ICHSM 2024

Invited Speakers

Prof. Dr. Ahmet Murat Ozbayoglu

TOBB University of Economics and Technology, Turkiye

A.M. Ozbayoglu graduated from the Department of Electrical Engineering at METU, Ankara, Turkey in 1991, then he got his Msc and PhD degrees from the department of Engineering Management at Missouri University of Science and Technology, USA in 1993 and 1996, respectively. After graduation, he joined MEMC Electronics (now became SunEdison), USA as a software project engineer, programmer and analyst working on silicon wafer manufacturing software and data automation projects. In 2005, he went back to academia by joining the Department of Computer Engineering of TOBB University of Economics and Technology, in Ankara, Turkey. His research interests include machine learning, pattern recognition, deep learning, financial forecasting, computational intelligence, machine vision. He has conducted 20 MSc and 2 PhD theses in theoretical and applied machine learning. He has published more than 40 journal and 100 international conference papers along with numerous white papers and technical reports. He has served in many academic and industrial projects as principal investigator, researcher and consultant. Also, he has been actively involved in social and technical committes both on and off-campus. He is a member of ACM and IEEE Computational Intelligence Society.

Speech title "Generative Adversarial Networks in Medical Imaging and Health Informatics"

Abstract-Generative Adversarial Networks (GANs) have emerged as a revolutionary technology in the field of medical imaging and health informatics, offering significant advancements in data augmentation, image reconstruction, and segmentation. Meanwhile, the transformative potential of GANs also have their benefits and drawbacks.
GANs facilitate data augmentation by generating realistic synthetic medical images, thus overcoming the limitations of small datasets and enhancing the training of machine learning models. This capability is particularly vital in medical domains where acquiring large volumes of labeled data is challenging and expensive. Through data augmentation, GANs contribute to improved diagnostic accuracy and robust predictive modeling.
In image reconstruction, GANs excel in restoring high-quality images from low-resolution or corrupted inputs. This has effective implications for enhancing the clarity and detail of medical images such as MRI, CT, and ultrasound scans. The ability to reconstruct images with high fidelity aids clinicians in making more accurate diagnoses and treatment plans, ultimately improving patient outcomes.
Segmentation, a critical task in medical image analysis, also benefits from the application of GANs. These networks can portray anatomical structures and pathological regions with remarkable precision, facilitating tasks such as tumor detection, organ segmentation, and lesion quantification. The improved segmentation accuracy offered by GANs supports better-informed clinical decisions and personalized patient care.
However, the implementation of GANs in medical imaging and health informatics is not without challenges. The training process of GANs is notoriously complex, requiring substantial computational resources and expertise to achieve convergence. Additionally, the synthetic data generated by GANs may occasionally introduce artifacts or biases, potentially impacting clinical interpretations. Ethical considerations surrounding data privacy and the synthetic nature of GAN-generated images also warrant careful examination.
Hence, GANs present a promising avenue for advancing medical imaging and health informatics, offering substantial benefits in data augmentation, image reconstruction, and segmentation. As the technology matures, addressing the associated challenges will be crucial to fully realizing its potential and ensuring its safe and effective integration into clinical practice.