Uhlolo lweMizimba yeDijithali: Indlela i-AI Eguqula ngayo Ukufumanisa Ngexesha

Ubuhlakani bokwenziwa obufakwe ekuhloleni izinwele busekelwe emkhakheni wokufunda kwemishini obizwa ngokuthi ukufunda okujulile, ikakhulukazi amanethiwekhi e-neural convolutional aqeqeshwe ukuze aqonde amaphethini ezithombeni. Lezi zinhlelo zifunda ngokucubungula izinkulungwane zezithombe ezibhalwe phansi ezihlobene nezokwelapha, zithuthukisa ikhono lokuhlonza izici ezibonakalayo ezihlobene nezimo ezahlukene zezinwele. Inqubo yokuqeqesha ibandakanya ukufaka izithombe ezitholwe ngodokotela bezinwele abanolwazi futhi ziqinisekiswe nge-biopsy lapho kudingeka.

Ngemva kwezikhathi eziningi, inethiwekhi ifunda ukuhlole amaphethini anembile ekusatshenzisweni kwemibala, umumo, izici zomngcele, izakhiwo, nezinye izici ezibonakalayo ezihambisana nezokwelapha ezithile. Umphumela uwuhlelo olungahlola isithombe esisha futhi lunikeze ukuhlola amathuba ezimo ezahlukene. Izinhlelo zokuhlola izinwele ze-AI zanamuhla azisebenzisi kuphela uhlelo lwe-ABCDE rule ngokuqondile.

Zihlonza amaphethini aphakeme kakhulu futhi anembile kunezindlela ezisebenzayo ezisekelwe ohlwini. Ezinye izici i-AI eyinika isisindo esikhulu zingase zingabi namagama ezokwelapha ngoba zimelela ubudlelwano bezibalo phakathi kwezinga le-pixel elinencazelo ngempela kodwa ezingabonakali kumuntu. Izinhlelo ze-AI zamanje zokuhlola izilonda zezinwele zifinyelele ezingeni lokunembile elihambisana nodokotela bezinwele abaqinisekisiwe ezimeni zocwaningo ezilawulwayo, nakuba ukusebenza kwangempela kubandakanya ezinye izinto ezithinta ukwethembeka.!!

Lezi zinhlelo zisebenza kahle ekucubunguleni inani elikhulu lezithombe ngokushesha nangokuphindaphindiwe, ngaphandle kokukhathala noma ukucabanga okungafanele okungathinta abahloli bomuntu. Nokho, ukuqonda ukuthi i-AI yokuhlola isebenza kanjani kusho futhi ukuqonda imikhawulo yayo. Lezi zinhlelo zikhulula ukuhlola amathuba, hhayi izinqumo eziphelele.

Ziqeqeshwe kumadathasethi athile angase angamehluko wonke ombala wezitho, izinhlobo zezilonda, noma izimo zokuthwebula izithombe ngokulinganayo. Ikhwalithi yesithombe esifakiwe ithinta kakhulu ikhwalithi yokuhlola okukhishwayo, okwenza proper photography technique kubalulekile.

Ngaphakathi kwezikhungo ze-dermatology, i-AI ifakwa njengensiza yokweseka izinqumo ethuthukisa esikhundleni sokushintsha ukuhlola komdokotela wezokwelapha. Lezi zinhlelo zokwelapha ngokuvamile zisebenza ngezithombe ze-dermoscopic ezithathwe ngempahla yokwelapha evamile, zinikeza izinga lekhwalithi yesithombe nokuhambisana okusekela ukunembile okuphezulu. Izinhlelo eziningi ze-AI zokwelapha zithole imvume yokulawula ukuze zisetshenziswe njengamathuluzi okuxilonga.

Lezi zinhlelo zihlola izithombe ze-dermoscopic ngesikhathi sangempela ngesikhathi sokuhlola isiguli futhi ziqhakambisa izilonda ezikhombisa izici ezihlobene ne-malignancy. Umdokotela wezokwelapha ubona ukuhlola kwe-AI eduze kokuhlola kwakhe, kudala ukuhlola okubili okungase kubonise okutholakele okungase kube khona kokubili umuntu noma umshini angakuphuthelwa yedwa. Izinkundla zokuthwebula umzimba wonke ezithuthukiswe yi-AI zimelela enye indlela yokwelapha.

Lezi zinhlelo zithwebula ubuso bonke bezinwele zomguli zisebenzisa isethi ye-multi-camera evamile, bese zisebenzisa ama-algorithms ukuze zihlukanise, zifake, futhi ziqhube zonke izilonda ezibonakalayo. Ezivakashi zokulandela, uhlelo luhlola ngokuzenzakalelayo izithombe ezintsha uma kuqhathaniswa ne-benchmark, kugqamisa izilonda ezishintshile ngosayizi, umumo, noma umbala futhi ziqhakambisa izilonda ezintsha ezingazange zibe khona ngesikhathi sokuvakasha kwangaphambilini. Lokhu kuhlola okuzenzakalelayo kwezinguquko kubhekana nenkinga enkulu kakhulu ku-dermatology: ukuhlole ngokwethembeka izinguquko ezincane phakathi kwezinkulungwane ze-moles kumguli oyedwa.

Ngisho nomdokotela wezokwelapha onolwazi oluningi akakwazi ukukhumbula ukubukeka okunembile kwe-mole ngayinye phakathi kwezivakashi zonyaka, okwenza ukuhlola okuqondisiwe kwe-AI kube ukuthuthukiswa okwenziwe ngempela emandleni okwelapha. Ucwaningo luqhubeka lwandisa izicelo ze-AI ku-dermatology ngaphezu kokuhlonza i-melanoma. Ama-algorithms athuthukiswa futhi athuthukiswa ukuze ahlonze i-basal cell carcinoma, i-squamous cell carcinoma, ama-actinic keratoses, nezinye izimo ezithile ezithokozisayo. Ezinye izinhlelo zihlola izikhalazo zezinwele ezindaweni zokunakekelwa okujwayelekile, zisiza odokotela abajwayelekile ukuthi bazi ukuthi yiziphi iziguli ezidinga ukudluliselwa okuphuthumayo ku-dermatology nokuthi yiziphi ezingaqhubeka zihlolwe ngokuphepha.

Ukuhlola izinwele kwe-AI okuhlangabezana nomthengi okwenziwa ngezinhlelo zokusebenza ze-smartphone kube nokufinyelela kwezobuchwepheshe obusekela ukuhlola kubantu abanokuxhumana kwe-inthanethi. Lezi zinhlelo zokusebenza zivumela abasebenzisi ukuthi bathwebule isithombe sezilonda zezinwele futhi bathole ukuhlola okuveliswe yi-AI phakathi kwemizuzwana, kwenza ukufinyelela kuhlola okuqala kube lula. Ukuthandwa kukhulu.

Akukho muntu onokufinyelela kalula kumdokotela wezokwelapha, kungaba ngenxa yokuhlukaniswa kwezindawo, izindleko, isikhathi sokulinda, noma ukuqwashisa. Uhlelo lokusebenza lwe-smartphone olungakwazi ukuveza izilonda ezingase zibe yingozi futhi lukhuthaze abasebenzisi ukuba bafune ukuhlolwa kochwepheshe lugcwalisa isikhala esibalulekile, ikakhulukazi ezindaweni ezingenamakhaya lapho odokotela bezokwelapha behluleka khona. Nokho, ukuhlola okusekelwe ku-smartphone kuza nezixwayiso ezinkulu.

Ikhwalithi yesithombe esivela kumakhamera wefoni yomthengi iyahluka kakhulu kuye ngedivayisi, izimo zokukhanya, ukuhamba kwesandla, kanye nokukhalipha ukusuka ezilondeni. Izinhlelo ze-AI eziqeqeshwe ngezithombe ze-dermoscopic ezisezingeni eliphezulu zingase zisebenze ngendlela ehlukile uma zihlola izithombe ezithathwe ngama-smartphones omthengi ngaphansi kwezimo zangempela ezihlukahlukene.!! Abasebenzisi kumele baqonde ukuthi lezi zinsiza zinikeza ukuhlola kokuqala, hhayi izinqumo.

Izifundo zokunembile eziphathelene nezinhlelo zokuhlola izinwele zomthengi zikhombise imiphumela ehlukene. Ezinye izinhlelo zisebenza kahle ezihlolweni ezilawulwayo, kanti ezinye zikhombise amazinga akhathazayo okungamanga lapho izilonda ezingase zibe yingozi zihlukaniswe njengokungabazeki. Ukuhlolwa kokulawula lezi zinhlelo kuhluka ngokwezwe, futhi hhayi zonke izinhlelo zokusebenza ezithengiswayo zidlule ekuhlolweni okuqinile kwezokwelapha.

Izinhlelo zokusebenza ze-smartphone zokuhlola izinwele eziphathekayo zikhombisa ukuziphatha kahle njengamathuluzi okuhlola asiza abasebenzisi ukuthi bazi ukuthi kufanele yini bafune ukuhlolwa kochwepheshe, hhayi njengokuphikisana nezokwelapha. Zikhuthaza abasebenzisi abathola ukuhlola okukhathazayo ukuba bahlele izikhathi zokuvakasha kochwepheshe, futhi eziningi zifaka okuqukethwe kwemfundo mayelana nokuzihlola nokukhumbuza.

Umthetho obaluleke kakhulu ekuqondeni indima ye-AI empilweni yezinwele ukuthi lezi zinto zenzelwe ukuhlanganisa ubuchwepheshe bomuntu esikhundleni sokubushintsha. Lo mqondo awuwona nje umqondo omuhle kodwa ubonisa imikhawulo yezezimali nezenzo eyenza ukuqapha komuntu kube kubalulekile. Izinhlelo ze-AI azinakho ukukwazi ukuthatha umlando wesiguli, ubuze ngokuqhubeka kwezimpawu, ubuze ngembali yomndeni, uthinte isilonda ukuze uhlole umumo nokuqina, noma uhlanganise yonke imininingwane yokwelapha ethinta ukuhlola komdokotela wezokwelapha.

Isithombe sithwebula kuphela ulwazi olubonakalayo, kanti ukuhlolwa okuphelele kwezokwelapha kuhlanganisa ulwazi lokuthinta, umlando, nolwazi lwezinhlelo okungatholakali kunoma iyiphi i-AI esekelwe ezithombeni. Odokotela bezokwelapha bring contextual reasoning that current AI systems cannot replicate. Baqonda ukuthi isilonda esisodwa esitholakala emathangeni omuntu omdala sinezincazelo ezihlukile kunezilonda ezifanayo ezitholakala emzimbeni wentsha.

Bazi uma umlando wemithi wesiguli, isimo sokuvikela, noma imvelaphi ye-genetic ishintsha amathuba ezixazululo ezahlukene. Babona izimpawu zokulimala kwelanga, amaphethini okukhula, nezinye izinkomba ezibonakalayo ezibonakala phezu kwezinwele ezibanzi. Imodeli efanele ibeka i-AI njengomsizi onekhono ophakamisayo ukunakekelwa kwezokwelapha ezingeni eliningi.

Kubaguli, amathuluzi okuzihlola asekelwe ku-AI akhuphula ukuqwashisa futhi anikeze ukuhlola kokuqala phakathi kwezivakashi zokuqonda. Kubodokotela bezokwelapha, izinhlelo ze-AI zokuhlola zisiza ukuthola iziguli ezidinga ukudluliselwa okuphuthumayo. Kubodokotela bezokwelapha, izinhlelo zokweseka izinqumo ze-AI zinikeza umbono wesibili ongase ubone okutholakele okuncane futhi uthuthukise ukuhambisana kokuxilonga.

Le modeli yokubambisana ibonise imiphumela emihle ezifundweni lapho ukuhlanganiswa kodokotela bezokwelapha ne-AI kudlula kokubili odokotela bezokwelapha bodwa noma i-AI yodwa. Amandla ahambisanayo okucabanga komuntu nezokuhlola amaphethini e-algorithm akha uhlelo lokuhlola olunamandla kakhulu kunokuthi noma iyiphi indlela esebenza ngokuzimela.

Umkhondo we-AI empilweni yezinwele ukhomba ezinhlelweni ezihlanganisiwe, ezitholakalayo, nezinekhono ezizoshintsha indlela izimo zezinwele ezitholakala futhi zihlolwe phakathi kweminyaka ezayo. Ukuthuthuka okuningi okukhulayo kukhombisa indlela le teknoloji ethuthuka ngayo. Izinhlelo ze-AI ezihlanganisa hhayi kuphela izithombe kodwa futhi nezimpawu ezibikwe ngababhalisi, umlando wezokwelapha, izici zokulawula, nolwazi lwezimo zendawo zisebenza.

Lezi zinhlelo eziphelele zifuna ukuhlinzeka ngokuqonda okujwayelekile okuphakeme kokuhlolwa okwenziwa udokotela wezokwelapha, kuhlanganisa imithombo eminingi yedatha esikhundleni sokuncika kuphela ekuhloleni okubonakalayo. Izindlela zokufunda ezihlangene zingase zixazulule enye yezinkinga zamanje zokuhlola izinwele ze-AI: ubias bedathasethi. Ngokufundisa ama-algorithms phakathi kwabantu abahlukahlukene ngaphandle kokuhlanganisa izithombe ezibucayi zezokwelapha, ukufunda okuhlangene kungakhiqiza amamodeli anembile kakhulu asebenza ngokwethembeka kuwo wonke umbala wezitho nezinhlobo.

Izinhlelo zamanje ngokuvamile zisebenza kahle kakhulu ezithombeni ezimnyama ngenxa yedatha yokufundisa ethinta abantu abamhlophe, futhi ukuxazulula le miphumela kuyinto ebalulekile yocwaningo. Ubuchwepheshe obugqokekayo bungase bukhuthaze ukuqapha okuqhubekayo noma okungapheli kwezinwele esikhathini esizayo. Izinsiza ezinekhono lokuthwebula zingase zibe nezithombe ezithile futhi zihlola izindawo zezinwele, ziqaphele izinguquko ngokuzenzakalelayo futhi zixwayise abasebenzisi ngezilonda ezidinga ukunakwa.

Lokhu kushintsha kusuka ekuhloleni okujwayelekile kuya ekuhloleni okuqhubekayo kungase kuhlonze izinguquko eziphakeme kakhulu. Ukuhlanganiswa nezincwadi zezempilo ezikwi-inthanethi kuzovumela imiphumela yokuhlola izinwele ze-AI ukuba ibe yingxenye yomlando wezokwelapha wesiguli, itholakala kubo bonke abahlinzeki bezempilo. Le nhlangano izovumela ukunakekelwa okuxhumanayo nokuhlonza izitayela zempilo yezinwele phakathi kokuphila kwesiguli.

Izinhlelo zokulawula ziyaqhubeka zishintsha ukuze zihambisane nobuchwepheshe. Njengoba amathuluzi wokuhlola izinwele ze-AI ethuthuka futhi esetshenziswa kabanzi, izindinganiso ezicacile zokuhlola, ukuhlinzeka ngokuqonda, nokubekwa kwezokwelapha kuzokwenza kokubili abathengi nabahlinzeki bezempilo baqonde indima efanele yalezi zinsiza ngaphakathi kwesistimu yokunakekelwa kwezempilo ebanzi.