Algorithm for Symmetric Additional Two-Dimensional Delineation of Computer X-Ray Images

No Thumbnail Available
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Samarkand branch of TUIT
Abstract
Description
The COVID-19 epidemic spread to all corners of the world, resulting in numerous infections and deaths. This research proposes a symmetric additional two-dimensional classification framework based on three main modules: the preprocessing module for weakly supervised segmentation (O-WSSPM), the asymmetric two-dimensional module (S-CBM), and the Fuzzy C-Means clustering visualization module (FCMM). The first module, O-WSSPM, extracts additional features from CT images to create a new Data1-Seg dataset, primarily focusing on preserving essential feature areas. The second module, S-CBM, utilizes two asymmetric networks to separate various features and obtain additional functionalities. The third module, FCMM, allows the visualization of lesions in non-contrast images. While the data volume is low, five-fold cross-validation is employed to improve diversity. The proposed network shows an average classification accuracy of 85.3%, demonstrating its superior performance when compared to the baseline six-category classification model.
COVID-19 epidemiyasi dunyoning barcha burchaklariga tarqaldi, natijada son-sanoqsiz infektsiyalar va oʻlimlar kuzatildi. Ushbu tadqiqotda uchta asosiy modulga boʻlingan zaif nazorat qilinadigan simmetrik qoʻshimcha ikki chiziqli tasniflash tarmogʻi taklif etiladi: optimallashtirishni qidiradigan zaif nazorat qilinadigan segmentatsiyani oldindan ishlov berish moduli (O-WSSPM), nosimmetrik qoʻshimcha ikki chiziqli modul (S-CBM) va FCM. Klaster vizualizatsiya moduli (FCMM). Birinchi modul, O-WSSPM, yangi Data1-Seg ma’lumotlar toʻplamini yaratish uchun KT tasvirlaridan ortiqcha fon xususiyatlarini olib tashlaydi, bu asosan asosiy xususiyat sohalarini saqlashga xizmat qiladi. Ikkinchi modul S-CBM asosan turli xil xususiyatlarni ajratib olish va shu bilan boy qoʻshimcha funktsiyalarni olish uchun nosimmetrik ikkita tarmoqdan foydalanadi. Uchinchi modul FCMM yorliqsiz tasvirlarda lezyonlarni vizualizatsiya qilish imkonini beradi. Ma’lumotlar hajmi past boʻlsa, namunalar xilma-xilligini yaxshilash uchun besh tomonlama ma’lumotlarni yaxshilash amalga oshiriladi. Besh marta oʻzaro tekshirish tajribasi oʻrtacha 85,3% tasniflash aniqligini koʻrsatadi va oltita ilgʻor tasniflash modeli bilan taqqoslashni biz taklif qilayotgan tarmoq yaxshiroq ishlashga ega ekanligini koʻrsatadi.
Keywords
COVID-19, deep learning, classification, weak supervision, segmentation, additional two-dimensional, FCM, COVID-19, chuqur oʻrganish, tasniflash, zaif nazorat, segmentatsiya, qoʻshimcha ikki chiziqli, FCM
Citation