Page 59 - חוברת תזות הנדסה ירוקה 2023

Characterization and Detection of Primary, Pre-Malignantand Malignant Cells Using Raman Spectroscopy andMachine Learning Algorithms
Daniel Hania1: danietu1@sce.ac.il
Prof. Ahmad Salman1, Prof. Mahmoud Huleihel2, Prof. Itshak Lapidot3
1SCE - Shamoon College of Engineering, Be’er-sheva, 2Ben-Gurion University of the Negev, 3Afeka Tel-Aviv Academic College of Engineering
Cancer is the most common fatal disease in Israel and around the globe. Annually, 19 million new cancer patients are diagnosed and about 10 million deaths are estimated due to cancer. In this era, when the life expectancy of the population is increasing, the number of cancer patients is also increasing. Cancer patients struggle in daily life with difficult treatments, pain, and financial, social, and personal hardship, which is also reflected in financial difficulties for their family members and society. Currently, the process of accurately diagnosing cancer using classical methods involves a series of tests, most of which are time-consuming, and cause discomfort and anxiety in the patients. While waiting to receive the interpretation of each test, the tumor continues to develop in the body and may require more extensive treatment. Moreover, performing these tests produces a lot of medical waste, termed genotoxic waste, as defined by the World Health Organization )WHO(. Genotoxic wastes are an extremely dangerous type of hospital waste that may cause cell mutation or cancer in those who come in contact with it. All the above illustrates the need for a fast, reliable, simple, and accurate modern diagnostic method to detect pre-cancerous and cancerous cells in the first stages of their development. Raman spectroscopy is known to be a reliable, fast, sensitive, and accurate method for identifying and characterizing biological systems at the molecular level, without the need for reagents that d become medical waste. It is based on the interaction between a laser beam and a diagnostic sample, and measures the vibrational energy of the tested sample, to get its Raman spectrum, its ‘biochemical fingerprint’.Inthisstudy,mouseembryonicprimarycells)mEPCs(representnormalhealthy cells, mouse fibroblast cells )MEFs( from the National Institutes of Health cell cultures )NIHs( represent precancerous cells, and MEFs transformed into cancer cells are labeled MBM )‘mouse bone marrow’(. During this study, 997 measurements were taken from 457 different cells from all three biological systems: normal, precancerous, and cancerous. The spectral differences between these three systems were minute; therefore, various methods of machine learning were used to identify and classify them. Individual cells were measured at three different locations )cell center, cytoplasm, and cell membrane( to identify the areas responsible for the main spectral differences between the three systemic models. A linear discriminant analysis )LDA( classifier, following principal component analysis )PCA( calculations, simultaneously classified these three biological systems with an accuracy of
59