Groundbreaking Cancer Research from Boğaziçi University Scientist

Artificial intelligence models called "PathoSeg" and "PathopixGAN," developed by Doç. Dr. Mehmet Turan and his team at Boğaziçi University Department of Computer Engineering, accelerate diagnostic processes for diseases such as cancer while improving diagnostic accuracy.

"The diagnostic process in pathology has long been limited to visual examinations conducted under a microscope. Thanks to the artificial intelligence models we developed, segmentation of cell and tissue regions can now be performed both faster and more precisely," said Doç. Dr. Mehmet Turan, whose research paper was published in Medical Image Analysis, a prestigious scientific journal from Elsevier. 

Artificial intelligence and machine learning research continues in multiple centres and laboratories at Boğaziçi University. Most recently, the latest work by Doç. Dr. Mehmet Turan and his team on applications of artificial intelligence and deep learning technologies in the field of pathology was published in Medical Image Analysis, a prestigious scientific journal from Elsevier. Doç. Dr. Turan stated that they aim to integrate this technology into clinical applications to support faster, more reliable and personalised treatment options.

"We have brought significant innovation to the diagnostic process"

Doç. Dr. Mehmet Turan, stating that the models "PathoSeg" and "PathopixGAN" they developed using artificial intelligence and deep learning technologies aim to take the diagnostic process for diseases such as cancer beyond microscopic examinations, said: "The diagnostic process in pathology has long been limited to visual examinations conducted under a microscope. With our work, we have brought significant innovation to this process through artificial intelligence. Thanks to our artificial intelligence model 'PathoSeg', segmentation of cell and tissue regions can now be performed both faster and more precisely. This makes the diagnostic process more efficient and allows for much more accurate detection of cancerous areas." He added that early detection of cancer cell metastasis is also possible with the model, noting that the model's superior performance increases diagnostic accuracy. He stated: "'PathoSeg' model increases diagnostic accuracy with its superior performance in segmenting cancerous cells and tissues, while at the same time reducing the workload of doctors. It can make a meaningful contribution to patient care by performing accurate and rapid analyses, particularly in critical areas such as early detection of metastasis or monitoring of treatment processes." 

"We are addressing data problems"

Doç. Dr. Turan, noting that issues arising in histopathology data are resolved through the "PathopixGAN" model, described it as follows: "With conventional data collection methods, there is serious imbalance in histopathology data. Particularly rare cases may be insufficient for the model to learn. 'PathopixGAN' generates highly realistic and diverse synthetic images, enabling our model to be trained with a broader dataset. Thus, we can successfully perform segmentation even of rarely seen pathological structures. This is an important step taken regarding data imbalance." 

 "We aim for the models to become reference points worldwide"

Emphasising that the models presented have a pioneering quality in their field, Doç. Dr. Turan said: "Academically, we are providing a robust model and data source for other researchers. We aim to be an important reference point in the process of adopting artificial intelligence. From an industry perspective, we hope to set an example in integrating artificial intelligence into clinical diagnostic processes and to contribute to increased use of artificial intelligence across healthcare services. Our goal is to integrate this technology into clinical applications to support faster, more reliable and personalised treatment options."