Enkosi ngokundwendwela indalo.com. Inguqulelo yesikhangeli usebenzisa inkxaso encinci css. Ngenxa yeziphumo ezilungileyo, sicebisa ukuba sisebenzise ingxelo entsha yesikhangeli sakho (okanye icime indlela yokuhambelana kwi-Intanethi Explorer). Okwangoku, ukuqinisekisa inkxaso eqhubekayo, sibonisa indawo ngaphandle kokuzimisela okanye iJavaScript.
Amazinyo athathwa njengolona phawu luchanekileyo lobudala bomzimba womntu kwaye bahlala besetyenziswa kuvavanyo lweminyaka yobudala. Sijolise ekuqinisekiseni ngoqikelelo lwamazinyo olwenziwe ngendawo ngokuthelekisa ngokuthelekisa ukuchaneka koqikelelo lweminyaka eli-18 yeminyaka eli-18 ngeendlela zemveli kunye noqikelelo lwedatha. Zizonke ii-ranography ezingama-2657 zePanoramic zaqokelelwa kwiKorea nabemi baseJapan baneminyaka eli-15 ukuya kweli-23. Bahlulahlulwe baba yiseti yoqeqesho, nganye iqulethe i-900 i-900 i-900 ye-radiographs, kunye neseti yovavanyo lwangaphakathi oluqukethe i-radimassicals yamaJapan. Sithelekise ukuchaneka kokucwangciswa nokusebenza kakuhle kweendlela zemveli ezineeseti zovavanyo lweemodeli zemigodi yedatha. Ukuchaneka kwendlela yesiqhelo kuvavanyo lwangaphakathi lweseti lingaphezulu kakhulu kuneyona modeli yemigodi yedatha, kwaye umahluko mncinci (i-0.21 IYAMBULA IXESHA LOKUGQIBELA). Intsebenzo yokwahlulahlula-hlula kweminyaka eli-18 yeminyaka ifana nezezibini zemveli kunye neemodeli zemigodi yemigodi. Ngenxa yoko, iindlela zendabuko zinokutshintshwa ziimodeli zemigodi yedatha xa usenza uvavanyo lobudala be-forensic usebenzisa ukuvuthwa kwesibini neyesithathu kwi-Relelen Relelescents kunye nabantu abancinci.
Uqikelelo lwamanqanaba amazinyo lusetyenziswe ngokubanzi kumayeza obunini bamanye amazwe kunye namazinyo e-Pediatric. Ngokukodwa, ngenxa yokunxibelelana okuphezulu phakathi kobudala bexesha lokungasebenzi kunye namazinyo, uvavanyo lobudala ngamanqanaba ophuhliso lwamazinyo yindlela ebalulekileyo yokuvavanya ubudala babantwana kunye neshumi elivisayo. Nangona kunjalo, i-labantu abancinci, uqikelela ukuba kwinqanaba lokukhula kwamazinyo lunemida kuba ukukhula kwamazinyo kuphelile, ngaphandle kwemolari yesithathu. Injongo esemthethweni yokumisela iminyaka yolutsha kunye nolutsha kukubonelela ngoqikelelo oluchanileyo kunye nobungqina bezenzululwazi bokuba bafikile kwiminyaka yobuninzi. Kwi-Medicine-Intstion yokufikisa kwabakwishumi elivisayo kunye nabantu abadala baseKorea, ubudala baqikelelwa besebenzisa indlela kaLee, kunye nomgaqo osemthethweni weminyaka eyaxelwa kwangaphambili kwi-OH et al 5.
Ukufundwa komatshini luhlobo lobukrelekrele bobugcisa (i-AI) ephinda-phinda i-vele ifunda kwaye iqulunqa inani elikhulu ledatha, icombulula iingxaki ngokwayo, kwaye iqhuba inkqubo yedatha. Ukufunda umatshini unokufumana iipateni ezifihliweyo ezifihliweyo kwimithamo emikhulu yedatha6. Ngokwahlukileyo, iindlela zeklasikhi, ezichitha kakhulu kwaye zichitha ixesha, zinokuphelisa ixesha xa kusenziwa imiqulu emikhulu yedatha enzima ekunzima ukuyiqhubela ngesandla7. Ke ngoko, izifundo ezininzi zenziwe kutshanje kusetyenziswa itekinoloji yekhompyuter yakutsha nje yokunciphisa iimpazamo zabantu kunye nenkqubo efanelekileyo yokuqhubela phambili idatha ye-Multidimel And8,11,11,11,11,11,11,11,11,11,13. In particular, deep learning has been widely used in medical image analysis, and various methods for age estimation by automatically analyzing radiographs have been reported to improve the accuracy and efficiency of age estimation13,14,15,16,17,18,19,20 . Umzekelo, uHalabi et al 13 wenza umatshini wokufunda i-algorithm esekwe kwinethiwekhi yokungcola (i-CNN) yokuqikelela ubudala be-skeletal. Olu phononongo luphakamisa imodeli esebenza ngokufunda kwimifanekiso yezonyango kwaye ibonisa ukuba ezi ndlela zinokuphucula ukuchaneka kovavanyo. I-li et al44 eqikelelwa kwi-Polvic X-ray isebenzisa i-CNN yokufunda i-CNN kwaye ithelekiswa neziphumo zoMsebenzi woMsebenzi kusetyenziswa uqikelelo lwe-Ossation. Bafumanise ukuba imodeli yokufunda i-CNN ibonise intsebenzo efanayo yokuqikelela njengemodeli yophando lwemveli. IGuo et al. Iphonononge [15] iphonononge ukuSebenza kokuSebenza kweTND yeTekhnoloji esekwe kwii-orthophotos zamazinyo, kwaye iziphumo zemodeli ye-CNN ziqondakalanga ukuba abantu baphunyezelela ukusebenza kwenqanaba lokwahlulahlula.
Uninzi lwezifundo ngoqikelelo lweminyaka yobudala lusebenzisa indlela yokufunda ngoomatshini Uqikelelo lobudala olusekwe ekufundeni olunzulu luxelwe ukuba luchaneke ngaphezu kweendlela zemveli. Nangona kunjalo, le ndlela inika ithuba elincinci lokubonisa isiseko sesayensi soqikelelo lobudala, njengezalathi zeminyaka esetyenziswa kuqikelelo. Kukho imbambano yomthetho malunga nokuba ngubani oqhuba uhlolo. Ngenxa yoko, ukuqikelela okusekwe ekufundeni okunzulu kunzima ukwamkela abasemagunyeni kwezolawulo. Ukumbiwa kwedatha (DM) yindlela enokufumanisa ukuba ayilindelekanga kuphela kodwa ulwazi olungalindelekanga njengendlela yokufumanisa ukulungiswa okufanelekileyo phakathi kwedatha enkulu yedatha yedatha ephakathi kwedatha enkulu yedatha yedatha ephakathi kwedatha enkulu yedatha yedatha ephakathi kwedatha enkulu yedatha. Ukufundwa komatshini kudla ngokusetyenziswa kwimigodi yedatha, kwaye zombini imveliso yedatha kunye nokufunda umatshini Sebenzisa i-algorithms efanayo yokufumanisa iipateni kwidatha. Uqikelelo lweminyaka yobudala usebenzisa uphuhliso lwamazinyo lusekwe kuvavanyo lomvavanyi wokukhula kwamazinyo ekujoliswe kuwo, kwaye olu vavanyo lubonakaliswa njengenqanaba lezinyo ngalinye. I-DM ingasetyenziselwa ukuhlalutya unxibelelwano phakathi kwenqanaba lovavanyo lwamazinyo kunye neminyaka eyiyo kwaye inakho ukutshintsha i-staltisticial stalticial. Ke ngoko, ukuba sisebenzisa ubuchwephesha be-DM ukuya kuqikelelwa komatshini, sinokuphumeza ukufunda komatshini kuqikelelo lobudala beminyaka yobudala ngaphandle kokukhathazeka malunga netyala elisemthethweni. Izifundo ezininzi zokuthelekisa zipapashiwe kwezinye iindlela ezinokubakho kwiindlela zemveli zemveli ezisetyenziswa kukhetho lwe-forensic kunye neendlela ezisekwe kwi-EBM zokumisela iminyaka yamazinyo. I-Shen El23 ibonise ukuba imodeli ye-DM ichanekile ngakumbi kune-foourser yemveli. I-Gailourg et Al24 isebenzise iindlela ezahlukeneyo ze-DM ukuqikelela ukuba nenqanaba le-Dederdjian25 kwaye iziphumo zibonise iindlela ze-DM zaza zangaphezulu kwexesha labemi baseFransi.
Ukuqikelela ubudala bamazinyo abakwishumi elivisayo kunye nabantu abadala abancinci, indlela ye-LEE 4 isetyenziswe ngokubanzi ngokuziqhelanisa neKorea. Le ndlela isebenzisa uhlalutyo lwesiko (ezinjenge-regress ezininzi) ukuvavanya ubudlelwane phakathi kwezifundo zaseKorea kunye nobudala bokungasebenzi. Kolu phononongo, iindlela zoqikelelo lweminyaka yokuqikelela ifunyenwe kusetyenziswa iindlela zenkcubeko zemveli zichazwa njenge "iindlela zemveli." Indlela ye-Lee yindlela yemveli, kwaye ukuchaneka kwayo kuqinisekisiwe ngo-OH et al. 5; Nangona kunjalo, ukwenziwa koqikelelo lobudala ngokusekwe kwimodeli ye-DM kwi-Korea ye-Korea forensic kusenokuthandabuza. Injongo yethu yayikukuqinisekisa ngokwazisa ngoqikelelo lobudala obusekwe kwimodeli ye-DM. Injongo yolu phononongo yayi (1) ukuthelekisa ukuchaneka kweemodeli ezimbini ze-DM ukuqikelela iminyaka eyi-7 ye-DM eneminyaka eli-18 kunye nezo zifudumeleyo zeSibini kunye nesithathu iingozi kwimihlathi yomibini yemihla ngemihla.
Kuthetha ukupheliswa komgangatho wobudala bexesha lokulinganisa ngokulandelelana kunye nohlobo lwezinyo kuboniswa kwi-intanethi kwiTheyibhile yeS1 (iseti yoqeqesho), kunye netheyibhile yovavanyo lwangaphakathi ye-S3 (iseti yovavanyo lwangaphandle). Ixabiso le-kappa le-Intra- kunye ne-metrobserver ngokuthembeka efunyenwe kwiseti yoqeqesho yayiyi-0.951 ne-0.947, ngokulandelelana. Iimpawu zexabiso kunye ne-95% yokuzithemba kwamaxabiso e-Kappa zibonisiwe kwitafile eyongezelelweyo ye-Intanethi s4. Ixabiso le-kappa latolikwa njenge "phantse ligqibelele", lingangqinelani neekhrayitheriya ze-hissis naseKoch26.
Xa uthelekisa impazamo epheleleyo (i-mae), indlela yendabuko iphinda isebenze nje imodeli ye-DM kubo bonke abakhweli kunye novavanyo lwangaphandle lwamadoda, i-MLP). Umahluko phakathi kwemodeli yesiko kunye nemodeli ye-DM kwi-MEM yangaphakathi yokuSeta yayiyiminyaka eyi-0.12-0.19 yeminyaka kunye ne-0.17-0.21 iminyaka yabasetyhini. Kwibhetri yovavanyo lwangaphandle, umahluko mncinci (0.001-0.05 iminyaka yamadoda kunye ne-0.05-0.09 yabasetyhini). Ukongeza, ingcambu ithetha impazamo yesikwere (i-RMSE) iphantsi kancinci kunendlela yemveli, ngomahluko omncinci (0.17-.4), 0.03-0.08 yeseti yovavanyo lwangaphandle). ). I-MLP ibonisa intsebenzo engcono kakhulu kune-facept veretron (i-SLP), ngaphandle kwemeko yovavanyo lwangaphandle lwentokazi. KwiMae naseRMse, uvavanyo lwangaphandle Yonke i-mae kunye ne-RMse iboniswe kwiTheyibhile 1 kunye nomfanekiso 1.
I-Mae kunye ne-RMse yemodeli yendabuko yemveli kunye nedatha. Ithetha impazamo epheleleyo Mae, ingcambu ithetha impazamo yesikwere i-RMSE, i-famitptnl slp, i-multilayent ye-Multilayent, indlela ye-CM ye-CM.
Intsebenzo yokwahlulahlula-hlula (kunye ne-cutoff yeminyaka eli-18) yemodeli yemveli neye-DM ibonakalisiwe ngokwemvakalelo, kwixabiso elifanelekileyo, ixabiso eliqinisekileyo (i-PPV), kunye nendawo ephantsi kwendawo yokusebenza (i-Auroc) 27 (Itheyibhile 2, umzobo 2 kunye nesongezelelo somzobo 1 kwi-intanethi). Ngokwemvakalelo yengqondo yovavanyo lwangaphakathi, iindlela zemveli ziyenziwa kakuhle ebantwini kwaye zisiba mbi phakathi kwabasetyhini. Nangona kunjalo, umahluko kwintsebenzo yokuhlelwa phakathi kweendlela zemveli kunye ne-SD yi-9.7% yamadoda (MLP) kunye ne-2.4% kuphela yabasetyhini (i-xgboost). Phakathi kweemodeli ze-DM, i-regring yelondolozwe (LR) ibonise ubuntununtunu obungcono kuzo zombini izini. Ngokuphathelele ubunyani beseti yovavanyo lwangaphakathi, yaqwalaselwa ukuba iimodeli ze-SD ezine zenziwa kakuhle emadodeni, ngelixa imodeli yendabuko eyenzi ngcono kwabasetyhini. Umahluko kwintsebenzo yokuhlelwa kwamadoda nabasetyhini bane-13.3% (i-MLP) kunye ne-13.1% (i-MLP), ebonisa ukuba umahluko kwi-model upend igqibelele uvakalelo. Phakathi kweemodeli ze-DM, umatshini we-vector yenkxaso (i-SVM), umthi wesigqibo (i-DT), kunye nehlathi elingenamsebenzi (i-RF) lenziwa kakuhle phakathi kwamadoda, ngelixa imodeli ye-LR yenziwe kakuhle phakathi kwabasetyhini. I-Auroc yemodeli yesiNtu kwaye zonke iimodeli ze-SD zazinkulu kune-0.925 (i-k-ekufuphi (i-k-ekufuphi), ebonisa ukusebenza kakuhle kwe-188. Kuvavanyo lovavanyo lwangaphandle, bekukho ukwehla kokusebenza kwendlela yokuhlelwa ngokwemvakalelo, i-auroc xa kuthelekiswa neseti yovavanyo lwangaphakathi. Ngaphaya koko, umahluko kwimvakalelo kunye nobungqina phakathi kokusebenza kwendlela yokwahlulahlulahlula izinto ezilungileyo nezimbi kakhulu ukusuka kwi-10% ukuya kwi-25% kwaye yayinkulu kunonyaka wokusetyenziswa kwangaphakathi.
Imvakalelo kunye nobuchule beemodeli zokudibanisa idatha zokuhlelwa xa kuthelekiswa neendlela zemveli nge-cutoff yeminyaka eli-18. I-vector vector ye-vector, i-LR Loglonces, umthi we-DT, i-RF yeHlabathi, i-xf xgboost, i-MLP yendlela.
Inyathelo lokuqala kolu phononongo yayikukuthelekisa ukuchaneka koqikelelo lwamazinyo afunyenwe kwiimodeli ezisixhenxe ze-DM kunye nezo zifunyenweyo zisebenzisa uReger Gratsion. I-Mae kunye ne-RMse yavavanywa kwiiseti zovavanyo lwangaphakathi lwesibini, kwaye umahluko phakathi kwendlela yemveli kunye nemozulu ye-DM yaqala ukusuka kwiintsuku ezingama-44 ukuya kwezingama-77 ze-RMse. Nangona indlela yendabuko yayichanekile ngakumbi kolu phando, kunzima ukugqiba kwelokuba loo mahluko mncinci enobukrelekrele okanye ukubaluleka okusebenzayo. Ezi ziphumo zibonisa ukuba ukuchaneka koqikelelo lwamanqanaba amahlakani usebenzisa imodeli ye-DM kufike ngokufanayo nendlela yemveli. Ukutsala ngokuthe ngqo neziphumo ezivela kwizifundo zangaphambili kunzima kuba akukho kuphonononge ukuchaneka kweemodeli ze-DM ezineendlela zemveli ezisebenzisa amazinyo ashicilelweyo kuluhlu olufanayo nakweli phando. I-Gailourg et Al24 I-MAE ethengiswayo kunye ne-RMSE phakathi kweendlela ezimbini zemveli (Deirjian Pmell25 kunye ne-Wilems25) kunye ne-10 dm imodeli yeminyaka eyi-4 ukuya kwengama-24. Bayichazile ukuba zonke iimodeli ze-DM zazichanekile ngakumbi kunendlela zemveli, ekwahlukileyo kwe-0.20 kunye ne-0.25 yeminyaka e-Mae kunye ne-0.47 kunye neendlela ze-Dedeji, ngokwahlukeneyo. Ukungangqinelani phakathi kwemodeli ye-SD kunye neendlela zemveli eziboniswe kwisifundo se-halibourg ithathela ingqalelo iingxelo ezininzi ze-deligjiations eziphezulu kubantu baseFransi abasekwe kulo isifundo. Kolu phando. I-Tai et al 34 isebenzise i-MLP Algorithm yokuqikelela ubudala ukusuka kwi-1636 yeefoto ze-chinedontic zaseTshayina kwaye kuthelekiswa nendlela yayo ye-Deviljian kunye ne-Wilems. Baxela ukuba i-MLP inobungqiwa obuphezulu kunendlela yemveli. Umahluko phakathi kwendlela ye-Dedeji kunye nendlela yesiqhelo yi <0.32 iminyaka, kwaye indlela ye-Willem yi-0.28 iminyaka, efana neziphumo zophando. Iziphumo zale zifundo zangaphambili24,34 ziyahambelana neziphumo zophando, kunye nokuchaneka koqikelelo lweminyaka yemodeli ye-DM kunye nendlela yesiqhelo ziyafana. Nangona kunjalo, ngokusekelwe kwiziphumo ezibonisiweyo, sinokugqiba ngokupheleleyo ukuba ukusetyenziswa kweemodeli ze-DM ukuya kwiminyaka yobudala kunokutshintsha iindlela ezikhoyo ngenxa yokunqongophala kwezifundo zangaphambili. Izifundo ezilandelwayo zisebenzisa iisampulu ezinkulu ziyafuneka ukuqinisekisa iziphumo ezifunyenwe kolu phando.
Phakathi kwezifundo kuvavanya ukuchaneka kwe-SD ekuqikeleleni ubudala bamazinyo, abanye babonisa ukuchaneka okuphezulu kunokufunda. I-Stepanovsky et al 35 Ithathwe kwiimodeli ze-SD ezingama-22 kwiirades ze-pinorac ezingama-976 zeCzech abahlali baseCzech ezineminyaka engama-2,7 ukuya kwengama-20,5 ubudala kwaye bavavanya ukuchaneka kwemodeli nganye. Bavavanya uphuhliso lwe-16 ephezulu kunye nezantsi kwecala elisekhohlo linamazinyo asisigxina esebenzisa iikhrayitheriya zokuhlelwa okhutshwe nguMoorrees et al 36. I-mae rangs ukusuka kwi-0.64 ukuya kwi-0.94 iminyaka kunye ne-RMSE ukusuka kwi-0.85 ukuya kwi-0.85 ukuya kwi-1.27 iminyaka emi-1,27, echanileyo kune-DM yeemodeli ezimbini ze-DM ezisetyenziswe kolu phando. Shen Et Al23 isebenzise indlela yokuqikelela ukuqikelela ubudala ngamazinyo asixhenxe asisigxina kumanqanaba amaTshayina ase-the ElaseSun 5 ukuya kwele-13 kwaye wathelekisa i-AVE Babonisa ukuba zonke iimodeli ze-DM ze-DM zinokuchaneka okuphezulu xa kuthelekiswa nefomula yemveli yendabuko. I-mae kunye ne-RMS kwiSifundo sikaShen sasingaphantsi kweso sifundo se-DM kolu phando. Ukonyusa ukuchaneka kwezifundo nguStepanovsky et al. 35 I-Shen et al. 23 Inokuba kungenxa yokubandakanywa kwezifundo ezincinci kwiisampulu zawo zokufunda. Kuba iqikelelo yobudala abathathi-nxaxheba abanamazinyo aphumeleleyo bachaneke ngakumbi njengoko inani lamazinyo linyuka ngexesha loqikelelo lwamazinyo linokuthathwa xa abathathi-nxaxheba abafundayo abancinci xa abathathi-nxaxheba abafundayo abancinci xa abathathi-nxaxheba abafundayo abancinci xa abathathi-nxaxheba abafundayo abancinci xa abathathi-nxaxheba abafundayo abancinci. Ukongeza, imposiso ye-MLP kuqikelelo lweminyaka yobudala incinci kakhulu kune-SLP, intsingiselo yokuba i-MLP ichaneke ngakumbi kune-SLP. I-MLP ithathelwa ingqalelo nje ingqikelelo yobudala, kungenzeka ukuba ngenxa yeendawo ezifihliweyo kwi-MLP38. Nangona kunjalo, kukho imeko engaphandle kwesampulu yangaphandle yabasetyhini (SLP 1.45, MLP 1.49). Ukufumana ukuba i-MLP ichaneke ngakumbi kune-SLP kwi-ADP kwi-Agespen ifuna izifundo ezongeziweyo.
Intsebenzo yokuqhekezwa kwemodeli ye-DM kunye nendlela yesiqhelo kumbundu weminyaka eli-18 wawuthelekiswa nalo. Zonke iimodeli ze-SD ze-SD kunye neendlela zendabuko kuvavanyo lwangaphakathi olubonakalise amanqanaba amkelekileyo uqheliselo olwenziwe ngocalulo lwesampulu yeminyaka eli-18 ubudala. Imvakalelo yamadoda nabafazi yayinkulu kune-87.7% kunye ne-94.9%, ngokulandelelana, kunye ne-89.3% ne-84.7%. I-Auroc yazo zonke iimodeli ezivavanyiweyo ikwadlula i-0.925. Ngokolwazi lwethu, akukho phando luvavanye ukwenziwa kweMfula yeMpilo ye-DM yeminyaka eli-18 yokwahlulahlula ukukhula kwamazinyo. Sinokuthelekisa iziphumo zolu phononongo ngokwenza umsebenzi wokuhlelwa kweemodyuli zokufunda ezinzulu kwii-ranographys zephanoramic. I-GUO Et Al.15 ibalile ukusebenza kwenkqubo yokufunda esekwe kwi-CNN kunye nendlela yesandla esekwe kwindlela ka-Devirjian kumgangatho othile wobudala. Imvakalelo kunye nokuthintelwa kwendlela ye-sumas yayingama-87% kwaye i-95.5%, ngokulandelelana, kunye nemvakalelo kunye nemvakalelo kunye nemvakalelo ye-CNN engaphezulu kwe-89.6%, ngokwahlukeneyo. Baye bagqiba kwelokuba iimodeli zokufunda ezinzulu zinokuyitshintsha okanye ziphunyezwe ekuhlolweni kwezandla kwi-Deridation Rures. Iziphumo zolu phononongo zibonise ukusebenza ngendlela efanayo; Kukholelwa ukuba ukwahlulahlulwa kusetyenziswa imodeli ye-DM kunokutshintsha iindlela zendabuko zoqikelelo lobudala. Phakathi kweemodeli, i-DM lr yayiyimodeli elungileyo ngokwemvakalelo yesampulu yendoda kunye nemvakalelo kunye nobungqina besampulu yabasetyhini. I-LR RUSK yesibini ngokubhekisele kumadoda. Ngaphezu koko, i-LR ithathwa njengenye yeemodeli ze-DM35 ezinobuhlobo kwaye ayinantsokothileyo kwaye kunzima ukuyiqhubela phambili. Ngokusekwe kwezi ziphumo, lr waqwalaselwa njengemodeli ye-creaff ye-creaff yeminyaka eli-18 ubudala kubantu baseKorea.
Ngokubanzi, ukuchaneka koqikelelo lweminyaka yobudala okanye ukusebenza kokuhlelwa kwiseti yovavanyo lwangaphandle kwakuhlwempuzekile okanye ezantsi xa kuthelekiswa neziphumo zovavanyo lwangaphakathi. Ezinye iingxelo zibonisa ukuba ukuchaneka kokucwangciswa okanye ukuhla kwempumelelo kwendlela yokuqikelela kubudala ngokusekwe kuluntu lwaseKorea5,39, kwaye kwafunyanwa imeko efanayo kwimodeli yangoku. Lo mkhwa ugqithisileyo uqatshelwe kwimodeli ye-DM. Ke ngoko, ukulungiselela ubudala obuchanekileyo, kwanaxa usebenzisa i-DM kwinkqubo yohlalutyo, iindlela ezivela kwidatha yendawo yohlalutyo, ezinjengeendlela zemveli, kufuneka zikhethiwe5,39,41,41,42. Kuba akucacanga ukuba iindlela zokufunda ezinzulu zinokubonisa iindlela ezifanayo, izifundo zithelekisa iindlela zemfundo, i-DM Models, kunye neendlela zokufunda ezikwiimodeli ezifanayo zinokuqinisekisa ukuba olu lwandla lungaphantsi komhlaba. Uvavanyo.
Sibonisa ukuba iindlela zemveli zinokutshintshwa ngoqikelelo lobudala ngokusekwe kwimodeli ye-DM kwimidlalo yoqikelelo lweminyaka yobudala kwi-Korea. Sikwafumanise ukubakho kokufumana ukufunda umatshini wokuphumeza kovavanyo lobudala beminyaka yobudala. Nangona kunjalo, kukho umda ocacileyo, efana nenani elingananke abathathi-nxaxheba kolu phononongo ukuze baqwalasele iziphumo, kwaye ukunqongophala kwezifundo zangaphambili ukuthelekisa kunye nokuqinisekisa iziphumo zolu phononongo. Kwixesha elizayo, izifundo ze-DM kufuneka zenziwe ngamanani amakhulu eesampulu kunye neesakhiwo ezahlukeneyo zokuphucula iindlela zayo ezisetyenziswayo xa kuthelekiswa neendlela zemveli. Ukuqinisekisa ukuba kungenzeka ukuba kusetyenziswe ubugqwetha ubudala bokuqikelela ukuba abantu abaninzi, kufuneka bathelekise ukuchaneka kokucwangciswa nokusebenza kwe-DM kunye neendlela zokufunda ezinzulu ngeendlela zemveli kwiindlela zemveli.
Olu phando lusebenzise imifanekiso ye-2,657 i-orthographic eqokelelwe eKorea kunye nabaseJapan abadala abaneminyaka eli-15 ukuya kwengama-23. Iirally zaseKorea zahlulahlulwa zangama-900 ezintlangeni (19.42 ± 2,65) iiseti zovavanyo lwangaphakathi ezingama-900 (19.52). I-Stoppit iseti yaqokelelwa kwiziko elinye (saseSeoul St. Siqokelele i-radiography ezingama-857 ukusuka kwenye idatha esekwe kuluntu (i-Isand University eyunivesithi, eJapan) yovavanyo lwangaphandle. I-Radiographys yezifundo zaseJapan (19.31 ± 2.6 Iminyaka) ikhethiwe njengovavanyo lwangaphandle. Idatha eqokelelweyo iphindaphinde ihlalutye amanqanaba ophuhliso lwamazinyo kwiiradegrafi ezithathwe ngexesha lonyango lwamazinyo. Yonke idatha eqokelelwe yayingaziwa ngaphandle kwesini, umhla wokuzalwa kunye nomhla we-radiograph. Ukubandakanywa kunye neekhrayitheriya zokukhokelwa kwakufana nezifundo ezipapashiweyo 4, 5. Iminyaka yobudala yesampulu ibalwa ngokukhupha umhla wokuzalwa ukusuka kumhla irediyo ethathiweyo. Iqela lesampulu lahlulahlulwe langamaqela alithoba. Iminyaka yobudala kunye nezesondo ziboniswa kwiTheyibhile 3 olu phononongo lwenziwa ngokungqinelana neBhodi ye-Helsinki kwaye yamkelwe yiBhodi yeZiko leSoul St. Ngenxa yokuyilwa okubuyileyo kolu phononongo, imvume eyazisiweyo ayinakufumaneka kuzo zonke izigulana ezijonga uviwo lwe-radiographic kwiinjongo zonyango. IYunivesithi yaseSeoul Korea yeYunivesithi yaseStAl St.
Amanqanaba okuphuhlisa i-bimaxillary yesibini neyesithathu kwavavanywa ngokwemigaqo ye-Demircan25. Kukhethwe izinyo elinye kuphela ukuba uhlobo olufanayo lufumaneka kumacala asekhohlo nasekunene kwihlathi yomhlathi. Ukuba amazinyo e-homologous kumacala omabini ayengamanqanaba ahlukeneyo ophuhliso, izinyo elinenqanaba eliphantsi lophuhliso lakhethwa ukuba likhuphe ukungaqiniseki kubudala obuqikelelweyo. Ikhulu le-radiography ekhethiweyo ngokungekho mthethweni kwiseti yoqeqesho lafunyanwa amanqaku amabini anamava ukuvavanya ukuthembeka kwenterportver emva kokumisela ukukhula komzimba. Ukuthembeka kwe-intraobserver kuvavanywa kabini kwizithuba zenyanga ezintathu ngoyena mntu ubukeleyo.
Inqanaba lezesondo kunye nenqanaba lesibini nelesithathu kwi-ralal yomhlathi nganye kwiiseti zoqeqesho zaqikelelwa ngumboniso ophambili we-DM, kwaye ukwahluka kwasetwa njengexabiso ekujoliswe kuzo. Iimodeli ze-SLP kunye ne-MLP, ezisetyenziswa ngokubanzi ngokufunda, zavavanywa ngokuchasene ne-algorithms yombuso. Imodeli ye-DM idibanisa imisebenzi ye-DM usebenzisa amanqanaba ophuhliso lwamazinyo amane kwaye idibanise ezi datha ukuba ziqikelele ubudala. I-SLP yinethiwekhi elula kwaye ayinamanye amazwe afihliweyo. Umsebenzi we-SLP ngokusekwe kumsebenzi oshishiweyo phakathi kweendawo. Imodeli ye-SLP kwiRetroligy ifana nemathematics kwi-trimoring yelayini emininzi. Ngokungafaniyo nemodeli ye-SLP, imodeli ye-MLP ineendawo ezininzi ezifihliweyo ezinemisebenzi yokusetyenziswa kwe-nonline. Uvavanyo lwethu lusebenzise ungqimba olufihliweyo kunye neendawo ezingama-20 kuphela ezifihliweyo ngemisebenzi yokuvula i-nonliar. Sebenzisa umnqweno we-graent njengeyona ndlela yokuphucula kunye ne-mae kunye ne-RMse njengoko ilahleko ilahleko ukuqeqesha imodeli yokufunda. Eyona mfuno ifunyenweyo isetyenzisiwe kwiiseti zovavanyo lwangaphakathi nangaphandle kunye neminyaka yamazinyo yaqikelelwa.
I-algorithm ye-algorithm yaphuhliswa esebenzisa ukuqola kwamazinyo amane kuqeqesho olusetwe lokuqikelela ukuba isampulu ineminyaka eli-18 ubudala okanye hayi. Ukwakha imodeli, sithabatha umatshini osixhenxe wokumelwa i-algorithms6,43: (1) lr, (2) i-SVM, (5) i-SVM, (5) i-RF, (7) ye-XGM . I-LR yenye yezona zinto zisetyenziswa kakhulu ze-algorithms44. Yi-algorithm eyongamileyo esetyenziselwa imbandezelo yokuqikelela ukuba nedatha yedatha kudidi oluthile ukusuka kwi-0 ukuya kwi-1 kwaye ihlule idatha enokwenzeka ngokusekwe kule ndlela; ikakhulu isetyenziselwa ukwahlulahlula-bhanyabhanya. I-KNN yenye yezona ngcalu zilula zokufunda i-algorithms45. Xa unikwe idatha ye-Igalelo eNtsha, ifumana idatha ikufuphi neseti esele ikho kwaye emva koko ibalele kwiklasi kunye nesantya esiphezulu. Sibeka i-3 inani labamelwane abaqwalaselweyo (k). I-SVM yi-algorithm ekwandisa umgama phakathi kweeklasi ezimbini ngokusebenzisa umsebenzi we-kernel ukwandisa indawo esecaleni kwendawo engachazwanga. Kule modeli, sisebenzisa i-bias = 1, amandla = 1, kunye ne-gamma = 1 njenge-hyperparameter ye-pollynomial kernel ye-polynomial kernel ye-polynomial kernel kernel. I-DT isetyenzisiwe kwimimandla eyahlukeneyo njenge-algorithm yokwahlula i-ALTE datha esekwe kwiqela lemigangatho emininzi ngokumela imithetho yesigqibo kwiCandelo leMithi47. Imodeli ilungiselelwe ngenani elincinci leerekhodi nge-2 kwaye isebenzise isalathiso se-gini njengomgangatho womgangatho. I-RF yindlela evakalayo edibanisa i-DTS ezininzi ukuphucula indlela yokusebenza ye-bootstraP evelisa iisampulu ze-bootstrap ezivelisa iisampulu ze-datataset ezingama-408. We used 100 trees, 10 tree depths, 1 minimum node size, and Gini admixture index as node separation criteria. Ukwahlulahlula kwedatha entsha kugqitywa yivoti enkulu. I-Xgboost yi-algorithm edibanisa ubuchule bokusebenzisa indlela ethatha indawo yokuqeqesha impazamo phakathi kwemodeli edlulileyo kunye ne-aunters impazamo isebenzisa i-GRAED49. Yi-algorithm esetyenziswa ngokubanzi ngenxa yokusebenza kwayo elungileyo kunye nokusebenza kwezixhobo, kunye nokuthenjwa okuphezulu njengomsebenzi wokulungisa okungaphezulu. Imodeli ixhotywe ngamavili ama-400 enkxaso. I-MLP yinethiwekhi ye-NOLL apho i-Oneptudd nganye okanye ngaphezulu i-veeppedrons ezininzi zinezithombe zomntu omnye okanye ezingaphezulu phakathi kwegalelo kunye negalelo lamanzi38. Ukusebenzisa le nto, unokwenza udidi olungasebenziyo apho nini nini uqhawulo-mveliso kwaye ufumane ixabiso legalelo kwaye ufumane ixabiso lesiphumo, ixabiso lesiphumo lithelekiswa nexabiso lesiphumo kunye nempazamo ekusasazeni. Sidale intambo efihliweyo kunye nee-neurons ezifihliweyo ezingama-20 kwingqimba nganye. Imodeli nganye esaphuhlisa yasetyenziswa kwiiseti zangaphakathi nangaphandle kwintsebenzo yovavanyo lokucwangcisa ngokubala imvakalelo, ukuqina, i-PPV, i-NPV, ne-Auroc. Imvakalelo ichazwa njengomyinge wesampulu eqikelelweyo ukuba ibe yiminyaka eli-18 ubudala okanye ngaphezulu kwisampulu eqikelelweyo ukuba ibe yiminyaka eli-18 ubudala okanye ngaphezulu. Ubungqina luhlobo lweesampulu ezingaphantsi kweminyaka eli-18 kwaye ezo uqikelelwa ziphantsi kweminyaka eli-18 ubudala.
Amanqanaba okuvavanywa avavanywa kwiseti yoqeqesho aguqulwa abe ngamanqanaba amanani ohlalutyo lweenkcukacha-manani. I-Properiariate Commariate kunye ne-retrol yentshukumo yenziwa ukuba iphuhlise iimodeli zokuqikelela kwifomula nganye yesondo kunye nethatyathwayo ezinokusetyenziselwa ukuqikelela ubudala. Sasebenzisa ezi fomula ukuqikelela ubudala bezilwanyana zovavanyo lwangaphakathi nangaphandle. Itheyibhile 4 ibonisa i-remonces kunye neemodeli zokuhlelwa ezisetyenziswa kolu phando.
I-Intra- kunye ne-interportver yokuthembeka ibalwa isebenzisa i-kappa ka-Kappa. Ukuvavanya ukuchaneka kweemodeli zeDM kunye nemveli, sabala iMae nase RMSS sisebenzisa i-AURE eqikelelweyo kunye neminyaka yovavanyo lwangaphakathi nangaphandle. Ezi mpazamo zihlala zisetyenziselwa ukuvavanya ukuchaneka koqikelelo lwemodeli. Incinci imposiso, ukuphakama kokuchaneka koqikelelo24. Thelekisa i-mae kunye ne-RMse yeseti yovavanyo lwangaphakathi nangaphandle lokubalwa usebenzisa i-DM kunye ne-Retroloss. UkuSebenza kokuhlelwa kweminyaka eli-18 yokusikwa kweminyaka eli-18 kwiinkcukacha-manani zemveli bekuvavanywa kusetyenziswa itafile ye-2 × 2 ye-2. Imvakalelo yokubala, i-PPV, i-PPV, i-NPV, ne-Auroc yovavanyo ithelekiswa nexabiso elilinganayo lemodeli yokuhlelwa kwe-DM. Idatha ibonakaliswa njengokuthetha ± ukuphambuka komgangatho okanye inombolo (%) kuxhomekeke kwiimpawu zedatha. Amanqanaba amabini akwi-p <0.05 athathwa njengento eyimfuneko. Zonke izinto eziqhelekileyo zohlalutyo lwenziwa zisebenzisa inguqulelo ye-SAS 9.4 (SAS iZiko, eKhadani, NC). Imodeli yokubuyisela i-DM yamiliselwa kwi-censos50 2.2.4 Backend kunye ne-tenssorflow51 1.8.0. Imodeli yokuhlelwa kwe-DM yamiliselwa kwindawo yohlalutyo loLwazi lwe-Wanikato loLwazi kunye ne-Konstanz yemizi-mnyama (iKnstanz)
Ababhali bayavuma ukuba idatha exhasa izigqibo zophononongo zinokufumaneka kwinqaku nakwizixhobo ezongezelelweyo. Iidatha eziveliswe kwaye / okanye zihlaziywe ngexesha lokufunda ziyafumaneka kumbhali ohambelana nesicelo esifanelekileyo.
I-RITZ-Timeme, S. ET Al. Uvavanyo lobudala: Imeko yobugcisa yokufezekisa iimfuno ezithile zophando lwe-forensic. ubunye. J. Unyango lwezomthetho. I-113, 129-136 (2000).
I-Schmeling, A., Ukuvuselela, uWSEIK, G., no-OLze, imeko yangoku yovavanyo lweminyaka yokuphila yezifundo zokuzibandakanya kulwaphulo-mthetho. I-Fournsics. iyeza. I-pathology. 1, 239-246 (2005).
I-Pan, J. Et Al. Indlela eguqulweyo yokuvavanya iminyaka engamazinyo yabantwana abaneminyaka emi-5 ukuya kweli-16 kwimpuma yeChina. ikliniki. Uvavanyo lomlomo. 25, 3463-3474 (2021).
U-Lee, ss njl. ubunye. J. Unyango lwezomthetho. 124, 659-665 (2010).
Owu, S., IKim, A A., SSE, Sy kunye noqikelelo lweminyaka eli-18 ngokusekwe ekuqondeni kwesibini neyesithathu. I-Plos enye 17, E02717 (2022).
Kim, jy, et al. Uhlalutyo lwedatha ekoyikiweyo yovavanyo olusekwe kwidatha lunokuqikelela ukulala konyango lweziphumo kwizigulana nge-OSA. isayensi. Ingxelo 11, 14911 (2021).
Han, m. et al. Ubudala obuchanekileyo bobudala bokufunda kunye okanye ngaphandle kokungenelela komntu? ubunye. J. Unyango lwezomthetho. 136, 821-831 (2022).
Khan, S. NO Shaheen, M. ukusuka kwi-dataning yedatha yokufumana imigodi yedatha. I-J.Intermation. isayensi. I-https: //oi.org/20.1177/01655151510872 (2021).
Khan, S. NOSHY, M. Wisprule: I-algorithm yokuqala yengqondo yokunxulumana nemigodi. I-J.Intermation. isayensi. https://doi.org/21777/01655151108695 (2022).
I-Shaheeen M Bala. Mat. Qhubeka. 68, 3305322 (2021).
UMuhammad M., Rehman Z., shaheen m., khan M. kunye neHabib M. I-Habib M. Yazisa. Itekhnoloji yeTekhnoloji. ulawulo. I-https: //oi.org/0.5755/j01.itC.49.2.4.27118 (20208 (2020).
I-Tabish, M., Tanoli, Z., kunye neShahin, M. Inkqubo yokuqonda umsebenzi kwiividiyo zemidlalo. imultimedia. Izicwangciso zezixhobo ze-https:
I-HALABI, SS et al. Umngeni wokufunda umatshini we-RNNA kwi-Pediatric Bone. I-radiology 290, 498-503 (2019).
Li, y. et al. Uqikelelo lweminyaka yobudala evela kwi-VLVIC X-ray usebenzisa imfundo enzulu. I-Euro. imitha. 29, 2322-23229 (2019).
Guo, yc, et al. Ukuhlelwa kweminyaka echanekileyo usebenzisa iindlela ze-aulting kunye nenethiwekhi ezinzulu zokulungisa ezivela kwimifanekiso eqingqiweyo ye-ortographic. ubunye. J. Unyango lwezomthetho. 135, 1589-1597 (2021).
UAlabama Dalora Et Al. Uqikelelo lwe-Bone Lisebenzisa iindlela ezahlukeneyo zokufunda umatshini: Uphononongo loncwadi olucwangcisiweyo kunye nohlalutyo lweMeta. I-Plos i-14, e0220242 (2019).
UDu, H., Li, G., Cheng, K. ubunye. J. Unyango lwezomthetho. 136, 811-819 (2022).
Kim S., Lee Yh Yok, Park Fk kunye OH ks ukugqiba kweminyaka yobudala abantu abaphilayo abasebenzisa imifanekiso ye-ralations esekwe kwi-molations. isayensi. Ingxelo 11, 1073 (2021).
I-Drtern, D., intlawulo, c., giuliani, n., kunye ne-irchchler, i-MSCHLERE, i-MSCHLERE. IEEE J. Izilumkiso zezempilo. 23, 1392-1403 (2019).
UCyng, q. ubunye. J. Unyango lwezomthetho. 135, 365-373 (2021).
Wu, wt, et al. Ukumbiwa kwedatha kwiDatha yeKlinikhi: Iinkcukacha eziqhelekileyo, amanyathelo, kunye neendlela ezinemifanekiso. Umhlaba. iyeza. ubutyebi. 8, 44 (2021).
Yang, J. Et Al. Intshayelelo yoovimba bonyango kunye neetekhnoloji zemigodi yedatha kwixesha ledatha elikhulu. J. Avid. Iyeza elisisiseko. 13, 57-69 (2020).
Shen, S. ET Al. Indlela ye-Wormer yeXesha lokuqikelela lisebenzisa ukufunda umatshini. I-BMC Impilo Yomlomo 21, 641 (2021).
I-Galburg A. ET Al. Ukuthelekisa iindlela ngeendlela zokufunda ezahlukeneyo zokuqikelela iminyaka yamazinyo usebenzisa indlela ye-Dederdjian. ubunye. J. Unyango lwezomthetho. 135, 665-675 (2021).
UDedejian, A., igolide, H. kunye ne-Tanner, i-JM yenkqubo entsha yokuvavanya iminyaka yamazinyo. Snort. ibhayoloji. 45, 211-227 (1973).
I-Fiatis, JR, naseKoch, amanyathelo e-GG SEPUST SESIFUNDO SOKUGQIBELA KWI-DECATICARET. I-biometir 33, 159-174 (1977).
I-Bhattacharjee s, prakash D, kim c, kim hk kunye choi hk. Uhlalutyo olubhaliweyo, i-morphological kunye nobalo lweenkcukacha-manani lwemigangatho emibini yemagnetic icinga ukusebenzisa ubuchwephesha bobukrelekrele bokwahlulahlula amathumba obuchopho abaphambili. Ulwazi lwezeMpilo. ubutyebi. I-https: //oi.org/20.4258/hir.2022.28.2.2.2.46 (2022).
IXESHA LOKUQALA: UJan-04 ukuya ku-024