Mun gode da ziyartar Nature.com. Sigar burauzar da kuke amfani da ita tana da ƙarancin tallafin CSS. Don samun sakamako mafi kyau, muna ba da shawarar amfani da sabuwar sigar burauzar ku (ko kashe yanayin daidaitawa a cikin Internet Explorer). A halin yanzu, don tabbatar da ci gaba da tallafi, muna nuna shafin ba tare da salo ko JavaScript ba.
Amfani da fasahar basirar wucin gadi ta asibiti (AI) yana ƙaruwa cikin sauri, amma manhajojin makarantun likitanci da ake da su suna ba da taƙaitaccen koyarwa da ta shafi wannan fanni. A nan mun bayyana wani kwas na horar da fasahar wucin gadi da muka tsara kuma muka gabatar wa ɗaliban likitanci na Kanada kuma muna ba da shawarwari don horo a nan gaba.
Hankali na wucin gadi (AI) a fannin likitanci na iya inganta ingancin wurin aiki da kuma taimakawa wajen yanke shawara a asibiti. Domin jagorantar amfani da hankali na wucin gadi lafiya, likitoci dole ne su fahimci hazakar wucin gadi. Sharhi da yawa suna ba da shawarar koyar da ra'ayoyin AI1, kamar bayyana samfuran AI da hanyoyin tabbatarwa2. Duk da haka, an aiwatar da tsare-tsare kaɗan, musamman a matakin ƙasa. Pinto dos Santos et al.3. An yi wa ɗaliban likitanci 263 tambayoyi kuma kashi 71% sun yarda cewa suna buƙatar horo a fannin hazakar wucin gadi. Koyar da hazakar wucin gadi ga masu sauraron likita yana buƙatar ƙira mai kyau wanda ya haɗa ra'ayoyi na fasaha da waɗanda ba na fasaha ba ga ɗaliban da galibi suna da ilimi mai zurfi a baya. Muna bayyana ƙwarewarmu ta isar da jerin tarurrukan bita na AI ga ƙungiyoyi uku na ɗaliban likitanci kuma muna ba da shawarwari don ilimin likitanci na nan gaba a AI.
An gudanar da taron bita na makonni biyar na Gabatarwa ga Ilimin Sirri na Artificial a Medicine ga ɗaliban likitanci sau uku tsakanin Fabrairu 2019 da Afrilu 2021. Jadawalin kowane bita, tare da taƙaitaccen bayani game da canje-canje a cikin kwas ɗin, a cikin Hoto na 1. Kwas ɗinmu yana da manyan manufofin koyo guda uku: ɗalibai sun fahimci yadda ake sarrafa bayanai a cikin aikace-aikacen ilimin sirri na artificial, suna nazarin littattafan ilimin sirri na artificial don aikace-aikacen asibiti, da kuma amfani da damar yin aiki tare da injiniyoyi waɗanda ke haɓaka ilimin sirri na artificial.
Shuɗi shine batun laccar kuma shuɗi mai haske shine lokacin tambaya da amsa mai hulɗa. Sashen launin toka shine abin da aka fi mayar da hankali a kai a cikin ɗan gajeren bita na wallafe-wallafe. Sassan lemu sune zaɓaɓɓun nazarin shari'o'i waɗanda ke bayyana samfuran ko dabarun fasahar fasahar kere-kere. Kore kwas ne mai jagora wanda aka tsara don koyar da basirar wucin gadi don magance matsalolin asibiti da kimanta samfura. Abubuwan da ke ciki da tsawon lokacin bitar sun bambanta dangane da kimanta buƙatun ɗalibai.
An gudanar da taron bita na farko a Jami'ar British Columbia daga Fabrairu zuwa Afrilu 2019, kuma dukkan mahalarta 8 sun ba da ra'ayoyi masu kyau4. Saboda COVID-19, an gudanar da taron bita na biyu a zahiri a watan Oktoba-Nuwamba 2020, inda ɗaliban likitanci 222 da mazauna 3 daga makarantun likitanci 8 na Kanada suka yi rijista. An ɗora faifan gabatarwa da lambar zuwa wani shafin yanar gizo mai buɗewa (http://ubcaimed.github.io). Babban ra'ayoyin da aka samu daga maimaitawar farko shine cewa laccocin sun yi tsauri sosai kuma kayan aikin sun yi yawa a ka'ida. Yin hidima ga yankuna shida daban-daban na Kanada yana haifar da ƙarin ƙalubale. Don haka, taron bita na biyu ya rage kowane zaman zuwa awa 1, ya sauƙaƙa kayan kwas ɗin, ya ƙara ƙarin nazarin shari'o'i, kuma ya ƙirƙiri shirye-shiryen boilerplate waɗanda suka ba mahalarta damar kammala ƙananan bayanai na lambar tare da ƙarancin gyara kurakurai (Akwati na 1). Babban ra'ayoyin da aka samu daga maimaitawar ta biyu sun haɗa da ra'ayoyi masu kyau kan darussan shirye-shirye da kuma buƙatar nuna shirin aikin koyon na'ura. Saboda haka, a cikin bitarmu ta uku, wacce aka gudanar ta hanyar kwamfuta ga ɗaliban likitanci 126 a watan Maris-Afrilu 2021, mun haɗa da ƙarin darussan coding masu hulɗa da juna da kuma zaman ra'ayoyin aiki don nuna tasirin amfani da ra'ayoyin bitar kan ayyuka.
Binciken Bayanai: Fannin nazari a kididdiga wanda ke gano alamu masu ma'ana a cikin bayanai ta hanyar nazari, sarrafawa, da kuma isar da tsare-tsaren bayanai.
Haƙar bayanai: tsarin gano da kuma cire bayanai. A cikin mahallin fasahar kere-kere, wannan galibi yana da girma, tare da masu canji da yawa ga kowane samfurin.
Rage girman girma: Tsarin canza bayanai tare da fasaloli da yawa zuwa ƙananan fasaloli yayin da ake kiyaye mahimman kaddarorin saitin bayanai na asali.
Halaye (a cikin mahallin basirar wucin gadi): halayen da za a iya aunawa na samfurin. Sau da yawa ana amfani da su a musanya da "kadara" ko "mai canzawa".
Taswirar Kunnawa Mai Sauri: Wata dabara da ake amfani da ita wajen fassara samfuran fasahar kere-kere (musamman hanyoyin sadarwa na jijiyoyi masu juyi), wanda ke nazarin tsarin inganta ɓangaren ƙarshe na hanyar sadarwa don gano yankunan bayanai ko hotuna waɗanda ke da matuƙar hasashen yanayi.
Tsarin Daidaitacce: Tsarin AI da ake da shi wanda aka riga aka horar da shi don yin ayyuka makamantan haka.
Gwaji (a cikin mahallin basirar wucin gadi): lura da yadda samfurin ke yin aiki ta amfani da bayanan da bai taɓa fuskanta ba a da.
Horarwa (a cikin mahallin basirar wucin gadi): Samar da samfurin bayanai da sakamako ta yadda samfurin zai daidaita sigogin ciki don inganta ikonsa na yin ayyuka ta amfani da sabbin bayanai.
Vektor: jerin bayanai. A cikin koyon na'ura, kowane ɓangaren jeri yawanci siffa ce ta musamman ta samfurin.
Tebur na 1 ya lissafa sabbin darussa na watan Afrilun 2021, gami da manufofin koyo da aka tsara don kowane batu. An yi wannan bita ne ga waɗanda suka fara zuwa matakin fasaha kuma ba ya buƙatar wani ilimin lissafi fiye da shekarar farko ta digirin likitanci. Ɗaliban likitanci 6 da malamai 3 masu digiri na gaba a fannin injiniya ne suka tsara wannan kwas ɗin. Injiniyoyi suna haɓaka ka'idar basirar wucin gadi don koyarwa, kuma ɗaliban likitanci suna koyon kayan aiki masu dacewa da asibiti.
Bitar ta kunshi laccoci, nazarin shari'o'i, da shirye-shirye masu jagora. A cikin lacca ta farko, mun yi bitar zaɓaɓɓun ra'ayoyi na nazarin bayanai a cikin kididdigar halittu, gami da hangen nesa na bayanai, komawa ga tsarin dabaru, da kwatanta kididdigar bayanai da na inductive. Kodayake nazarin bayanai shine tushen basirar wucin gadi, mun ware batutuwa kamar hakar bayanai, gwajin mahimmanci, ko hangen nesa mai hulɗa. Wannan ya faru ne saboda ƙarancin lokaci kuma saboda wasu ɗaliban da ke karatun digiri na farko sun sami horo a baya a fannin ilimin halittu kuma suna son rufe ƙarin batutuwa na koyon injina na musamman. Laccar da ta biyo baya ta gabatar da hanyoyin zamani kuma ta tattauna tsarin matsalolin AI, fa'idodi da iyakokin samfuran AI, da gwajin samfura. Laccocin ya cika da adabi da bincike mai amfani kan na'urorin fasahar wucin gadi da ake da su. Muna jaddada ƙwarewar da ake buƙata don kimanta tasiri da yuwuwar samfurin don magance tambayoyin asibiti, gami da fahimtar iyakokin na'urorin fasahar wucin gadi da ake da su. Misali, mun nemi ɗalibai su fassara jagororin raunin kai na yara waɗanda Kupperman et al., 5 suka gabatar waɗanda suka aiwatar da tsarin yanke shawara na hankali na wucin gadi don tantance ko duban CT zai yi amfani bisa ga gwajin likita. Muna jaddada cewa wannan misali ne gama gari na AI wanda ke ba da nazarin hasashen ga likitoci don fassara, maimakon maye gurbin likitoci.
A cikin misalan shirye-shiryen bootstrap na tushen buɗewa da ake da su (https://github.com/ubcaimed/ubcaimed.github.io/tree/master/programming_examples), mun nuna yadda ake yin nazarin bayanai na bincike, rage girma, loda samfura na yau da kullun, da horo. da gwaji. Muna amfani da litattafan rubutu na Google Colaboratory (Google LLC, Mountain View, CA), waɗanda ke ba da damar aiwatar da lambar Python daga mai binciken yanar gizo. A cikin Hoto na 2 ya ba da misali na aikin shirye-shirye. Wannan aikin ya ƙunshi annabta cututtukan daji ta amfani da Wisconsin Open Breast Imaging Dataset6 da algorithm na yanke shawara na itace.
Gabatar da shirye-shirye a duk tsawon mako kan batutuwa masu alaƙa da juna kuma zaɓi misalai daga aikace-aikacen AI da aka buga. Abubuwan shirye-shirye ana haɗa su ne kawai idan an ɗauke su da mahimmanci don samar da haske game da ayyukan asibiti na gaba, kamar yadda ake kimanta samfura don tantance ko sun shirya don amfani a gwaje-gwajen asibiti. Waɗannan misalan sun ƙare da cikakken aikace-aikacen ƙarshe-zuwa-ƙarshe wanda ke rarraba ciwace-ciwacen a matsayin marasa lahani ko masu illa bisa ga sigogin hoton likita.
Bambancin ilimin da aka samu a baya. Mahalarta taronmu sun bambanta a matakin ilimin lissafi. Misali, ɗaliban da suka ƙware a fannin injiniyanci suna neman ƙarin bayani mai zurfi, kamar yadda za su yi nasu canjin Fourier. Duk da haka, tattauna tsarin Fourier a cikin aji ba zai yiwu ba saboda yana buƙatar zurfin ilimin sarrafa sigina.
Yawan halarta. Yawan halarta ya ragu, musamman a tsarin yanar gizo. Mafita ita ce a bi diddigin halarta da kuma bayar da takardar shaidar kammala karatun. An san makarantun likitanci da sanin rubuce-rubucen ayyukan ɗaliban da suka yi a fannin ilimi, wanda zai iya ƙarfafa ɗalibai su nemi digiri.
Tsarin Kwas: Saboda AI ta mamaye ƙananan fannoni da yawa, zaɓar manyan ra'ayoyi na zurfin da faɗin da ya dace na iya zama ƙalubale. Misali, ci gaba da amfani da kayan aikin AI daga dakin gwaje-gwaje zuwa asibitin muhimmin batu ne. Duk da yake muna rufe tsarin sarrafa bayanai, gina samfuri, da tabbatarwa, ba ma haɗa da batutuwa kamar manyan nazarin bayanai, hangen nesa mai hulɗa, ko gudanar da gwaje-gwajen asibiti na AI ba, maimakon haka muna mai da hankali kan mafi kyawun ra'ayoyin AI. Ka'idar jagorancinmu ita ce inganta karatu da rubutu, ba ƙwarewa ba. Misali, fahimtar yadda samfurin ke sarrafa fasalulluka na shigarwa yana da mahimmanci don fassarawa. Hanya ɗaya ta yin hakan ita ce amfani da taswirar kunna gradient, wanda zai iya hango waɗanne yankuna na bayanan ake iya hasashensu. Duk da haka, wannan yana buƙatar lissafin multivariate kuma ba za a iya gabatar da shi ba8. Ƙirƙirar kalmomi gama gari ya kasance ƙalubale saboda muna ƙoƙarin bayyana yadda ake aiki da bayanai a matsayin vectors ba tare da tsarin lissafi ba. Lura cewa kalmomi daban-daban suna da ma'ana iri ɗaya, misali, a cikin ilimin cututtuka, ana siffanta "halaye" a matsayin "canji" ko "halaye."
Rike ilimi. Saboda amfani da AI yana da iyaka, har yanzu ba a ga yadda mahalarta za su riƙe ilimi ba. Manhajojin makarantun likitanci galibi suna dogara ne akan maimaitawa ta sarari don ƙarfafa ilimi yayin juyawa na aiki,9 wanda kuma ana iya amfani da shi ga ilimin AI.
Ƙwarewa ta fi muhimmanci fiye da karatu da rubutu. An tsara zurfin kayan ba tare da tsauraran lissafi ba, wanda hakan ya kasance matsala yayin ƙaddamar da darussa na asibiti a fannin fasahar kere-kere. A cikin misalan shirye-shirye, muna amfani da tsarin samfuri wanda ke ba mahalarta damar cike filayen da gudanar da software ba tare da gano yadda za su kafa cikakken yanayin shirye-shirye ba.
Damuwa game da basirar wucin gadi da aka magance: Akwai damuwa sosai cewa basirar wucin gadi za ta iya maye gurbin wasu ayyukan asibiti3. Don magance wannan batu, mun bayyana iyakokin AI, gami da gaskiyar cewa kusan duk fasahar AI da masu kula da su suka amince da ita suna buƙatar kulawar likita11. Haka kuma muna jaddada mahimmancin son zuciya saboda algorithms suna da saurin nuna son zuciya, musamman idan saitin bayanai bai bambanta ba12. Saboda haka, ana iya yin kwaikwayon wani rukuni ba daidai ba, wanda ke haifar da yanke shawara na asibiti mara adalci.
Albarkatu suna samuwa ga jama'a: Mun ƙirƙiri albarkatu da ake da su a bainar jama'a, gami da nunin faifai na lacca da lambar kwamfuta. Duk da cewa samun damar shiga abubuwan da ke cikin tsarin synchronous yana da iyaka saboda yankunan lokaci, abubuwan da ke cikin tushen bude hanya ce mai dacewa don koyo asynchronous tunda ƙwarewar AI ba ta samuwa a duk makarantun likitanci.
Haɗin gwiwa tsakanin fannoni daban-daban: Wannan bita haɗin gwiwa ne da ɗaliban likitanci suka fara don tsara darussa tare da injiniyoyi. Wannan yana nuna damar haɗin gwiwa da gibin ilimi a fannoni biyu, yana ba mahalarta damar fahimtar rawar da za su iya takawa a nan gaba.
Bayyana ƙwarewar AI. Bayyana jerin ƙwarewa yana samar da tsari mai daidaito wanda za a iya haɗa shi cikin manhajar likitanci da ke da ƙwarewa. Wannan bita a halin yanzu yana amfani da Matakan Manufar Koyo na 2 (Fahimta), 3 (Aikace-aikace), da 4 (Nazari) na Tsarin Tsarin Bloom. Samun albarkatu a manyan matakan rarrabuwa, kamar ƙirƙirar ayyuka, na iya ƙara ƙarfafa ilimi. Wannan yana buƙatar yin aiki tare da ƙwararrun likitoci don tantance yadda za a iya amfani da batutuwan AI ga ayyukan asibiti da kuma hana koyar da batutuwa masu maimaitawa waɗanda aka riga aka haɗa a cikin manhajar likitanci ta yau da kullun.
Ƙirƙiri nazarin shari'o'i ta amfani da AI. Kamar misalan asibiti, koyon shari'o'i bisa ga shari'o'i na iya ƙarfafa ra'ayoyi marasa ma'ana ta hanyar nuna mahimmancin su ga tambayoyin asibiti. Misali, wani binciken bita ya binciki tsarin gano cutar sankarau ta hanyar AI na Google 13 don gano ƙalubalen da ke kan hanya daga dakin gwaje-gwaje zuwa asibiti, kamar buƙatun tabbatarwa na waje da hanyoyin amincewa da ƙa'idoji.
Yi amfani da ilimin gogewa: Kwarewar fasaha tana buƙatar yin aiki mai zurfi da kuma maimaita amfani da shi don ƙwarewa, kamar yadda ake yi wa ɗaliban da ke koyon aiki a asibiti. Wata mafita mai yuwuwa ita ce samfurin aji da aka juya, wanda aka ruwaito yana inganta riƙe ilimi a cikin ilimin injiniya14. A cikin wannan samfurin, ɗalibai suna yin bita kan kayan nazari daban-daban kuma lokacin aji an keɓe shi don magance matsaloli ta hanyar nazarin shari'o'i.
Ma'aunin mahalarta fannoni daban-daban: Muna tunanin ɗaukar AI wanda ya haɗa da haɗin gwiwa a fannoni daban-daban, ciki har da likitoci da ƙwararrun lafiya masu alaƙa da juna waɗanda ke da matakai daban-daban na horo. Saboda haka, manhajojin karatu na iya buƙatar a haɓaka su ta hanyar tattaunawa da malamai daga sassa daban-daban don daidaita abubuwan da ke cikin su zuwa fannoni daban-daban na kula da lafiya.
Hankali na wucin gadi fasaha ce mai zurfi kuma manyan manufofinsa suna da alaƙa da lissafi da kimiyyar kwamfuta. Horar da ma'aikatan kiwon lafiya don fahimtar hazakar wucin gadi yana gabatar da ƙalubale na musamman a cikin zaɓin abun ciki, mahimmancin asibiti, da hanyoyin isarwa. Muna fatan cewa fahimtar da aka samu daga bita na AI a cikin Ilimi za ta taimaka wa masu ilimi na gaba su rungumi hanyoyin kirkire-kirkire don haɗa AI cikin ilimin likitanci.
Rubutun Python na Google Colaboratory yana buɗe kuma ana iya samunsa a: https://github.com/ubcaimed/ubcaimed.github.io/tree/master/.
Prober, KG da Khan, S. Sake Tunani kan ilimin likitanci: kira zuwa ga aiki. Akkad. magani. 88, 1407–1410 (2013).
McCoy, LG da sauransu. Me ɗaliban likitanci suke buƙatar sani game da basirar wucin gadi? Lambobin NPZh. Magani 3, 1–3 (2020).
Dos Santos, DP, da sauransu. Ra'ayoyin ɗaliban likitanci game da basirar wucin gadi: wani bincike mai cibiyoyi da yawa. EURO. radiation. 29, 1640–1646 (2019).
Fan, KY, Hu, R., da Singla, R. Gabatarwa ga koyon injina ga ɗaliban likitanci: aikin gwaji. J. Med. koyarwa. 54, 1042–1043 (2020).
Cooperman N, da sauransu. Gano yara da ke cikin ƙarancin haɗarin kamuwa da cutar kwakwalwa mai tsanani bayan raunin kai: wani bincike mai zuwa na ƙungiyar. Lancet 374, 1160–1170 (2009).
Street, WN, Wolberg, WH da Mangasarian, OL. Cire fasalin nukiliya don gano ciwon nono. Kimiyyar Biomedical. Sarrafa hoto. Kimiyyar Biomedical. Weiss. 1905, 861–870 (1993).
Chen, PHC, Liu, Y. da Peng, L. Yadda ake haɓaka samfuran koyon injina don kiwon lafiya. Nat. Matt. 18, 410–414 (2019).
Selvaraju, RR da sauransu. Grad-cam: Fassarar gani na hanyoyin sadarwa masu zurfi ta hanyar fassara ta hanyar gradient bisa ga gradient. Takardun Taron Duniya na IEEE kan hangen nesa na kwamfuta, 618–626 (2017).
Kumaravel B, Stewart K da Ilic D. Haɓakawa da kimanta samfurin karkace don tantance ƙwarewar likitanci bisa shaida ta amfani da OSCE a cikin ilimin likitanci na digiri na farko. BMK Medicine. koyarwa. 21, 1–9 (2021).
Kolachalama VB da Garg PS Koyon injina da ilimin likitanci. Lambobin NPZh. magani. 1, 1–3 (2018).
van Leeuwen, KG, Schalekamp, S., Rutten, MJ, van Ginneken, B. da de Rooy, M. Artificial hankali a cikin rediyo: 100 kasuwanci kayayyakin da su kimiyya shaida. EURO radiation. 31, 3797-3804 (2021).
Topol, EJ Magunguna masu inganci: haɗuwar hankali na ɗan adam da na wucin gadi. Maganin Halitta. 25, 44–56 (2019).
Bede, E. da sauransu. Kimantawa mai zurfi ta ɗan adam game da tsarin koyo da aka tura a asibitin don gano cutar retinopathy ta masu ciwon suga. Takardun taron CHI na 2020 kan abubuwan da suka shafi ɗan adam a cikin tsarin kwamfuta (2020).
Kerr, B. Aji mai juyi a fannin ilimin injiniya: Bitar bincike. Takardun Taron Ƙasa da Ƙasa na 2015 kan Koyon Haɗin gwiwa Mai Haɗaka (2015).
Marubutan sun gode wa Danielle Walker, Tim Salcudin, da Peter Zandstra daga Ƙungiyar Binciken Halittu ta Biomedical Imaging and Artificial Intelligence a Jami'ar British Columbia saboda tallafi da kuɗaɗen da suka bayar.
RH, PP, ZH, RS da MA sune ke da alhakin tsara abubuwan da ke cikin koyarwar bitar. RH da PP sune ke da alhakin tsara misalan shirye-shirye. KYF, OY, MT da PW sune ke da alhakin tsara ayyukan da kuma nazarin bitar. RH, OY, MT, RS sune ke da alhakin ƙirƙirar adadi da tebura. RH, KYF, PP, ZH, OY, MY, PW, TL, MA, RS sune ke da alhakin tsara da kuma gyara takardar.
Ma'aikatar Sadarwa ta gode wa Carolyn McGregor, Fabio Moraes, da Aditya Borakati saboda gudummawar da suka bayar wajen bitar wannan aikin.
Lokacin Saƙo: Fabrairu-19-2024