Post by Emin Eray Ertan

I am incredibly happy to share an exciting step in my academic journey in the field of Artificial Intelligence and Machine Learning! We have submitted our paper, which we have been working on for a long time, titled "Compression Sensitivity of Metastatic Regions in Whole Slide Images: A Comparative Study of JPEG, JPEG2000, and Autoencoder-Based Codecs," to the ASYU conference, and it has officially entered the peer-review process. Focus of Our Study: In digital pathology, Whole Slide Images (WSI) can reach gigabytes in size, making image compression mandatory for practical storage and transmission. However, the impact of these compression processes, especially on diagnostically critical metastatic (tumor) tissues, had not been sufficiently investigated in the literature. In this study, using the Camelyon16 dataset, we compared the compression sensitivity of tumor and normal tissues. Key Findings: Vulnerability of tumor tissues: Metastatic regions show a significantly higher sensitivity to compression artifacts compared to normal tissues across all classic JPEG compression levels. Superiority of JPEG2000: At mid and high-quality settings, JPEG2000 successfully preserves structural integrity without creating a statistically significant difference between tissue types. Artificial Intelligence (Autoencoder) Codecs: The shallow-architecture autoencoder models we examined fall behind classic codecs with their current training configurations, offering limited suitability for medical imaging at this stage. As a first-year undergraduate student representing Kayseri University, making such a concrete contribution to the field of digital pathology and medical AI is an invaluable experience for me. I eagerly await a positive outcome from the review process and look forward to sharing the details of our research with our colleagues at the conference. #ArtificialIntelligence #MachineLearning #DigitalPathology #ImageCompression #DeepLearning #ASYU #MedicalImaging