging 2000; 13 (04) 178-90.
• 6 Mello-Thoms C, Dunn SM, Nodine CF, Kundel HL. An analysis of perceptual errors in reading mammograms using quasi-local spatial frequency spectra. Journal of Digital Imaging 2001; 14 (03) 117-23.
• 7 Kerut EK, Given MB, Mc Ilwain E, Allen G, Espinoza C, Giles TD. Echocardio-graphic texture analysis using the wavelet transform: Differentiation of early heart muscle disease. Ultrasound in Medicine and Biology 2000; 26 (09) 1445-53.
• 8 Harms H, Aus HM, Haucke M, Gunzer U. Segmentation of stained blood cell images measured at high scanning density with high magnification and high numerical aperture optics. Cytometry 1992; 7: 522-31.
• 9 Stolz W, Vogth T, Landthaler M, Hempfer S, Bingler P, Abmayr W. Differentiation between maligant melanomas and benign melanocytic nevi by computerized DNA cytometry of imprint specimens. J Cutan Pathol 1994; 21: 7-15.
• 10 Brown AR. Combined immunocytochemical staining and image analysis for the study of lymphocyte specificity and function in situ. J Immunol Methods 1990; 130: 410-4.
• 11 Hamilton PW, Bartels PH, Montironi R. et al. Automated histometry in quantitative prostate pathology. Analytical and Quantitative Cytology and Histology 1998; 20: 443-60.
• 12 Sahn EE, Thiers BH, Nicholson JH, Maize JC. A new method for comprehensive automatic morphologic image analysis of Langerhans cells. Am J Dermatopathol 1988; 10: 410-4.
• 13 Handels H, Roß T, Kreusch J, Wolff HH, Pöppl SJ. Computer-Supported Diagnosis of Melanoma in Profilometry. Methods Infor Med 1999; 38: 43-9.
• 14 Smolle J. Computer recognition of skin structures using discriminant and cluster analysis. Skin Research and Technology 2000; 6: 58-63.
• 15 Smolle J. Diagnostic Assessment of Cutaneous Melanoma and Common Nevi Using Tissue Counter Analysis. Analytical and Quantitative Cytology and Histology, 2000; 22 (04) 299-306.
• 16 Wiltgen M, Gerger A, Smolle J. Tissue counter analysis of benign common nevi and malignant melanoma. Int J Med Informatics 2003; 69 (01) 17-28.
• 17 Haralick RM, Shanmugam K, Dinstein I. Textural Features for Image Classification. EEE Transactions on Systems Man Cybernetics 1973; 3: 610-21.
• 18 Smolle J, Gerger A, Weger W, Kutzner H, Tronnier M. Tissue Counter Analysis of Histologic Sections of Melanoma: Influence of Mask Size and Shape, Feature Selection, Statistical Methods and Tissue Preparation. Anal Cell Pathol 2002; 24 (2, 3) 59-67.
• 19 Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and Regression Trees. New York, London: Chapman & Hall; 1993
• 20 Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes in C: The Art of Scientific Computing. 2nd ed. Cambridge University Press; 1992: 591-606.