Ubhubhani we-COVID-19 wathatha umhlaba ngoqhwithela ekuqaleni kuka-2020 kwaye ukusukela ngoko uye waba ngoyena nobangela uphambili wokufa kumazwe aliqela, kubandakanya iTshayina, i-USA, iSpain kunye ne-United Kingdom. Abaphandi basebenza kakhulu ekuphuhliseni iindlela ezisebenzayo zokuxilonga usulelo lwe-COVID-19, kwaye uninzi lwabo lujolise kwingqwalasela yabo kwindlela ubukrelekrele bokwenziwa (AI) obunokuxhaswa ngayo ngale njongo.
Izifundo ezininzi ziye zaxela ukuba iinkqubo ezisekwe kwi-AI zinokusetyenziselwa ukubona i-COVID-19 kwimifanekiso ye-X-ray esifubeni kuba esi sifo sithande ukuvelisa iindawo ezinobovu kunye namanzi emiphungeni, abonakala njengamabala amhlophe kwi-X-ray scan. . Nangona iimodeli ezahlukeneyo zokuxilonga ze-AI ezisekelwe kulo mgaqo ziye zacetywa, ukuphuculwa kokuchaneka, isantya, kunye nokusebenza kwazo kuhlala kuseyona nto iphambili.
Ngoku, iqela lezazinzulu elikhokelwa nguNjingalwazi uGwanggil Jeon weYunivesithi yeSizwe yase-Incheon, eKorea, liphuhlise isakhelo esizenzekelayo se-COVID-19 esiguqula izinto zibe yinotshi ngokudibanisa iindlela ezimbini ezinamandla ezisekwe kwi-AI. Inkqubo yabo inokuqeqeshwa ukuba yahlule ngokuchanekileyo phakathi kwemifanekiso yeX-ray yesifuba yezigulana ze-COVID-19 kwezingeyiyo i-COVID-19. Iphepha labo lenziwa lafumaneka kwi-intanethi nge-27 ka-Okthobha ngo-2021, laza lapapashwa nge-21 kaNovemba ka-2021, kuMqulu 8, uKhupho 21 lwe-IEEE Internet of Things Journal.
Ii-algorithms ezimbini ezisetyenziswe ngabaphandi zi-Faster R-CNN kunye ne-ResNet-101. Eyokuqala imodeli esekelwe kumatshini wokufunda osebenzisa inethiwekhi yesiphakamiso sommandla, enokuqeqeshwa ukuchonga imimandla echaphazelekayo kumfanekiso wegalelo. Eyesibini yinethiwekhi ye-neural yokufunda nzulu equka i-101 layers, eyayisetyenziswa njengomqolo. I-ResNet-101, xa iqeqeshwe ngedatha eyaneleyo yokufaka, imodeli enamandla yokuqaphela umfanekiso. UProf. ukuchaneka okumangalisayo kwe-101%.
Iqela lophando likholelwa ukuba isicwangciso salo sinokungqina ukuba luncedo ekubonweni kwangoko kwe-COVID-19 ezibhedlele nakumaziko ezempilo oluntu. Ukusebenzisa iindlela zokuxilonga ngokuzenzekelayo ezisekwe kwitekhnoloji ye-AI kunokuthatha umsebenzi kunye noxinzelelo kwiingcali zemitha kunye nezinye iingcali zonyango, ebezijongene nomthwalo omkhulu okoko ubhubhane waqalayo. Ngaphezu koko, njengoko izixhobo zonyango zanamhlanje ziqhagamshelwa kwi-Intanethi, kuya kwenzeka ukondla inani elikhulu ledatha yoqeqesho kwimodeli ecetywayo; oku kuya kubangela ukuchaneka okuphezulu ngakumbi, kwaye kungekuphela nje kwi-COVID-19, njengoko uNjingalwazi Jeon esitsho: "Indlela yokufunda enzulu esetyenziswa kuphononongo lwethu iyasebenza kwezinye iintlobo zemifanekiso yezonyango kwaye inokusetyenziswa ukuxilonga izifo ezahlukeneyo."
INTO ONOKUYITHATHA KWELI NQAKU:
- Several studies have reported that AI-based systems can be used to detect COVID-19 in chest X-ray images because the disease tends to produce areas with pus and water in the lungs, which show up as white spots in the X-ray scans.
- The first one is a machine learning-based model that uses a region-proposal network, which can be trained to identify the relevant regions in an input image.
- Moreover, as more modern medical devices become connected to the Internet, it will be possible to feed vast amounts of training data to the proposed model.
