دسته بندي تصاوير پزشكي ضايعات پوستي با استفاده از شبكه عصبي كپسولي

Classification of medical images of skin lesions using capsular neural network

ﺷﺒﮑﻪ ﻫﺎي ﻋﻤﯿﻖ ﻧﻮﻋﯽ از روش ﻫﺎي ﯾﺎدﮔﯿﺮي ﻫﺴﺘﻨﺪ ﮐﻪ ﻗﺎﺑﻠﯿﺖ ﻣﺪل ﮐﺮدن رواﺑﻂ ﺳﻄﺢ ﺑﺎﻻي ﻣﻮﺟﻮد در داده ﻫﺎ را دارﻧﺪ. ﯾﮑﯽ از ﭘﺮﮐﺎرﺑﺮدﺗﺮﯾﻦ اﻧﻮاع ﺷﺒﮑﻪ ﻫﺎي ﻋﻤﯿﻖ، ﺷﺒﮑﻪ ﻫﺎي ﭘﯿﭽﺸﯽ ﯾﺎ ﮐﺎﻧﻮﻟﻮﺷﻨﯽ ﻫﺴﺘﻨﺪ ﮐﻪ ﺑﺎ ﺑﻬﺮه ﮔﯿﺮي از ﻻﯾﻪ ﻫﺎي ﮐﺎﻧﻮﻟﻮﺷﻦ ﺑـﺮ روي ﺗﺼـﺎوﯾﺮ ﻗﺎدر ﺑﻪ ﻣﺪل ﮐﺮدن واﺑﺴﺘﮕﯽ ﻫﺎي ﻣﮑﺎﻧﯽ در آن ﻫﺎ ﻫﺴﺘﻨﺪ، اﻣﺎ ﺳﺎﺧﺘﺎرﻫﺎي ﺳﻠﺴـﻠﻪ-ﻣﺮاﺗﺒـﯽ ﻣﮑـﺎﻧﯽ درون ﺗﺼـﻮﯾﺮ را در ﻧﻈـﺮ ﻧﻤﯽ ﮔﯿﺮﻧـﺪ. ﺷﺒﮑﻪ ﻫﺎي ﮐﭙﺴﻮﻟﯽ ﯾﮑﯽ از اﯾﺪه ﻫﺎي ﺟﺪﯾﺪي ﻫﺴﺘﻨﺪ ﮐﻪ ﺑﺮاي ﻣﺪل ﺳﺎزي ﺳﺎﺧﺘﺎر ﺳﻠﺴﻠﻪ-ﻣﺮاﺗﺒﯽ وﯾﮋﮔﯽ ﻫﺎ در ﺗﺼﻮﯾﺮ ﭘﯿﺸـﻨﻬﺎد ﺷـﺪه اﻧﺪ و در آﻧﻬﺎ از ﮐﭙﺴﻮل ﯾﺎ ﻧﻮرون ﻫﺎي ﮔﺮوه ﺑﻨﺪي ﺷﺪه ﺑﻪ ﻫﻤﺮاه ﯾـﮏ اﻟﮕـﻮرﯾﺘﻢ ﻣﺴـﯿﺮﯾﺎﺑﯽ ﭘﻮﯾـﺎ اﺳـﺘﻔﺎده ﻣﯽ ﺷـﻮد. ﺑـﺎ وﺟـﻮد ﮐـﺎرآﯾﯽ اﯾـﺪه ي ﺷﺒﮑﻪ ﻫﺎي ﮐﭙﺴﻮﻟﯽ ﺑﺮ روي ﻣﺠﻤﻮﻋﻪ داده ﻫﺎي ﺳﺎده، ﻋﻤﻠﮑﺮد اﯾﻦ ﺷﺒﮑﻪ ﻫﺎ ﺑﺮ روي داده ﻫﺎي ﭘﯿﭽﯿﺪه ﻫﻨﻮز در اﺑﻬﺎم اﺳـﺖ. در اﯾـﻦ ﻣﻘﺎﻟـﻪ ﻋﻤﻠﮑﺮد اﯾﻦ ﺷﺒﮑﻪ ﺑﺮ روي ﻣﺠﻤﻮﻋﻪ داده ي ﭘﯿﭽﯿﺪه ي ﺳﺮﻃﺎن ﭘﻮﺳﺖ ﻣـﻮرد ﺑﺮرﺳـﯽ ﻗـﺮار ﮔﺮﻓﺘـﻪ اﺳـﺖ ﮐـﻪ ﺑـﻪ دﻟﯿـﻞ اﻫﻤﯿـﺖ ﺗﺸـﺨﯿﺺ ﺿﺎﯾﻌﺎت ﭘﻮﺳﺘﯽ در ﭘﺰﺷﮑﯽ، ﭘﯿﭽﯿﺪﮔﯽ ﺗﺼﺎوﯾﺮ، ﺗﻌﺪاد زﯾﺎد آﻧﻬﺎ و ﻧﺎﻣﺘﻌﺎدل ﺑﻮدن دﺳﺘﻪ ﻫﺎ در آن اﻧﺘﺨﺎب ﺷـﺪه اﺳـﺖ. ﺑـﺮاي اﺳـﺘﺨﺮاج ﺑﻬﺘﺮ ﺗﻨﻮع ﻣﻮﺟﻮد در ﺿﺎﯾﻌﺎت ﭘﻮﺳﺘﯽ، ﺗﻐﯿﯿﺮاﺗﯽ در ﻻﯾﻪ ﻫﺎي اﺑﺘـﺪاﯾﯽ ﺷـﺒﮑﻪ داده ﺷـﺪ و ﺑـﻪ دﻟﯿـﻞ ﻋـﺪم ﺗـﻮازن در ﻣﺠﻤﻮﻋـﻪ داده ي ذﮐـﺮ ﺷﺪه، ﺗﻐﯿﯿﺮاﺗﯽ در ﺗﺎﺑﻊ ﻫﺰﯾﻨﻪ ي ﺷﺒﮑﻪ اﻋﻤﺎل ﺷﺪ. ﺗﺄﺛﯿﺮ اﺳﺘﻔﺎده از ﺗﻮاﺑﻊ ﻓﻌﺎل ﺳﺎزي ﻣﺨﺘﻠﻒ در ﺷـﺒﮑﻪ ﻧﯿـﺰ ﻣـﻮرد ﺑﺮرﺳـﯽ ﻗـﺮار ﮔﺮﻓـﺖ. ﻧﺘﺎﯾﺞ ﺑﻪ دﺳﺖ آﻣﺪه ﻧﺸﺎن ﻣﯽ دﻫﺪ اﯾﺪه ي ﺷﺒﮑﻪ ﮐﭙﺴﻮﻟﯽ ﺑﺎ اﻧﺠﺎم ﺗﻨﻈﯿﻤﺎت ﻣﺘﻨﺎﺳﺐ ﻣﯽ ﺗﻮاﻧﺪ ﺑﺮ روي ﻣﺠﻤﻮﻋﻪ داده ﻫﺎي ﭘﯿﭽﯿﺪه ﻧﯿـﺰ ﺑـﻪ ﻧﺤﻮ ﻣﻄﻠﻮﺑﯽ ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﯿﺮد.

Deep networks are a type of learning method that can model high-level relationships in data. One of the most widely used types of deep models are convolutional networks that are able to model spatial dependencies in images using convolutional layers, but do not consider the hierarchical spatial structures within the image. Capsule networks are one of the new ideas proposed for modeling the hierarchical structure of features in the image, which use grouped capsules or neurons with a dynamic routing algorithm. Despite the effectiveness of the idea of ​​capsule networks on simple data sets, the performance of these networks on complex data is still unclear. In this paper, the performance of this network is examined on a complex skin cancer dataset, which has been selected due to the importance of skin lesions diagnosis in medicine, the complexity and huge number of images and the imbalance of categories. In order to better extract the diversity of skin lesions, changes were made in the initial layers of the network. Also, due to the imbalance in the mentioned data set, changes were made in the cost function of the network. The effect of using different activation functions in the network was also investigated. The results show that the idea of ​​a capsule network can be used optimally on complex data sets by making appropriate adjustments