According to Mehr news agency, quoting Tehran University, a group of researchers from Tehran University’s technical schools, headed by Ali Masoudinejad, professor of systems biology and bioinformatics at the Faculty of Engineering Sciences, with the cooperation of Behnaz Haji-Molahsini, a PhD student; Ahmadreza Iranpour, a student at Utrecht University in the Netherlands and Souda Imani, a student at Pazmani University in Hungary, and other researchers of the System Biology and Bioinformatics Laboratory, succeeded in conducting two innovative scientific researches in the field of deep learning and personalized medicine.
Masoudinejad said in this regard: These studies, which were published in the international journals of Elsevier Publications with an impact factor of 13 and 6.3, respectively, focus on the use of artificial intelligence models to improve the accuracy of biological and pathological data analysis.
He added: In the first research, we categorized the methods into four main groups of methods based on deep learning such as GAN and autoencoder, traditional methods such as histogram matching, hybrid models and a new method based on signal processing and showed that each of these approaches has its own advantages and limitations.
Professor of Biology Systems of Tehran University continued: The results of our study emphasize the importance of preserving biological information in the normalization process and its role in increasing the accuracy of computerized diagnosis systems.
Referring to the details of the research, Masoudinejad explained: In addition to a detailed review of previous studies, we presented a new framework for a systematic comparison between normalization methods, which provides the possibility of evaluating the performance of the methods in different conditions.
He added: This framework can help digital pathology researchers and specialists in choosing the most optimal method according to the type of data and the purpose of the research.
Masoudinejad said about the second study: In this study, the focus was on multi-omics data analysis and the application of deep learning models to predict survival in endometrioid uterine cancer patients.
He explained: In this research, gene expression, DNA methylation, and proteome data from the TCGA-UCEC project were analyzed, and a new autoencoder with a dedicated cost function was designed to better identify complex nonlinear relationships between biological characteristics and survival rates.
This researcher added: The results showed that this approach extracts information related to survival more accurately than usual methods and led to the identification of key molecular pathways, such as vitamin D pathway and galanin receptor, which are related to the prognosis of patients.
In the end, Masoudinejad noted: In our opinion, the integration of deep learning methods with the analysis of biological and image data can lead to a deeper understanding of the molecular mechanisms of diseases and the development of personalized treatment strategies.
Latest Passing over countries : Spain | Dominica | United Arab Emirates
RCO NEWS
