iindaba

Ukususela ekubeni i-IBM Watson yaqala ngo-2007, abantu baye baqhubeka belandela uphuhliso lwezobukrelekrele bezonyango (AI). Inkqubo ye-AI yonyango esebenzisekayo nenamandla inamandla amakhulu okubumba kwakhona yonke imiba yamayeza anamhlanje, eyenza ukuba ukhathalelo olukrelekrele, oluchanekileyo, olusebenzayo, nolubandakanyayo, izisa impilo kubasebenzi bezonyango kunye nezigulana, kwaye ngaloo ndlela iphucula kakhulu impilo yabantu. Kwiminyaka eyi-16 edlulileyo, nangona abaphandi be-AI bezonyango beqokelele kwiinkalo ezincinci ezahlukeneyo, ngeli nqanaba, abakakwazi ukuzisa intsomi yesayensi.

Kulo nyaka, ngophuhliso loguqulo lwetekhnoloji ye-AI efana ne-ChatGPT, i-AI yonyango yenze inkqubela phambili enkulu kwiinkalo ezininzi. Impumelelo engazange ibonwe ngaphambili kwisakhono se-AI yonyango: Ijenali yeNdalo iye yaqalisa ngokuqhubekayo uphando lwemodeli yolwimi olukhulu lwezonyango kunye nemodeli yonyango olusisiseko; UGoogle ukhupha iMed-PaLM kunye nendlalifa yayo, ifikelele kwinqanaba lobuchwephesha kwimibuzo yeemviwo yeGqirha lezonyango lase-US. Iijenali ezinkulu zemfundo ziya kugxila kwi-AI yonyango: Indalo ikhupha imbono kwimodeli esisiseko ye-AI yonyango jikelele; Ukulandela uchungechunge lokuphononongwa kwe-AI kwiMedicine ekuqaleni kwalo nyaka, i-New England Journal of Medicine (NEJM) yapapasha uphononongo lwayo lokuqala lwezempilo lwedijithali ngoNovemba 30, kwaye yaqalisa ukukhutshwa kokuqala kwe-NEJM sub-journal NEJM AI ngoDisemba 12. Umhlaba wokuhlalisa i-AI yezobugqirha ukhula ngakumbi: I-JAMA sub-journal yapapasha inkqubo yokwabelana ngedatha yomfanekiso wezonyango jikelele; I-US Food and Drug Administration (FDA) iphuhlisa izikhokelo eziqulunqiweyo zokulawulwa kwe-AI yonyango.

Apha ngezantsi, sijonga inkqubela ebalulekileyo eyenziwe ngabaphandi kwihlabathi liphela kwicala le-AI yonyango esebenzayo ngo-2023.

Imodeli esisiseko ye-AI yonyango

Ukwakhiwa kwemodeli esisiseko ye-AI yonyango ngokungathandabuzekiyo lolona gxininiso lophando lushushu kulo nyaka. Iijenali zeNdalo zipapashe amanqaku ophononongo malunga nemodeli yeSiseko yeHlabathi yokhathalelo lwempilo kunye nemodeli yolwimi olukhulu lwenkathalo yezempilo apha enyakeni. Uhlalutyo lweMifanekiso yezoNyango, iphephancwadi eliphambili kwishishini, lihlaziywe kwaye lijonge phambili kwimingeni kunye namathuba ophando olusisiseko lwemodeli ekuhlalutyweni komfanekiso wezonyango, kwaye licebise ingcamango "yomnombo wemodeli esisiseko" ukushwankathela kunye nokukhokela ukuphuhliswa komzekelo wophando olusisiseko lwe-AI yonyango. Ikamva leemodeli ezisisiseko ze-AI kukhathalelo lwempilo liya licaca ngakumbi. Ukuzoba kwimizekelo ephumeleleyo yeemodeli ezinkulu zeelwimi ezifana ne-ChatGPT, usebenzisa iindlela zokuqeqesha kwangaphambili ezizilawulayo kunye nokuqokelela okukhulu kwedatha yoqeqesho, abaphandi kwinkalo ye-AI yonyango bazama ukwakha i-1) imifuziselo yesiseko yesifo, i-2) imodeli yesiseko ngokubanzi, kunye ne-3) imodeli enkulu ye-multimodal edibanisa uluhlu olubanzi lweemodemu zeparameter kunye ne-parimeter enkulu

Imodeli ye-AI yokuFumana idatha yezoNyango

Ukongeza kwiimodeli ezinkulu ze-AI ezidlala indima enkulu kwimisebenzi yohlalutyo lwedatha yekliniki esezantsi, ekufumaneni idatha yeklinikhi ephezulu, iteknoloji emelwe yimizekelo ye-AI yokuvelisa nayo iye yavela. Inkqubo, isantya, kunye nomgangatho wokufunyanwa kwedatha kunokuphuculwa kakhulu nge-algorithms ye-AI.

Ekuqaleni kwalo nyaka, i-Nature Biomedical Engineering yapapasha uphando oluvela kwiYunivesithi yaseTurkey yaseStraits egxile ekusebenziseni i-AI yokuvelisa ukusombulula ingxaki yokuxilongwa kwe-pathologic-incediswa ngumfanekiso kwizicelo zeklinikhi. I-Artifacts kwiithishu zecandelo elinomkhenkce ngexesha lotyando zingumqobo kuvavanyo olukhawulezayo lokuxilonga. Nangona i-formalin kunye neparafini efakwe kwi-tissue (FFPE) ibonelela ngesampulu ekumgangatho ophezulu, inkqubo yayo yokuvelisa ithatha ixesha kwaye ihlala ithatha iiyure eziyi-12-48, iyenza ingafanelanga ukusetyenziswa kutyando. Iqela lophando ke ngoko licebise i-algorithm ebizwa ngokuba yi-AI-FFPE, enokwenza ukubonakala kwezicubu kwicandelo elinomkhenkce elifana ne-FFPE. I-algorithm ilungise ngempumelelo i-artifacts yamacandelo afriziwe, iphucule umgangatho womfanekiso, kwaye igcine iimpawu ezifanelekileyo zeklinikhi ngexesha elinye. Ekuqinisekisweni kweklinikhi, i-algorithm ye-AI-FFPE iphucula kakhulu ukuchaneka kokuxilongwa kwee-pathologists kwii-tumor subtypes, ngelixa linciphisa kakhulu ixesha lokuxilongwa kweklinikhi.

IiNgxelo zeCell Medicine zibika umsebenzi wophando liqela elivela kwiKholeji yeSithathu yeKliniki yeYunivesithi yaseJilin, iSebe leRadiology, iSibhedlele saseZhongshan esiManyaniswa neYunivesithi yaseFudan, kunye neYunivesithi yaseShanghai yeSayensi kunye neTeknoloji [25]. Olu pho nonongo luphakamisa injongo enzulu yokufunda ngokubanzi kunye nesakhelo sokudibanisa ukwakhiwa ngokutsha (Hybrid DL-IR) enokuguquguquka okuphezulu kunye nokuguquguquka, ebonisa ukusebenza kakuhle kokwakhiwa kwakhona komfanekiso kwi-MRI ekhawulezayo, i-CT yedosi ephantsi, kunye ne-PET ekhawulezayo. I-algorithm inokufikelela kwi-MR Single-organ multi-sequence scanning in 100 seconds, ukunciphisa umthamo we-radiation ukuya kwi-10% kuphela yomfanekiso we-CT, kwaye uphelise ingxolo, kwaye unokwakha kwakhona izilonda ezincinci ezivela ekufumaneni i-PET kunye ne-2 ukuya kwii-4 ukukhawuleza, ngelixa unciphisa umphumo wezinto ezihambayo.

I-AI yezoNyango ngokuBambisana nabasebenzi bezoNyango

Uphuhliso olukhawulezayo lwe-AI yonyango luye lwakhokelela kwiingcali zonyango ukuba zicinge nzulu kwaye zihlolisise indlela yokusebenzisana ne-AI ukuphucula iinkqubo zeklinikhi. NgoJulayi kulo nyaka, i-DeepMind kunye neqela lophando lamaziko amaninzi licebise ngokudibeneyo inkqubo ye-AI ebizwa ngokuba yi-Complementary Driven Clinical Workflow Delay (CoDoC) . Inkqubo yokuxilonga iqala ukufunyaniswa yinkqubo ye-AI eqikelelweyo, emva koko igwetywe ngenye inkqubo ye-AI kwisiphumo sangaphambili, kwaye ukuba kukho ukungathandabuzeki, ukuxilongwa ekugqibeleni kwenziwa ngugqirha ukuze kuphuculwe ukuchaneka kokuxilongwa kunye nokusebenza kakuhle kokulinganisela. Xa kuziwa ekuhlolweni komhlaza webele, i-CoDoC yanciphisa amazinga angalunganga ngama-25% kunye nezinga elifanayo elibi lobuxoki, ngelixa linciphisa umsebenzi wonyango ngama-66%, xa kuthelekiswa nenkqubo yangoku "yokulamla kabini" e-UK. Ngokumalunga nokuhlelwa kwe-TB, iireyithi ezintle zobuxoki zancitshiswa nge-5 ukuya kwi-15 ipesenti kunye nezinga elifanayo elibi lobuxoki xa kuthelekiswa ne-AI ezimeleyo kunye nokuhamba komsebenzi wekliniki.

Ngokufanayo, u-Annie Y. Ng et al., weNkampani yeKheiron eLondon, e-UK, wazisa abafundi be-AI abongezelelweyo (ngentsebenziswano nabavavanyi babantu) ukuba baphinde bahlolisise iziphumo xa kungekho ziphumo zokukhumbula kwinkqubo yolamlo ephindwe kabini, eyaphucula ingxaki yokufunyaniswa okuphosiweyo ekuqaleni kokuhlolwa komhlaza webele, kwaye inkqubo yayiphantse ingabikho ziphumo zobuxoki. Olunye uphando, olukhokelwa liqela kwiYunivesithi yaseTexas McGovern Medical School kwaye lugqitywe kumaziko amane e-stroke, kusetyenziswa i-computed tomography angiography (CTA) -based based AI iteknoloji ukuzenzekelayo ukufunyanwa kwe-vascular occlusive ischemic stroke (LVO). Oogqirha kunye neengcali zeradiologists bafumana izilumkiso zexesha lokwenyani kwiifowuni zabo eziphathwayo kwimizuzu nje yokugqitywa komfanekiso we-CT, ukwazisa ngobukho obunokwenzeka be-LVO. Le nkqubo ye-AI iphucula ukuhamba komsebenzi we-hospital for acute ischemic stroke, ukunciphisa ixesha lokungena emnyango ukusuka ekwamkelweni ukuya kunyango kunye nokubonelela ngamathuba okuhlangula ngempumelelo. Iziphumo zipapashwa kwi-JAMA Neurology.

Imodeli yokhathalelo lwempilo ye-AI yeNzuzo yeHlabathi

I-2023 iya kubona nomsebenzi omningi olungileyo osebenzisa i-AI yonyango ukufumana iimpawu ezingabonakaliyo kwiliso lomntu ukusuka kwidatha efumaneka ngokulula, eyenza ukuxilongwa kwendawo yonke kunye nokuhlolwa kwangaphambili kwinqanaba. Ekuqaleni konyaka, i-Nature Medicine yapapasha izifundo ezenziwe yi-Zhongshan Eye Centre yeYunivesithi yase-Sun Yat-sen kunye neSibhedlele esiBambiseneyo seSibini saseFujian Medical University. Besebenzisa ii-smartphones njengeetheminali zesicelo, basebenzise imifanekiso enjengeekhathuni zevidiyo ukukhuthaza abantwana ukuba bajonge kwaye barekhode indlela abajonga ngayo abantwana kunye neempawu zobuso, kwaye bahlalutya ngakumbi iimodeli ezingaqhelekanga zisebenzisa iimodeli zokufunda nzulu ukuchonga ngempumelelo izifo zamehlo ezili-16, kubandakanya i-cataracts yokuzalwa, i-ptosis yokuzalwa kunye ne-glaucoma yokuzalwa, ngokuchaneka komndilili ongaphezulu kwe-85%. Oku kubonelela ngendlela esebenzayo nekulula ukwenza yaziwe iindlela zobugcisa zovavanyo olukhulu kwangethuba lokonakala kokubonakala kweemveku kunye nezifo zamehlo ezinxulumeneyo.

Ekupheleni konyaka, i-Nature Medicine yabika umsebenzi owenziwe ngamaziko angaphezu kwe-10 ezonyango kunye nophando emhlabeni jikelele, kuquka ne-Shanghai Institute of Pancreatic Disease kunye neSibhedlele sokuqala esiBambiseneyo seYunivesithi yaseZhejiang. Umbhali wasebenzisa i-AI ekuhlolweni komhlaza we-pancreatic wabantu abangabonakaliyo kumaziko oviwo lwenyama, izibhedlele, njl., Ukufumanisa iimpawu ze-lesion kwimifanekiso ye-CT scan enzima ukuyibona ngeso lenyama yodwa, ukuze kuphunyezwe ngokufanelekileyo kunye nokungabikho kokuhlaselwa kwangaphambili komhlaza we-pancreatic. Ekuphononongeni idatha evela kwizigulane ezingaphezu kwe-20,000, imodeli iphinde ichonge iimeko ze-31 zezilonda eziphosiweyo zonyango, eziphucule kakhulu iziphumo zekliniki.

Ukwabelana ngeDatha yezoNyango

Ngo-2023, uninzi lweendlela ezigqibeleleyo zokwabelana ngedatha kunye neemeko eziyimpumelelo ziye zavela kwihlabathi jikelele, ziqinisekisa intsebenziswano yamaziko amaninzi kunye nokuvuleka kwedatha phantsi kwesiseko sokukhusela ubumfihlo bedatha kunye nokhuseleko.

Okokuqala, ngoncedo lweteknoloji ye-AI ngokwayo, abaphandi be-AI baye bafaka isandla ekwabelaneni ngedatha yezokwelapha. UQi Chang kunye nabanye abavela kwiYunivesithi yaseRutgers eUnited States bapapashe inqaku kwiNdawo yoNxibelelwano, ephakamisa isakhelo sokufunda sobumbano i-DSL esekelwe kuthungelwano lwe-adversarial synthetic, esebenzisa i-AI yokuvelisa ukuqeqesha idatha ethile eyenziwe yamaziko amaninzi, kwaye emva koko ithatha indawo yedatha yokwenyani yamaziko amaninzi kunye nedatha eyenziweyo. Qinisekisa uqeqesho lwe-AI olusekwe kumaziko amaninzi edatha ngelixa ukhusela ubumfihlo bedatha. Iqela elifanayo likwavula umthombo wedatha yemifanekiso eveliswayo ye-pathological kunye neenkcazo ezihambelanayo. Imodeli yokwahlula oqeqeshwe kwisethi yedatha eyenziweyo inokufezekisa iziphumo ezifanayo kwidatha yangempela.

Iqela lika-Dai Qionghai elivela kwiYunivesithi yaseTsinghua lipapashe iphepha kwi-npj yeMpilo yeDijithali, ephakamisa i-Relay Learning, esebenzisa idatha enkulu ye-multi-site ukuqeqesha iimodeli ze-AI phantsi kwesiseko solawulo lwedatha yendawo kwaye akukho uxhumano lwenethiwekhi ye-cross-site. Ilinganisa ukhuseleko lwedatha kunye neenkxalabo zabucala kunye nokusukela ukusebenza kwe-AI. Iqela elifanayo emva koko liphuhliswe ngokudibeneyo kwaye liqinisekisiwe i-CAIMEN, isifuba se-CT pan-mediastinal diagnostic system esekelwe kwi-federal learning, ngokubambisana neSibhedlele sokuQala esiManyano neSibhedlele saseGuangzhou Medical University kunye nezibhedlele ze-24 kulo lonke ilizwe. Inkqubo, enokusetyenziswa kwii-12 eziqhelekileyo ze-mediastinal tumors, iphumelele i-44,9 yepesenti yokuchaneka okungcono xa isetyenziswe yodwa kunokuba isetyenziswe ziingcali zabantu kuphela, kunye neepesenti ze-19 ezingcono zokuchaneka kokuxilongwa xa iingcali zabantu zincedwa yiyo.

Kwelinye icala, amalinge amaninzi asendleleni yokwakha iiseti zedatha yezonyango ezikhuselekileyo, zehlabathi, ezinkulu. NgoNovemba ka-2023, u-Agustina Saenz kunye nabanye abasuka kwiSebe le-Biomedical Informatics kwiSikolo sezoNyango saseHarvard bapapashwa kwi-Intanethi kwiLancet Digital Health isakhelo sehlabathi sokwabelana ngedatha yemifanekiso yezonyango ebizwa ngokuba yi-Artificial Intelligence Data for All Healthcare (MAIDA). Basebenza kunye nemibutho yezempilo emhlabeni wonke ukubonelela ngesikhokelo esibanzi malunga nokuqokelela idatha kunye nokuchongwa, usebenzisa i-template ye-US Federal Demonstration Partner (FDP) ukulinganisa ukwabelana ngedatha. Baceba ukukhulula ngokuthe ngcembe iiseti zedatha eziqokelelwe kwimimandla eyahlukeneyo kunye neSetingi zeklinikhi kwihlabathi jikelele. Idatha yokuqala kulindeleke ukuba ikhutshwe ekuqaleni kuka-2024, kunye nokunye okuza kuza njengoko intsebenziswano isanda. Le projekthi ilinge elibalulekileyo lokwakha isethi yehlabathi, enkulu kunye neentlobo ezahlukeneyo zedatha ye-AI ekhoyo esidlangalaleni.

Emva kwesi siphakamiso, i-UK Biobank ibeke umzekelo. I-UK Biobank ikhuphe idatha entsha kwi-30 kaNovemba kuyo yonke i-genome yokulandelelana kwabathathi-nxaxheba be-500,000. Uvimba weenkcukacha, opapasha ulandelelwano olupheleleyo lwegenome yevolontiya ngalinye lama-500,000 aseBritani, lolona luhlu lwedatha lwegenome lomntu likhulu ehlabathini. Abaphandi kwihlabathi liphela banokucela ukufikelela kule datha engachongiweyo kwaye bayisebenzise ukuphanda isiseko sofuzo sempilo kunye nesifo. Idatha yemfuza ibisoloko inovakalelo oluphezulu ukuze ingqinwe kwixesha elidlulileyo, kwaye le mpumelelo yembali ye-UK Biobank ingqina ukuba kunokwenzeka ukwakha isiseko sedatha esivulekileyo, esingenamfihlo sehlabathi jikelele. Ngobu buchwepheshe kunye nedathabheyisi, i-AI yonyango ibophelelekile ukuba ingenise umtsi olandelayo.

Ukuqinisekiswa kunye noVavanyo lwe-AI yezoNyango

Xa kuthelekiswa nophuhliso olukhawulezayo lwetekhnoloji ye-AI yonyango ngokwayo, uphuhliso lokuqinisekisa kunye novavanyo lwe-AI yonyango lucotha kancinci. Ukuqinisekiswa kunye novavanyo kwintsimi ye-AI ngokubanzi kaninzi ayihoyi iimfuno zangempela zeekliniki kunye nezigulane ze-AI. Ulingo lweklinikhi olulawulwa ngokungenamkhethe lusebenza nzima kakhulu ukuhambelana nokuphindaphindwa okukhawulezileyo kwezixhobo ze-AI. Ukuphucula inkqubo yokuqinisekisa kunye novavanyo olufanelekileyo kwizixhobo ze-AI zonyango ngokukhawuleza eyona nto ibaluleke kakhulu ukukhuthaza i-AI yonyango yophando lwe-leapfrog ngokwenene kunye nophuhliso ukuya kwikliniki.

Kwiphepha lophando likaGoogle nge-Med-PaLM, elipapashwe kwiNdalo, iqela likwapapashe i-MultiMedQA benchmark yovavanyo, esetyenziselwa ukuhlola amandla eemodeli zeelwimi ezinkulu zokufumana ulwazi lwezonyango. Umgangatho udibanisa iiseti zedatha ezintandathu ezikhoyo zonyango ze-Q&A, ezibandakanya ulwazi lwezonyango lobuchwephesha, uphando kunye neminye imiba, kunye nedatha yedatha yedatha yophando lwezonyango kwi-intanethi, eqwalasela ugqirha-isigulane kwi-intanethi ye-Q&A, ezama ukuqeqesha i-AI ibe ngugqirha oqeqeshiweyo kwiinkalo ezininzi. Ukongeza, iqela liphakamisa isakhelo esisekelwe kuvavanyo lomntu oluthathela ingqalelo imilinganiselo emininzi yenyani, ukuqonda, ukuqiqa, kunye nokuthambekela okunokwenzeka. Le yenye yezona nzame zophando ezimele ukuvavanya i-AI kukhathalelo lwezempilo olupapashwe kulo nyaka.

Nangona kunjalo, ngaba into yokuba iimodeli ezinkulu zeelwimi zibonisa inqanaba eliphezulu lokukhowuda kulwazi lwezonyango kuthetha ukuba iimodeli ezinkulu zolwimi zifanelekile kwimisebenzi yezonyango yehlabathi yokwenyani? Kanye njengokuba umfundi wezonyango ophumelele iimviwo zobugqirha bobuchwephesha ngamanqaku agqibeleleyo usekude nogqirha oyintloko yedwa, indlela yovavanyo ecetywe nguGoogle isenokungabi yimpendulo egqibeleleyo kwisihloko sovavanyo lwe-AI yezonyango kwiimodeli ze-AI. Ukususela ekuqaleni kwe-2021 kunye ne-2022, abaphandi baye bacebisa izikhokelo zokunika ingxelo ezifana ne-Decid-AI, SPIRIT-AI, kunye ne-INTRPRT, ngethemba lokukhokela ukuphuhliswa kwangaphambili kunye nokuqinisekiswa kwe-AI yonyango phantsi kweemeko zokuqwalasela izinto ezifana nokusebenza kweklinikhi, ukhuseleko, izinto zabantu, kunye nokucaca / ukutolika. Kutshanje, Ijenali yendalo yezonyango ipapashe isifundo ngabaphandi abavela kwiYunivesithi yaseOxford kunye neYunivesithi yaseStanford malunga nokuba basebenzise "ukuqinisekiswa kwangaphandle" okanye "ukuqinisekiswa okuphindaphindiweyo kwendawo. "Ukuqinisekisa izixhobo ze-AI.

Ubume obungenamkhethe bezixhobo ze-AI kwakhona lukhokelo olubalulekileyo lokuvavanya olufumene ingqalelo kulo nyaka kuwo omabini amanqaku eSayensi kunye ne-NEJM. I-AI idla ngokubonisa i-bias kuba ilinganiselwe kwiidatha zoqeqesho. Olu tyekelo lunokubonisa ukungalingani kwezentlalo, nto leyo eqhubeka iguqukela kucalucalulo lwe-algorithmic. AmaZiko ezeMpilo eSizwe asanda kuqalisa inyathelo le-Bridge2AI, eliqikelelwa ukuba lixabisa i-130 yezigidi zeedola, ukwakha iiseti zedatha ezahlukeneyo (ngokuhambelana neenjongo zenyathelo le-MAIDA elikhankanywe ngasentla) ezinokusetyenziswa ukuqinisekisa ukungaqiniseki kwezixhobo zonyango ze-AI. Le miba ayiqwalaselwa yi-MultiMedQA. Umbuzo wendlela yokulinganisa kunye nokuqinisekisa iimodeli ze-AI zonyango usadinga ingxoxo ebanzi kunye nenzulu.

NgoJanuwari, i-Nature Medicine yapapasha isicatshulwa soluvo esibizwa ngokuba "Isizukulwana Esilandelayo soBubungqina obuSekwe kuNyango" evela kwiVivek Subbiah yeYunivesithi yaseTexas MD Anderson Cancer Centre, ephonononga imida yolingo lwezonyango oluveziweyo kumxholo we-COVID-19 bhubhane kwaye yalatha impikiswano phakathi kokutsha kunye nokubambelela kwinkqubo yophando lweklinikhi. Ekugqibeleni, libonisa ikamva lohlengahlengiso lwezilingo zeklinikhi - isizukulwana esilandelayo sovavanyo lweklinikhi usebenzisa i-intelligence yokufakelwa, oko kukuthi, ukusetyenziswa kobukrelekrele bokwenziwa kwinani elikhulu leenkcukacha zophando lwembali, idatha yehlabathi lokwenyani, idatha yeklinikhi emininzi, idatha yesixhobo esinxitywayo ukufumana ubungqina obuphambili. Ngaba oku kuthetha ukuba itekhnoloji ye-AI kunye neenkqubo zokuqinisekiswa kweklinikhi ze-AI zinokuqinisa kunye kunye nokuvela ngokubambisana kwixesha elizayo? Lo ngumbuzo ovulelekileyo nonika ingcinga ka-2023.

Ukulawulwa kwe-AI yezoNyango

Ukuqhubela phambili kweteknoloji ye-AI nayo ibeka imingeni kulawulo lwe-AI, kwaye abenzi bomgaqo-nkqubo emhlabeni jikelele baphendula ngokucophelela nangenyameko. Ngo-2019, i-FDA yaqala ukupapasha iSikhokelo soLawulo esiCetywayo soTshintsho lweSoftware kwiZixhobo zoBukrelekrele boBukrelekrele (iDrafti yeNgxoxo), ichaza indlela enokubakho yokuphononongwa kwangaphambili kwe-AI kunye nohlengahlengiso oluqhutywa kumatshini wokufunda. Ngo-2021, i-FDA iphakamise "i-Artificial Intelligence / Machine Learning-based Software njengeSicwangciso sokuSebenza kweSixhobo sezoNyango", esicacisa amanyathelo amahlanu olawulo lwezonyango lwe-AI. Kulo nyaka, i-FDA iphinde yaphinda yanikezela ngokungeniswa kwangaphambili kweeMpakethi zeSoftware yeSixhobo ukunika ulwazi malunga nezindululo zokungeniswa kweemarike kwangaphambili zovavanyo lwe-FDA lokhuseleko kunye nokusebenza kakuhle kweempawu zesoftware yesixhobo, kubandakanywa ezinye iimpawu zesixhobo sesoftware ezisebenzisa iimodeli zokufunda zoomatshini eziqeqeshwe ngeendlela zokufunda koomatshini. Umgaqo-nkqubo olawulayo we-FDA uye wavela kwisindululo sokuqala ukuya kwisikhokelo esisebenzayo.

Ukulandela ukupapashwa kweNdawo yeDatha yeMpilo yaseYurophu ngoJulayi kulo nyaka uphelileyo, i-EU iphinde yenze uMthetho woBukrelekrele beArtificial Intelligence. Eyangaphambili ijolise ekusebenziseni kakuhle idatha yezempilo ukubonelela ngononophelo lwezempilo oluphezulu, ukunciphisa ukungalingani, kunye nedatha yenkxaso yokuthintela, ukuxilongwa, unyango, ukuveliswa kwezinto ezintsha zesayensi, ukwenza izigqibo kunye nomthetho, ngelixa uqinisekisa ukuba abemi be-EU banolawulo olukhulu kwidatha yabo yezempilo. Le yokugqibela yenza kucace ukuba inkqubo yokuxilongwa kwezonyango yinkqubo ye-AI esemngciphekweni omkhulu, kwaye kufuneka yamkele ulongamelo olomeleleyo ekujoliswe kulo, ulongamelo lomjikelo wobomi bonke kunye nokuhlolwa kwangaphambili kovavanyo. I-Arhente ye-European Medicines Agency (EMA) iye yapapasha iPhepha lokuQinisekisa iDrafti malunga nokusetyenziswa kwe-AI ukuxhasa ukuphuhliswa kweziyobisi, ukulawulwa kunye nokusetyenziswa, ngokugxininisa ekuphuculeni ukuthembeka kwe-AI ukuqinisekisa ukhuseleko lwesigulane kunye nokunyaniseka kweziphumo zophando lweklinikhi. Ngokubanzi, indlela yokulawula ye-EU ngokuthe ngcembe iqala ukumila, kwaye iinkcukacha zokuphunyezwa kokugqibela zinokuba neenkcukacha ngakumbi kwaye zingqongqo. Ngokuchaseneyo nommiselo ongqongqo we-EU, iplani yolawulo ye-AI yase-UK yenza kucace ukuba urhulumente uceba ukuthatha indlela ethambileyo kwaye angawisi amatyala amatsha okanye amise abalawuli abatsha okwangoku.

E-China, iZiko loHlolo loBugcisa boNyango (i-NMPA) yoLawulo lweeMveliso zoNyango lweSizwe luye lwakhupha ngaphambili amaxwebhu afana ne-"Review Points of Deep Learning Assisted Decision Software", "Imigaqo-siseko yoBhaliso lweZixhobo zoNyango lwezoBuchule (iDrafti yokuHlola)" kunye ne "Setyhula kwiMigaqo yokuLawula iNgcaciso yeMveliso ye-Artificial Intelligence Medical Definition Management (Nombolo 47 ngo-2021)”. Kulo nyaka, "Isishwankathelo seziphumo zokuqala zokuhlelwa kwemveliso yesixhobo sonyango ngo-2023" sakhululwa kwakhona. Olu luhlu lwamaxwebhu lwenza inkcazo, ukuhlelwa kunye nokulawulwa kweemveliso zesoftware yobukrelekrele bokwenziwa kucace kwaye kube lula ukusebenza, kwaye ibonelela ngesikhokelo esicacileyo sokubekwa kwemveliso kunye nezicwangciso zobhaliso lwamashishini ahlukeneyo kushishino. I-China Medical Artificial Intelligence Conference eyayibanjelwe e-Hangzhou ukusuka nge-21 ukuya kwi-23 kaDisemba yaseka iqonga elikhethekileyo kulawulo lwezonyango lwedijithali kunye nophuhliso olukumgangatho ophezulu wezibhedlele zikawonkewonke kunye novavanyo lwezixhobo zonyango olungelulo kunye novavanyo lweqonga lophuhliso lomgangatho wobuchwephesha ngelo xesha.

Ukuqukumbela

Kwi-2023, i-AI yonyango sele iqalile ukudibanisa kuyo yonke inkqubo yonyango ephezulu kunye nezantsi, egubungela ukuqokelela idatha yesibhedlele, ukudibanisa, ukuhlalutya, ukuxilongwa kunye nonyango, kunye nokuhlolwa koluntu, kunye ne-organically isebenzisana nabasebenzi bezonyango / izifo, ebonisa amandla okuzisa impilo-ntle kwimpilo yabantu. Uphando lwe-AI lwezonyango olusebenzisekayo luqala ukuntwela kokusa. Kwixesha elizayo, inkqubela ye-AI yonyango ayixhomekeke kuphela kuphuhliso lwezobuchwepheshe ngokwayo, kodwa ifuna intsebenziswano epheleleyo yoshishino, iyunivesiti kunye nophando lwezonyango kunye nenkxaso yabenzi bomgaqo-nkqubo kunye nabalawuli. Le ntsebenziswano ye-cross-domain iyisitshixo ekufezekiseni iinkonzo zonyango ezidibeneyo ze-AI, kwaye ngokuqinisekileyo ziya kukhuthaza ukuphuhliswa kwempilo yabantu.