ردیابی خودکار سلول‌ها درتصاویر میکروسکوپی گذر-زمانی

ردﯾﺎﺑﯽ ﺳﻠﻮل ﻫﺎ در ﻃﻮل زﻣﺎن ﯾﮑﯽ از ﻣﻮﺛﺮﺗﺮﯾﻦ اﻧﻮاع ﻣﻄﺎﻟﻌﻪ ﺑﺮای درک ﺳﺎزوﮐﺎر ﻫﺎی رﺷﺪ، ﺗﺮﻣﯿﻢ ﺑﺎﻓﺖ و درﻣﺎن ﺑﯿﻤﺎری در ﻣﻮﺟﻮدات زﻧﺪه اﺳﺖ. ﻫﺪف از اﯾﻦ ﺗﺤﻘﯿﻖ اراﺋﻪ روﺷﯽ ﺟﺪﯾﺪ ﺑﺎ ﮐﻤﮏ ﯾﺎدﮔﯿﺮی ﻣﺎﺷﯿﻨﯽ ﺑﺮای ردﯾﺎﺑﯽ ﺳﻠﻮل ﻫﺎ ﺑﻪ ﺷﮑﻞ ﺧﻮدﮐﺎر در ﺗﺼﺎوﯾﺮ ﻣﯿﮑﺮوﺳﮑﻮﭘﯽ ﮔﺬر-زﻣﺎﻧﯽ اﺳﺖ. ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﭘﯿﺸﺮﻓﺖ ﻫﺎی ﭼﺸﻤﮕﯿﺮ در ﻓﻨﺎوری ﻣﯿﮑﺮوﺳﮑﻮپ در ﺳﺎل ﻫﺎی اﺧﯿﺮ، از ﻣﯿﮑﺮوﺳﮑﻮپ ﮔﺬر-زﻣﺎﻧﯽ ﻣﻌﻤﻮﻻً ﺑﺮای ﻣﻄﺎﻟﻌﻪ ﺳﻠﻮل ﻫﺎی درون اﻧﺪام ﻫﺎ اﺳﺘﻔﺎده ﻣﯽ ﺷﻮدﮐﻪ اﻣﮑﺎن ﺗﺠﺰﯾﻪ و ﺗﺤﻠﯿﻞ ﻣﺴﺘﻘﯿﻢ رﻓﺘﺎرﻫﺎی ﺳﻠﻮﻟﯽ را ﻓﺮاﻫﻢ ﻣﯽ ﮐﻨﺪ.در اﯾﻦ روش ﺑﺮای ردﯾﺎﺑﯽ ﺳﻠﻮل ﻫﺎ، ﭘﺲ از ﻗﻄﻌﻪ ﺑﻨﺪی ﺗﺼﺎوﯾﺮ ﺑﺎ ﯾﺎدﮔﯿﺮی ﻋﻤﯿﻖ، ﻋﻤﻞ ﺗﺸﺨﯿﺺ روﯾﺪاد ﺗﻘﺴﯿﻢ ﺳﻠﻮﻟﯽ ﺑﺎ ﻣﻨﺤﻨﯽ ﮐﺎﺳﯿﻨﯽ و ﻋﻤﻞ ﭘﯿﮕﺮد ﺳﻠﻮل ﻫﺎ در ﻓﺮﯾﻢ ﻫﺎی ﻣﺘﻮاﻟﯽ ﺑﺎ روش ﺗﻄﺒﯿﻖ ﻗﺎﻟﺐ ﻣﺤﺪود ﺷﺪه اﻧﺠﺎم ﻣﯽ ﺷﻮد. ﻋﻤﻠﮑﺮد ردﯾﺎب ﭘﯿﺸﻨﻬﺎدی ﺗﻮﺳﻂ ﻣﻌﯿﺎر tra ﮐﻪ ﯾﮏ ﻣﻌﯿﺎر اﺳﺘﺎﻧﺪارد ﺑﺮای ارزﯾﺎﺑﯽ اﻟﮕﻮرﯾﺘﻢ ﻫﺎی ردﯾﺎﺑﯽ ﺳﻠﻮل ﻫﺎ اﺳﺖ، اﻧﺪازه ﮔﯿﺮی ﺷﺪ. ردﯾﺎب ﭘﯿﺸﻨﻬﺎدی روی دو ﻣﺠﻤﻮﻋﻪ داده dic-c2dh-hela و fluo-n2dh-sim+ دارای ﻋﻤﻠﮑﺮد 96% و 95% در ردﯾﺎﺑﯽ اﺳﺖ و در ﻣﻌﯿﺎرﻫﺎی ﺗﺮﮐﯿﺒﯽ ﻗﻄﻌﻪ ﺑﻨﺪی و ردﯾﺎﺑﯽ ﺑﻬﺒﻮد 5 درﺻﺪی را ﻧﺸﺎن داد. ﺑﺮ اﺳﺎس آزﻣﺎﯾﺸﺎت ﺑﺎ اﻓﺰاﯾﺶ دﻗﺖ ﻗﻄﻌﻪ ﺑﻨﺪی و ﺗﺸﺨﯿﺺ ﺗﻘﺴﯿﻢ ﺳﻠﻮﻟﯽ، ﻋﻤﻠﮑﺮد ﮐﻠﯽ اﻟﮕﻮرﯾﺘﻢ ردﯾﺎﺑﯽ ﺳﻠﻮل ﻫﻢ اﻓﺰاﯾﺶ ﻣﯽ ﯾﺎﺑﺪ.

automatic cell tracking in time-lapse microscopy images

tracking cells over time is one of the most effective ways to understand the mechanisms of growth, tissue repair, and disease treatment in living organisms. the goal of this research is to present a new method using machine learning for automatic cell tracking in time-lapse microscopy images. with the remarkable advancements in microscopy technology in recent years, time-lapse microscopy is commonly used to study cells within organs, enabling direct analysis of cellular behaviors. in this method, after segmenting the images using deep learning, the detection of cell division events is performed using the cassini curve, and cell tracking across consecutive frames is carried out using a constrained template matching method. the performance of the proposed tracker was measured using the tra metric, a standard criterion for evaluating cell tracking algorithms. the proposed tracker achieved 96% and 95% performance in tracking on two datasets, dic-c2dh-hela and fluo-n2dh-sim+, respectively, and showed a 5% improvement in combined segmentation and tracking metrics. based on the experiments, increasing the accuracy of segmentation and cell division detection also enhances the overall performance of the cell tracking algorithm.