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.
Ana ɗaukar haƙora a matsayin mafi daidaiton alamar shekarun jikin ɗan adam kuma galibi ana amfani da su a cikin kimanta shekarun bincike. Mun yi nufin tabbatar da kimanta shekarun haƙoran da aka dogara da bayanai ta hanyar kwatanta daidaiton kimantawa da aikin rarrabuwa na ƙa'idar shekaru 18 tare da hanyoyin gargajiya da kimanta shekarun da aka dogara da bayanai ta hanyar haƙoran. An tattara jimillar hotunan radiyo 2657 daga 'yan asalin Koriya da Japan masu shekaru 15 zuwa 23. An raba su zuwa tsarin horo, kowannensu yana ɗauke da hotunan radiyo 900 na Koriya, da kuma gwajin ciki wanda ya ƙunshi hotunan radiyo 857 na Japan. Mun kwatanta daidaiton rarrabuwa da ingancin hanyoyin gargajiya tare da gwajin samfuran haƙoran bayanai. Daidaiton hanyar gargajiya akan saitin gwajin ciki ya ɗan fi na samfurin haƙoran bayanai, kuma bambancin yana da ƙanƙanta (matsakaicin kuskuren cikakku < shekaru 0.21, kuskuren tushe na murabba'i < shekaru 0.24). Ayyukan rarrabuwa na yanke shekaru 18 suma iri ɗaya ne tsakanin hanyoyin gargajiya da samfuran haƙoran bayanai. Saboda haka, ana iya maye gurbin hanyoyin gargajiya da samfuran haƙar bayanai yayin yin kimanta shekarun bincike ta amfani da balaga na haƙoran biyu da na uku a cikin matasa da matasa 'yan Koriya.
Ana amfani da kimanta shekarun haƙori sosai a fannin likitancin shari'a da kuma ilimin hakora na yara. Musamman ma, saboda yawan alaƙar da ke tsakanin shekarun da suka gabata da kuma ci gaban haƙori, kimanta shekaru ta hanyar matakan ci gaban haƙori muhimmin ma'auni ne don tantance shekarun yara da matasa1,2,3. Duk da haka, ga matasa, kimanta shekarun haƙori bisa ga girman haƙori yana da iyakokinsa saboda ci gaban haƙori ya kusan cika, ban da haƙori na uku. Manufar shari'a ta ƙayyade shekarun matasa da matasa ita ce samar da kimantawa daidai da shaidar kimiyya ko sun kai shekarun girma. A cikin aikin likitanci na matasa da matasa a Koriya, an kiyasta shekaru ta amfani da hanyar Lee, kuma an annabta iyakar shari'a ta shekaru 18 bisa ga bayanan da Oh et al 5 suka ruwaito.
Koyon na'ura wani nau'in basirar wucin gadi ne (AI) wanda ke koyo da rarraba bayanai da yawa akai-akai, yana magance matsaloli da kansa, kuma yana jagorantar shirye-shiryen bayanai. Koyon na'ura na iya gano tsare-tsare masu amfani a cikin manyan bayanai6. Sabanin haka, hanyoyin gargajiya, waɗanda ke ɗaukar aiki da ɗaukar lokaci, na iya samun iyakoki lokacin da ake mu'amala da manyan bayanai masu rikitarwa waɗanda ke da wahalar sarrafawa da hannu7. Saboda haka, an gudanar da bincike da yawa kwanan nan ta amfani da sabbin fasahohin kwamfuta don rage kurakuran ɗan adam da kuma sarrafa bayanai masu girma dabam-dabam yadda ya kamata8,9,10,11,12. Musamman ma, an yi amfani da zurfin koyo sosai a cikin nazarin hotunan likita, kuma an ruwaito hanyoyi daban-daban don kimanta shekaru ta hanyar nazarin hotunan rediyo ta atomatik don inganta daidaito da ingancin kimanta shekaru13,14,15,16,17,18,19,20. Misali, Halabi et al 13 sun haɓaka tsarin koyon na'ura bisa ga hanyoyin sadarwa na jijiyoyi masu canzawa (CNN) don kimanta shekarun kwarangwal ta amfani da hotunan rediyo na hannayen yara. Wannan binciken ya gabatar da samfurin da ke amfani da koyon na'ura ga hotunan likita kuma ya nuna cewa waɗannan hanyoyin na iya inganta daidaiton ganewar asali. Li et al14 sun kiyasta shekaru daga hotunan X-ray na ƙashin ƙugu ta amfani da wani bincike mai zurfi na CNN kuma sun kwatanta su da sakamakon komawa baya ta amfani da kimanta matakin ossification. Sun gano cewa samfurin CNN mai zurfi yana nuna aikin kimanta shekaru iri ɗaya da samfurin komawa baya na gargajiya. Binciken Guo et al. [15] sun kimanta aikin rarraba juriyar shekaru na fasahar CNN bisa ga orthophotos na hakori, kuma sakamakon samfurin CNN ya tabbatar da cewa mutane sun fi ƙarfin aikin rarraba shekaru.
Yawancin bincike kan kimanta shekaru ta amfani da koyon na'ura suna amfani da hanyoyin koyo mai zurfi13,14,15,16,17,18,19,20. An ruwaito cewa kimanta shekaru bisa ga zurfafan koyo ya fi daidai fiye da hanyoyin gargajiya. Duk da haka, wannan hanyar ba ta ba da dama sosai don gabatar da tushen kimiyya don kimanta shekaru, kamar alamun shekaru da aka yi amfani da su a cikin kimantawa. Akwai kuma takaddama ta shari'a kan wanda ke gudanar da binciken. Saboda haka, kimanta shekaru bisa ga zurfafan koyo yana da wuya a yarda da hukumomin gudanarwa da na shari'a. Haƙar bayanai (DM) wata dabara ce da za ta iya gano ba wai kawai ana tsammani ba har ma da bayanan da ba a zata ba a matsayin hanyar gano alaƙa mai amfani tsakanin adadi mai yawa na bayanai6,21,22. Sau da yawa ana amfani da koyon na'ura a haƙar bayanai, kuma duka haƙar bayanai da koyon na'ura suna amfani da mahimman algorithms iri ɗaya don gano alamu a cikin bayanai. Kimanta shekaru ta amfani da haɓaka haƙori ya dogara ne akan kimantawar mai bincike game da balagar haƙoran da aka yi niyya, kuma an bayyana wannan kimantawa a matsayin mataki ga kowane haƙori da aka yi niyya. Ana iya amfani da DM don nazarin alaƙar da ke tsakanin matakin kimanta haƙoran da ainihin shekaru kuma yana da yuwuwar maye gurbin nazarin ƙididdiga na gargajiya. Saboda haka, idan muka yi amfani da dabarun DM don kimanta shekaru, za mu iya aiwatar da koyon injina a cikin kimanta shekarun bincike ba tare da damuwa game da alhakin shari'a ba. An buga wasu nazarce-nazarce na kwatantawa kan hanyoyin da za a iya amfani da su na gargajiya da aka yi amfani da su a cikin aikin bincike da hanyoyin EBM don tantance shekarun hakori. Shen et al23 sun nuna cewa samfurin DM ya fi daidaito fiye da dabarar Camerer ta gargajiya. Galibourg et al24 sun yi amfani da hanyoyi daban-daban na DM don hasashen shekaru bisa ga ma'aunin Demirdjian25 kuma sakamakon ya nuna cewa hanyar DM ta yi fice a cikin hanyoyin Demirdjian da Willems wajen kimanta shekarun al'ummar Faransa.
Don kimanta shekarun haƙoran matasa 'yan Koriya da matasa, ana amfani da hanyar Lee ta 4 sosai a cikin aikin binciken laifuka na Koriya. Wannan hanyar tana amfani da nazarin kididdiga na gargajiya (kamar komawa baya da yawa) don bincika alaƙar da ke tsakanin mutanen Koriya da shekarun tarihi. A cikin wannan binciken, hanyoyin kimanta shekaru da aka samu ta amfani da hanyoyin kididdiga na gargajiya an bayyana su a matsayin "hanyoyin gargajiya." Hanyar Lee hanya ce ta gargajiya, kuma Oh et al. 5 sun tabbatar da daidaitonta; duk da haka, amfani da kimanta shekaru bisa ga samfurin DM a cikin aikin binciken laifuka na Koriya har yanzu abin tambaya ne. Manufarmu ita ce a tabbatar da yuwuwar amfani da kimanta shekaru bisa ga samfurin DM. Manufar wannan binciken ita ce (1) kwatanta daidaiton samfuran DM guda biyu wajen kimanta shekarun haƙori da (2) kwatanta aikin rarrabuwa na samfuran DM guda 7 a shekaru 18 da waɗanda aka samu ta amfani da hanyoyin kididdiga na gargajiya Girman haƙora na biyu da na uku a cikin muƙamuƙi biyu.
Ma'auni da daidaitattun bambance-bambance na shekaru na lokaci ta hanyar mataki da nau'in haƙori an nuna su akan layi a cikin Tebur na Ƙarin S1 (satin horo), Tebur na Ƙarin S2 (satin gwaji na ciki), da Tebur na Ƙarin S3 (satin gwaji na waje). Ƙimar kappa don amincin mai lura da ciki da na interobserver da aka samu daga saitin horon sune 0.951 da 0.947, bi da bi. Ƙimar P da tazara ta amincewa 95% don ƙimar kappa an nuna su a cikin teburin ƙarin kan layi S4. An fassara ƙimar kappa a matsayin "kusan cikakke", daidai da ƙa'idodin Landis da Koch26.
Idan aka kwatanta matsakaicin kuskuren da ya dace (MAE), hanyar gargajiya ta fi ƙarfin samfurin DM ga dukkan jinsi da kuma a cikin gwajin maza na waje, ban da perceptron mai layi da yawa (MLP). Bambancin da ke tsakanin samfurin gargajiya da samfurin DM akan saitin gwajin MAE na ciki shine shekaru 0.12–0.19 ga maza da shekaru 0.17–0.21 ga mata. Ga batirin gwajin waje, bambance-bambancen sun fi ƙanƙanta (shekaru 0.001–0.05 ga maza da shekaru 0.05–0.09 ga mata). Bugu da ƙari, kuskuren tushen matsakaicin murabba'i (RMSE) ya ɗan yi ƙasa da hanyar gargajiya, tare da ƙananan bambance-bambance (0.17–0.24, 0.2–0.24 ga saitin gwajin ciki na maza, da 0.03–0.07, 0.04–0.08 ga saitin gwajin waje). ). MLP yana nuna ɗan aiki mafi kyau fiye da Perceptron Layer ɗaya (SLP), sai dai idan yanayin saitin gwajin waje na mata ne. Ga MAE da RMSE, gwajin waje ya nuna maki sama da gwajin ciki na dukkan jinsi da samfura. Duk MAE da RMSE an nuna su a cikin Jadawali na 1 da Hoto na 1.
MAE da RMSE na samfuran komawa ga ma'adinan bayanai na gargajiya da na asali. Kuskuren Ma'ana Ma'ana Ma'ana, Kuskuren Ma'ana Ma'ana Ma'ana, Kuskuren Ma'ana Ma'ana Ma'ana, Kuskuren Ma'ana Ma'ana Ma'ana, SLP mai layi ɗaya, MLP mai layi da yawa, hanyar CM ta gargajiya.
An nuna aikin rarrabuwa (tare da yanke shekaru 18) na samfuran gargajiya da na DM dangane da hankali, takamaiman aiki, ƙimar hasashen mai kyau (PPV), ƙimar hasashen mara kyau (NPV), da yanki a ƙarƙashin lanƙwasa halayyar aiki na mai karɓar (AUROC) 27 (Tebur 2, Hoto na 2 da Hoto na Ƙari na 1 akan layi). Dangane da yanayin ƙarfin batirin gwajin ciki, hanyoyin gargajiya sun fi kyau a tsakanin maza da mafi muni a tsakanin mata. Duk da haka, bambancin aikin rarrabuwa tsakanin hanyoyin gargajiya da SD shine 9.7% ga maza (MLP) kuma 2.4% kawai ga mata (XGBoost). Daga cikin samfuran DM, koma-baya na dabaru (LR) ya nuna mafi kyawun jin daɗi a cikin jinsi biyu. Dangane da takamaiman saitin gwajin ciki, an lura cewa samfuran SD guda huɗu sun yi kyau a cikin maza, yayin da samfurin gargajiya ya fi kyau a cikin mata. Bambance-bambancen aikin rarrabuwa ga maza da mata shine 13.3% (MLP) da 13.1% (MLP), bi da bi, yana nuna cewa bambancin aikin rarrabuwa tsakanin samfuran ya wuce jin daɗi. Daga cikin samfuran DM, na'urar vector mai tallafi (SVM), bishiyar yanke shawara (DT), da samfuran daji na bazuwar (RF) sun fi kyau a tsakanin maza, yayin da samfurin LR ya fi kyau a tsakanin mata. AUROC na samfurin gargajiya da duk samfuran SD ya fi 0.925 (maƙwabcin k-mafi kusa (KNN) a cikin maza), yana nuna kyakkyawan aikin rarrabuwa wajen bambance samfuran shekaru 1828. Ga saitin gwajin waje, akwai raguwar aikin rarrabuwa dangane da hankali, takamaiman aiki da AUROC idan aka kwatanta da saitin gwajin ciki. Bugu da ƙari, bambancin da ke tsakanin aikin rarrabuwa na mafi kyau da mafi munin samfura ya kasance daga 10% zuwa 25% kuma ya fi girma fiye da bambancin da ke cikin saitin gwajin ciki.
Sanin hankali da takamaiman samfuran rarrabuwar bayanai idan aka kwatanta da hanyoyin gargajiya tare da yankewa na shekaru 18. KNN k maƙwabci mafi kusa, injin vector na tallafi na SVM, komawar logistic na LR, itacen yanke shawara na DT, RF random forest, XGB XGBoost, MLP multilayer perceptron, hanyar CM ta gargajiya.
Mataki na farko a cikin wannan binciken shine a kwatanta daidaiton kiyasin shekarun haƙori da aka samu daga samfuran DM guda bakwai da waɗanda aka samu ta amfani da hanyar komawa ga al'ada. An tantance MAE da RMSE a cikin gwaje-gwaje na ciki ga jinsi biyu, kuma bambancin da ke tsakanin hanyar gargajiya da samfurin DM ya kasance daga kwanaki 44 zuwa 77 ga MAE da kuma daga kwanaki 62 zuwa 88 ga RMSE. Duk da cewa hanyar gargajiya ta ɗan fi daidai a cikin wannan binciken, yana da wuya a kammala ko irin wannan ƙaramin bambanci yana da mahimmanci na asibiti ko na aiki. Waɗannan sakamakon sun nuna cewa daidaiton kiyasin shekarun haƙori ta amfani da samfurin DM kusan iri ɗaya ne da na hanyar gargajiya. Kwatanta kai tsaye da sakamakon binciken da ya gabata yana da wahala saboda babu wani bincike da ya kwatanta daidaiton samfuran DM tare da hanyoyin ƙididdiga na gargajiya ta amfani da dabarar yin rikodin haƙora a cikin shekarun da suka gabata kamar yadda yake a cikin wannan binciken. Galibourg et al24 sun kwatanta MAE da RMSE tsakanin hanyoyin gargajiya guda biyu (hanyar Demirjian25 da hanyar Willems29) da samfuran DM guda 10 a cikin al'ummar Faransa masu shekaru 2 zuwa 24. Sun bayar da rahoton cewa duk samfuran DM sun fi daidaito fiye da hanyoyin gargajiya, tare da bambance-bambancen shekaru 0.20 da 0.38 a cikin MAE da shekaru 0.25 da 0.47 a cikin RMSE idan aka kwatanta da hanyoyin Willems da Demirdjian, bi da bi. Bambancin da ke tsakanin samfurin SD da hanyoyin gargajiya da aka nuna a cikin binciken Halibourg ya yi la'akari da rahotanni da yawa30,31,32,33 cewa hanyar Demirdjian ba ta kimanta shekarun hakora daidai a cikin al'ummomi banda 'yan Kanada na Faransa waɗanda binciken ya dogara da su ba. a cikin wannan binciken. Tai et al 34 sun yi amfani da tsarin MLP don hango shekarun hakora daga hotunan orthodontic na China 1636 kuma sun kwatanta daidaitonsa da sakamakon hanyar Demirjian da Willems. Sun bayar da rahoton cewa MLP yana da daidaito mafi girma fiye da hanyoyin gargajiya. Bambancin da ke tsakanin hanyar Demirdjian da hanyar gargajiya shine <0.32 shekaru, kuma hanyar Willems shekaru 0.28 ne, wanda yayi kama da sakamakon wannan binciken. Sakamakon waɗannan binciken da suka gabata24,34 suma sun yi daidai da sakamakon wannan binciken, kuma daidaiton kimanta shekaru na samfurin DM da hanyar gargajiya iri ɗaya ne. Duk da haka, bisa ga sakamakon da aka gabatar, za mu iya yanke hukunci da kyau cewa amfani da samfuran DM don kimanta shekaru na iya maye gurbin hanyoyin da ake da su saboda rashin nazarin da aka yi a baya na kwatantawa da nassoshi. Ana buƙatar nazarin da aka yi amfani da manyan samfura don tabbatar da sakamakon da aka samu a cikin wannan binciken.
Daga cikin binciken da aka gwada daidaiton SD wajen kimanta shekarun haƙori, wasu sun nuna daidaito mafi girma fiye da bincikenmu. Stepanovsky da abokan aikinsa 35 sun yi amfani da samfuran SD guda 22 a kan hotunan radiyo na mazauna Czech 976 masu shekaru 2.7 zuwa 20.5 kuma sun gwada daidaiton kowanne samfurin. Sun kimanta ci gaban jimillar haƙoran dindindin guda 16 na sama da na ƙasa na hagu ta amfani da ka'idojin rarrabuwa da Moorrees et al 36 suka gabatar. MAE yana tsakanin shekaru 0.64 zuwa 0.94 kuma RMSE yana tsakanin shekaru 0.85 zuwa 1.27, waɗanda suka fi daidaito fiye da samfuran DM guda biyu da aka yi amfani da su a wannan binciken. Shen da abokan aikinsa 23 sun yi amfani da hanyar Cameriere don kimanta shekarun haƙoran dindindin guda bakwai a cikin haƙoran hagu a cikin mazauna gabashin China masu shekaru 5 zuwa 13 kuma sun kwatanta shi da shekarun da aka kiyasta ta amfani da layin layi, SVM da RF. Sun nuna cewa duk samfuran DM guda uku suna da daidaito mafi girma idan aka kwatanta da dabarar Cameriere ta gargajiya. MAE da RMSE a cikin binciken Shen sun yi ƙasa da waɗanda ke cikin samfurin DM a cikin wannan binciken. Ƙara daidaiton binciken da Stepanovsky et al. 35 da Shen et al. 23 suka yi na iya zama saboda haɗa ƙananan yara a cikin samfuran bincikensu. Saboda ƙididdigar shekaru ga mahalarta masu hakora masu tasowa ya zama daidai yayin da adadin hakora ke ƙaruwa yayin haɓaka hakori, daidaiton hanyar kimanta shekaru da aka samu na iya yin rauni lokacin da mahalarta binciken suka kasance ƙanana. Bugu da ƙari, kuskuren MLP a cikin kimanta shekaru ya ɗan ƙanƙanta fiye da na SLP, ma'ana cewa MLP ya fi SLP daidai. Ana ɗaukar MLP a matsayin mafi kyau don kimanta shekaru, wataƙila saboda ɓoyayyun yadudduka a cikin MLP38. Duk da haka, akwai banda ga samfurin waje na mata (SLP 1.45, MLP 1.49). Binciken cewa MLP ya fi SLP daidai a kimanta shekaru yana buƙatar ƙarin nazarin baya.
An kuma kwatanta aikin rarraba samfurin DM da hanyar gargajiya a matakin shekaru 18. Duk samfuran SD da aka gwada da hanyoyin gargajiya a cikin saitin gwajin ciki sun nuna matakan wariya kusan masu karɓuwa ga samfurin mai shekaru 18. Jin zafi ga maza da mata ya fi 87.7% da 94.9%, bi da bi, kuma takamaiman ya fi 89.3% da 84.7%. AUROC na duk samfuran da aka gwada shi ma ya wuce 0.925. A gwargwadon iliminmu, babu wani bincike da ya gwada aikin samfurin DM na rarraba shekaru 18 bisa ga girman haƙori. Za mu iya kwatanta sakamakon wannan binciken da aikin rarraba samfuran koyo mai zurfi akan hotunan radiyo na panoramic. Guo et al.15 sun ƙididdige aikin rarrabawa na samfurin koyo mai zurfi na CNN da hanyar hannu bisa ga hanyar Demirjian don wani matakin shekaru. Hankali da takamaiman hanyar da aka yi amfani da ita wajen tantancewa sun kai kashi 87.7% da 95.5%, bi da bi, kuma jin daɗin da takamaiman samfurin CNN ya kai kashi 89.2% da 86.6%, bi da bi. Sun kammala da cewa zurfin koyo na iya maye gurbin ko kuma ya fi ƙarfin kimantawa ta hannu wajen rarraba ƙa'idodin shekaru. Sakamakon wannan binciken ya nuna irin wannan aikin rarrabuwa; Ana kyautata zaton rarrabuwa ta amfani da samfuran DM na iya maye gurbin hanyoyin ƙididdiga na gargajiya don kimanta shekaru. Daga cikin samfuran, DM LR shine mafi kyawun samfurin dangane da jin daɗin samfurin maza da jin daɗin da takamaiman samfurin mace. LR tana matsayi na biyu a cikin takamaiman samfurin maza. Bugu da ƙari, ana ɗaukar LR a matsayin ɗaya daga cikin samfuran DM35 mafi sauƙin amfani kuma ba shi da rikitarwa kuma yana da wahalar sarrafawa. Dangane da waɗannan sakamakon, an ɗauki LR a matsayin mafi kyawun samfurin rarrabuwa ga yara 'yan shekara 18 a cikin al'ummar Koriya.
Gabaɗaya, daidaiton kimanta shekaru ko aikin rarrabuwa akan saitin gwajin waje ya kasance mara kyau ko ƙasa idan aka kwatanta da sakamakon gwajin ciki. Wasu rahotanni sun nuna cewa daidaiton rarrabuwa ko inganci yana raguwa lokacin da aka yi amfani da kimanta shekaru bisa ga yawan mutanen Koriya ga yawan mutanen Japan5,39, kuma an sami irin wannan tsari a cikin wannan binciken. An kuma lura da wannan yanayin tabarbarewa a cikin samfurin DM. Saboda haka, don kimanta shekaru daidai, koda lokacin amfani da DM a cikin tsarin bincike, ya kamata a fifita hanyoyin da aka samo daga bayanan yawan mutanen asali, kamar hanyoyin gargajiya,5,39,40,41,42. Tunda ba a san ko samfuran ilmantarwa masu zurfi na iya nuna irin wannan yanayin ba, ana buƙatar nazarin kwatanta daidaiton rarrabuwa da inganci ta amfani da hanyoyin gargajiya, samfuran DM, da samfuran ilmantarwa masu zurfi akan samfuran iri ɗaya don tabbatar da ko basirar wucin gadi za ta iya shawo kan waɗannan bambance-bambancen launin fata a cikin iyakantaccen shekaru. kimantawa.
Mun nuna cewa hanyoyin gargajiya za a iya maye gurbinsu da kimanta shekaru bisa ga tsarin DM a cikin aikin kimanta shekarun bincike a Koriya. Mun kuma gano yiwuwar aiwatar da koyon injina don tantance shekarun bincike. Duk da haka, akwai iyakoki bayyanannu, kamar ƙarancin adadin mahalarta a cikin wannan binciken don tantance sakamakon, da kuma rashin binciken da aka yi a baya don kwatantawa da tabbatar da sakamakon wannan binciken. A nan gaba, ya kamata a gudanar da nazarin DM tare da adadi mai yawa na samfura da kuma al'ummomi daban-daban don inganta amfaninsa na aiki idan aka kwatanta da hanyoyin gargajiya. Don tabbatar da yuwuwar amfani da basirar wucin gadi don kimanta shekaru a cikin al'ummomi da yawa, ana buƙatar nazarin gaba don kwatanta daidaiton rarrabuwa da ingancin DM da samfuran ilmantarwa masu zurfi tare da hanyoyin gargajiya a cikin samfuran iri ɗaya.
Binciken ya yi amfani da hotunan rubutu guda 2,657 da aka tattara daga manyan mutanen Koriya da Japan masu shekaru 15 zuwa 23. An raba hotunan rediyo na Koriya zuwa saitin horo guda 900 (shekaru 19.42 ± 2.65) da kuma saitin gwaje-gwaje na ciki guda 900 (shekaru 19.52 ± 2.59). An tattara saitin horon a wata cibiya (Asibitin Seattle St. Mary), kuma an tattara saitin gwajin a cibiyoyi biyu (Asibitin Hakori na Jami'ar Kasa ta Seoul da Asibitin Hakori na Jami'ar Yonsei). Mun kuma tattara hotunan rediyo guda 857 daga wani bayanai na al'umma (Jami'ar Likitanci ta Iwate, Japan) don gwajin waje. An zaɓi hotunan rediyo na mutanen Japan (shekaru 19.31 ± 2.60) a matsayin saitin gwajin waje. An tattara bayanai daga baya don nazarin matakan ci gaban hakori akan hotunan rediyo na panoramic da aka ɗauka yayin maganin hakori. Duk bayanan da aka tattara ba a san su ba sai dai jinsi, ranar haihuwa da ranar rediyo. Sharuɗɗan haɗawa da cirewa iri ɗaya ne da na binciken da aka buga a baya 4, 5. An ƙididdige ainihin shekarun samfurin ta hanyar cire ranar haihuwa daga ranar da aka ɗauki hoton. An raba rukunin samfurin zuwa ƙungiyoyi tara na shekaru. An nuna shekarun da aka raba da jinsi a cikin Jadawali na 3. An gudanar da wannan binciken ne bisa ga Sanarwar Helsinki kuma Hukumar Bita ta Cibiya (IRB) ta Seoul St. Mary's Hospital na Jami'ar Katolika ta Koriya (KC22WISI0328) ta amince da shi. Saboda tsarin da aka tsara na baya-bayan nan na wannan binciken, ba a iya samun izini daga duk marasa lafiya da ake yi wa gwajin hoton rediyo don dalilai na magani ba. Asibitin St. Mary na Jami'ar Seoul ta Koriya (IRB) ya yi watsi da buƙatar izini mai tushe.
An tantance matakan ci gaban haƙoran bimaxillary na biyu da na uku bisa ga ka'idojin Demircan25. An zaɓi haƙori ɗaya kawai idan an sami irin wannan haƙorin a gefen hagu da dama na kowane muƙamuƙi. Idan haƙoran da suka yi kama da juna a ɓangarorin biyu suna cikin matakai daban-daban na ci gaba, an zaɓi haƙorin da ke da matakin ci gaba mai ƙasa don la'akari da rashin tabbas a cikin shekarun da aka kiyasta. An yi amfani da na'urar daukar hoto guda ɗari da aka zaɓa bazuwar daga saitin horo ta hanyar masu lura biyu masu ƙwarewa don gwada amincin masu lura da juna bayan daidaitawa kafin a tantance matakin balaga na haƙori. An tantance amincin mai lura da intraobserver sau biyu a tazara na watanni uku ta babban mai lura.
An kiyasta matakin jinsi da ci gaban haƙoran na biyu da na uku na kowane muƙamuƙi a cikin saitin horo ta hanyar wani babban mai lura da aka horar da shi tare da samfuran DM daban-daban, kuma an saita ainihin shekarun a matsayin ƙimar manufa. An gwada samfuran SLP da MLP, waɗanda ake amfani da su sosai a cikin koyon injina, akan algorithms na komawa baya. Tsarin DM yana haɗa ayyukan layi ta amfani da matakan ci gaban haƙora huɗu kuma yana haɗa waɗannan bayanan don kimanta shekaru. SLP shine mafi sauƙin hanyar sadarwa ta jijiyoyi kuma baya ɗauke da yadudduka ɓoyayye. SLP yana aiki bisa ga watsawa tsakanin nodes. Tsarin SLP a cikin komawa baya yayi kama da komawa baya da yawa a lissafi. Ba kamar samfurin SLP ba, samfurin MLP yana da yadudduka ɓoyayye da yawa tare da ayyukan kunnawa marasa layi. Gwaje-gwajenmu sun yi amfani da Layer ɓoyayye tare da nodes ɓoyayye 20 kawai tare da ayyukan kunnawa marasa layi. Yi amfani da saukowa mai sauƙi azaman hanyar ingantawa da MAE da RMSE azaman aikin asara don horar da samfurin koyon injinmu. An yi amfani da mafi kyawun samfurin komawa baya da aka samu akan saitin gwaje-gwaje na ciki da na waje kuma an kiyasta shekarun haƙoran.
An ƙirƙiri wani tsari na rarrabuwa wanda ke amfani da balaga na haƙora huɗu a kan saitin horo don yin hasashen ko samfurin yana da shekaru 18 ko a'a. Don gina samfurin, mun samo algorithms guda bakwai na koyo na injina6,43: (1) LR, (2) KNN, (3) SVM, (4) DT, (5) RF, (6) XGBoost, da (7) MLP. LR yana ɗaya daga cikin algorithms na rarrabuwa da aka fi amfani da su44. Tsarin ilmantarwa ne mai kulawa wanda ke amfani da koma-baya don annabta yuwuwar bayanai na wani rukuni daga 0 zuwa 1 kuma yana rarraba bayanan a matsayin na wani rukuni mafi yuwuwa bisa ga wannan yuwuwar; galibi ana amfani da shi don rarrabuwar binary. KNN yana ɗaya daga cikin algorithms na koyo na injina mafi sauƙi45. Idan aka ba shi sabbin bayanai na shigarwa, yana samun bayanan k kusa da saitin da ke akwai sannan ya rarraba su cikin aji tare da mafi girman mita. Mun saita 3 don adadin maƙwabta da aka yi la'akari da su (k). SVM tsari ne wanda ke haɓaka nisan da ke tsakanin azuzuwan biyu ta amfani da aikin kernel don faɗaɗa sararin layi zuwa sararin da ba layi ba wanda ake kira filayen46. Don wannan samfurin, muna amfani da bias = 1, iko = 1, da gamma = 1 a matsayin hyperparameters don ƙwayar polynomial. An yi amfani da DT a fannoni daban-daban a matsayin algorithm don raba dukkan bayanai da aka saita zuwa ƙananan ƙungiyoyi da yawa ta hanyar wakiltar ƙa'idodin yanke shawara a cikin tsarin itace47. An tsara samfurin tare da mafi ƙarancin adadin bayanai a kowace ƙulli na 2 kuma yana amfani da ma'aunin Gini a matsayin ma'aunin inganci. RF hanya ce ta tarin bayanai wacce ke haɗa DTs da yawa don inganta aiki ta amfani da hanyar haɗa bootstrap wanda ke samar da mai rarrabuwa mai rauni ga kowane samfurin ta hanyar zana samfuran girman iri ɗaya sau da yawa daga ainihin bayanan48. Mun yi amfani da bishiyoyi 100, zurfin bishiyoyi 10, mafi ƙarancin girman ƙulli 1, da ma'aunin haɗin Gini a matsayin ma'aunin rabuwar ƙulli. Rarraba sabbin bayanai ana ƙaddara su ta hanyar kuri'a mafi rinjaye. XGBoost algorithm ne wanda ke haɗa dabarun haɓakawa ta amfani da hanyar da ke ɗaukar bayanan horo a matsayin kuskuren tsakanin ainihin ƙimar da aka annabta na samfurin da ya gabata kuma yana ƙara kuskuren ta amfani da gradients49. Tsarin aiki ne da ake amfani da shi sosai saboda kyakkyawan aiki da ingancin albarkatu, da kuma babban aminci a matsayin aikin gyara da ya wuce gona da iri. An sanye wannan samfurin da ƙafafun tallafi 400. MLP wata hanyar sadarwa ce ta jijiyoyi inda ɗaya ko fiye da perceptrons ke samar da yadudduka da yawa tare da layuka ɗaya ko fiye da ɓoye tsakanin layukan shigarwa da fitarwa38. Ta amfani da wannan, zaku iya yin rarrabuwa mara layi inda lokacin da kuka ƙara layin shigarwa kuma kuka sami ƙimar sakamako, ana kwatanta ƙimar sakamakon da aka annabta da ainihin ƙimar sakamako kuma kuskuren ya dawo. Mun ƙirƙiri wani Layer ɓoyayye tare da ƙwayoyin jijiyoyi 20 ɓoye a kowane Layer. Kowane samfurin da muka ƙirƙira an yi amfani da shi ga saitin ciki da na waje don gwada aikin rarrabuwa ta hanyar ƙididdige hankali, takamaiman aiki, PPV, NPV, da AUROC. Ana bayyana hankali a matsayin rabo na samfurin da aka kiyasta yana da shekaru 18 ko sama da haka zuwa samfurin da aka kiyasta yana da shekaru 18 ko sama da haka. Musamman shine rabon samfuran da ba su kai shekaru 18 ba da waɗanda aka kiyasta suna ƙasa da shekaru 18.
An mayar da matakan hakori da aka tantance a cikin tsarin horo zuwa matakan lambobi don nazarin ƙididdiga. An yi amfani da tsarin jujjuyawar layi da na dabaru masu yawa don ƙirƙirar samfuran hasashen kowane jinsi da kuma samar da dabarun komawa baya waɗanda za a iya amfani da su don kimanta shekaru. Mun yi amfani da waɗannan dabarun don kimanta shekarun haƙori na gwaje-gwaje na ciki da na waje. Tebur na 4 yana nuna samfuran komawa baya da rarrabuwa da aka yi amfani da su a cikin wannan binciken.
An ƙididdige amincin masu lura da juna ta amfani da ƙididdigar kappa ta Cohen. Don gwada daidaiton samfuran DM da na gargajiya, mun ƙididdige MAE da RMSE ta amfani da shekarun da aka kiyasta da ainihin saitin gwaje-gwaje na ciki da waje. Waɗannan kurakuran galibi ana amfani da su don kimanta daidaiton hasashen samfuri. Ƙaramin kuskuren, mafi girman daidaiton hasashen24. Kwatanta MAE da RMSE na saitin gwaje-gwaje na ciki da waje da aka ƙididdige ta amfani da DM da na gargajiya na koma-baya. An kimanta aikin rarrabuwa na yanke shekaru 18 a cikin ƙididdigar gargajiya ta amfani da teburin gaggawa na 2 × 2. An kwatanta ƙwarewar lissafi, takamaiman aiki, PPV, NPV, da AUROC na saitin gwajin tare da ƙimar da aka auna na samfurin rarrabuwa na DM. Ana bayyana bayanai a matsayin matsakaicin ± karkacewar misali ko lamba (%) dangane da halayen bayanai. An ɗauki ƙimar P mai gefe biyu <0.05 a matsayin mahimmanci a kididdiga. An yi duk nazarin ƙididdiga na yau da kullun ta amfani da sigar SAS 9.4 (SAS Institute, Cary, NC). An aiwatar da tsarin DM regression a Python ta amfani da Keras50 2.2.4 backend da Tensorflow51 1.8.0 musamman don ayyukan lissafi. An aiwatar da tsarin rarraba DM a cikin Muhalli na Nazarin Ilimi na Waikato da kuma dandalin nazarin bayanai na Konstanz (KNIME) 4.6.152.
Marubutan sun yarda cewa ana iya samun bayanan da ke goyon bayan sakamakon binciken a cikin labarin da kuma ƙarin kayan aiki. Bayanan da aka samar da/ko aka yi nazari a kansu yayin binciken suna samuwa daga marubucin da ya dace bisa buƙata mai ma'ana.
Ritz-Timme, S. da sauransu. Kimanta shekaru: yanayin zamani don biyan takamaiman buƙatun aikin binciken laifuka. ƙasashen duniya. J. Likitan Shari'a. 113, 129–136 (2000).
Schmeling, A., Reisinger, W., Geserik, G., da Olze, A. Matsayin da ake ciki a yanzu na tantance shekarun masu binciken laifuka na shari'a don dalilan shari'a. Bincike. magani. Cututtuka. 1, 239–246 (2005).
Pan, J. et al. Hanya da aka gyara don tantance shekarun hakora na yara 'yan shekara 5 zuwa 16 a gabashin China. na asibiti. Binciken baki. 25, 3463–3474 (2021).
Lee, SS da sauransu. Tarihin ci gaban haƙora na biyu da na uku a cikin 'yan Koriya da aikace-aikacensa don kimanta shekarun bincike. ƙasashen duniya. J. Likitan Shari'a. 124, 659–665 (2010).
Oh, S., Kumagai, A., Kim, SY da Lee, SS Daidaiton kiyasin shekaru da kimantawa na ƙa'idar shekaru 18 bisa ga girman haƙoran biyu da na uku a cikin Koriya da Japan. PLoS ONE 17, e0271247 (2022).
Kim, JY, da sauransu. Binciken bayanai na koyon injin kafin tiyata na iya hasashen sakamakon maganin tiyatar barci ga marasa lafiya da ke fama da OSA. kimiyya. Rahoton 11, 14911 (2021).
Han, M. da sauransu. Daidaiton shekaru daga koyon injina ko ba tare da sa hannun ɗan adam ba? Ƙasashen duniya. J. Likitan Shari'a. 136, 821–831 (2022).
Khan, S. da Shaheen, M. Daga Haƙar Bayanai zuwa Haƙar Bayanai. J. Bayani. kimiyya. https://doi.org/10.1177/01655515211030872 (2021).
Khan, S. da Shaheen, M. WisRule: Tsarin Fahimta na Farko don Haƙar Dokokin Haɗaka. J. Bayani. kimiyya. https://doi.org/10.1177/01655515221108695 (2022).
Shaheen M. da Abdullah U. Karm: Haƙar bayanai na gargajiya bisa ƙa'idodin haɗin gwiwa bisa ga mahallin. ƙididdige. Matt. ci gaba. 68, 3305–3322 (2021).
Muhammad M., Rehman Z., Shaheen M., Khan M. da Habib M. Gano kamanceceniya mai zurfi bisa ga ma'anar ma'ana ta amfani da bayanan rubutu. sanarwa. fasaha. sarrafawa. https://doi.org/10.5755/j01.itc.49.4.27118 (2020).
Tabish, M., Tanoli, Z., da Shahin, M. Tsarin gane ayyuka a bidiyon wasanni. multimedia. Aikace-aikacen Kayan aiki https://doi.org/10.1007/s11042-021-10519-6 (2021).
Halabi, SS da sauransu. Kalubalen Koyon Inji na RSNA a Zamanin Ƙashi na Yara. Radiology 290, 498–503 (2019).
Li, Y. da sauransu. Kimanta shekarun binciken asibiti daga hasken X-ray na ƙugu ta amfani da zurfin koyo. EURO. hasken rana. 29, 2322–2329 (2019).
Guo, YC, da sauransu. Daidaitaccen rarrabuwar shekaru ta amfani da hanyoyin hannu da hanyoyin sadarwa na jijiyoyi masu zurfi daga hotunan hasashen orthographic. ƙasa da ƙasa. J. Likitan Shari'a. 135, 1589–1597 (2021).
Alabama Dalora da sauransu. Kimanta shekarun ƙashi ta amfani da hanyoyi daban-daban na koyon injin: bitar adabi mai tsari da kuma nazarin meta-analysis. PLoS ONE 14, e0220242 (2019).
Du, H., Li, G., Cheng, K., da Yang, J. Kimanta shekarun da suka shafi yawan jama'a na Baƙar fata 'Yan Amurka da Sinawa bisa ga yawan ɗakin pulp na haƙoran farko ta amfani da hoton kwamfuta mai siffar cone-beam. ƙasa da ƙasa. J. Legal medicine. 136, 811–819 (2022).
Kim S., Lee YH, Noh YK, Park FK da Oh KS Ƙayyade ƙungiyoyin shekaru na mutane masu rai ta amfani da hotunan haƙoran farko da aka dogara da hankali ta hanyar wucin gadi. Kimiyyar. Rahoton 11, 1073 (2021).
Stern, D., Payer, C., Giuliani, N., da Urschler, M. Kimanta shekaru ta atomatik da rarrabuwar shekaru mafi rinjaye daga bayanan MRI masu canzawa da yawa. IEEE J. Biomed. Faɗakarwar Lafiya. 23, 1392–1403 (2019).
Cheng, Q., Ge, Z., Du, H. da Li, G. Kimanta shekaru bisa ga rarrabuwar ɗakunan ɓawon 3D na haƙoran farko daga ƙirar mazugi ta hanyar haɗa zurfin koyo da matakan saiti. ƙasa da ƙasa. J. Likitan Shari'a. 135, 365–373 (2021).
Wu, WT, da sauransu. Haƙar bayanai a cikin manyan bayanai na asibiti: bayanai na gama gari, matakai, da samfuran hanyoyi. Duniya. magani. albarkatu. 8, 44 (2021).
Yang, J. da sauransu. Gabatarwa ga Bayanan Lafiya da Fasahar Haƙar Bayanai a Zamanin Babban Bayanai. J. Avid. Magungunan asali. 13, 57–69 (2020).
Shen, S. da sauransu. Hanyar Camerer don kimanta shekarun haƙori ta amfani da na'urar koyon injin. BMC Oral Health 21, 641 (2021).
Galliburg A. da sauransu. Kwatanta hanyoyin koyon injina daban-daban don hasashen shekarun haƙori ta amfani da hanyar tsara Demirdjian. ƙasa da ƙasa. J. Likitan Shari'a. 135, 665–675 (2021).
Demirdjian, A., Goldstein, H. da Tanner, JM Sabuwar tsarin tantance shekarun haƙori. snort. ilmin halitta. 45, 211–227 (1973).
Landis, JR, da Koch, GG Ma'aunin yarjejeniyar masu lura kan bayanai na rukuni. Biometrics 33, 159–174 (1977).
Bhattacharjee S, Prakash D, Kim C, Kim HK da Choi HK. Nazarin rubutu, tsari da kididdiga na hoton maganadisu mai girma biyu ta amfani da dabarun basirar wucin gadi don bambance manyan ciwace-ciwacen kwakwalwa. Bayanin lafiya. albarkatu. https://doi.org/10.4258/hir.2022.28.1.46 (2022).
Lokacin Saƙo: Janairu-04-2024