Web-based tissue microarray image data analysis: Initial validation testing through prostate cancer Gleason grading

Citation
Gs. Bova et al., Web-based tissue microarray image data analysis: Initial validation testing through prostate cancer Gleason grading, HUMAN PATH, 32(4), 2001, pp. 417-427
Citations number
13
Categorie Soggetti
Research/Laboratory Medicine & Medical Tecnology","Medical Research Diagnosis & Treatment
Journal title
HUMAN PATHOLOGY
ISSN journal
00468177 → ACNP
Volume
32
Issue
4
Year of publication
2001
Pages
417 - 427
Database
ISI
SICI code
0046-8177(200104)32:4<417:WTMIDA>2.0.ZU;2-2
Abstract
Tissue microarray technology promises to enhance tissue-based molecular res earch by allowing improved conservation of tissue resources and experimenta l reagents, improved internal experimental control, and increased sample nu mbers per experiment, Organized, well-validated collection and analysis of the voluminous image data produced by tissue microarray technology is criti cal to maximize its value. Web-based technology for visual analysis and sea rchable storage of microarray image data could provide optimal flexibility for research groups in meeting this goal, but this approach has not been ex amined scientifically. Toward this goal, a prostate tissue microarray block containing 432 tissue cores (0.6 mm diameter) was constructed. Moderately compressed (200 kb).jpg images of each tissue spot were acquired and were s aved using a naming convention developed by the SPORE Prostate Tissue Micro array Collaborative Group. Four hundred three tissue array spot images were uploaded into a database developed for this study and were converted to .f px format to decrease Internet transmission limes for high-resolution image data. In phase T of the image analysis portion of the study, testing and p reliminary analysis of the Web technology was performed by 2 pathologists ( M.A.R. and G.S.B.). In phase II, 2 pathologists (J.I.E. and T.M.W.) with no previous exposure to this technology and no knowledge of the structure of the study were presented a set of 130 sequential tissue spot images via the Web on their office computers. In phase III, the same pathologists were pr esented a set of 193 images, including all 130 from phase II and 63 others, with image presentation order randomized. With each zoomable tissue spot i mage, each pathologist was presented with a nested set of questions regardi ng overall interpretability of the image, presence or absence of cancer, an d predominant and second most frequent Gleason grade. In phases II and III of the study, 319 of 323 (99%) image presentations using this Web technolog y were rated interpretable. Comparing the 2 pathologists' readings in phase s II and III, Gleason grade determinations by each pathologist were identic al in 179 of 221 (81%,) determinations and were within 1 point of each othe r in 221 of 221 (100%) determinations, a performance rate similar to if not better than that previously reported for direct microscopic Gleason gradin g. Interobserver comparison of Gleason score determinations and intraobserv er comparisons for Gleason grade and score also showed a pattern of uniform ity similar to those reported in direct microscope-based Gleason grading st udies. Interobserver (7.5%) and intraobserver (5% and 3%) variability in de termining whether diagnosable cancer was present point out the existence of a "threshold effect" that has rarely been studied but may provide a basis for identification of features that are most amenable to improved diagnosti c standardization. In summary, storage and analysis of tissue microarray sp ot images using Web-based technology is feasible and practical, and the qua lity of images obtained using the techniques described here appears adequat e for most tissue-based pathology research applications.