Ndatenda nekushanyira Nature.com.Shanduro yebrowser yauri kushandisa ine tsigiro yeCSS shoma.Kuti uwane mibairo yakanaka, tinokurudzira kushandisa imwe vhezheni itsva yebrowser yako (kana kudzima kuenderana modhi muInternet Explorer).Zvichakadaro, kuti tive nechokwadi cherutsigiro runoenderera, tiri kuratidza saiti isina kutaera kana JavaScript.
Mazino anoonekwa sechiratidzo chakanyatsojeka chezera remuviri wemunhu uye anowanzoshandiswa mukuongorora zera rekuongorora.Isu takavavarira kusimbisa fungidziro yezera rekucherwa kwemazino nekuenzanisa kurongeka kwefungidziro uye kupatsanurwa kwechikamu chemakore gumi nemasere nemaitiro echinyakare uye fungidziro yezera remigodhi.Huwandu hwe2657 panoramic radiographs dzakaunganidzwa kubva kuvagari vekuKorea neJapan vane makore gumi nemashanu kusvika makumi maviri nematatu.Ivo vakakamurwa kuita seti yekudzidziswa, imwe neimwe iine mazana mapfumbamwe eKorea radiographs, uye yemukati bvunzo seti ine 857 Japanese radiographs.Isu takafananidza kurongeka uye kugona kwemaitiro echinyakare nemaseti ekuedzwa emamodhi ekuchera data.Iko kurongeka kwemaitiro echinyakare pane yemukati bvunzo seti yakakwira zvishoma pane iyo yedata mining modhi, uye mutsauko idiki (kureva mhosho yakakwana <0.21 makore, midzi inoreva square error <0.24 years).Kuitwa kwemhando ye18-year cutoff kwakafananawo pakati pemaitiro echinyakare nemhando dzemigodhi yedata.Nekudaro, nzira dzechinyakare dzinogona kutsiviwa nemhando dzemigodhi yedata paunenge uchiita ongororo yezera uchishandisa kukura kwechipiri uye chechitatu molars muKorea pwere nevakuru vechidiki.
Dental age estimation inoshandiswa zvakanyanya muforensic medicine uye vana dentistry.Kunyanya, nekuda kwehukama hwepamusoro pakati pezera renguva nekukura kwemazino, kuongororwa kwezera nematanho ekukura kwemazino chinhu chakakosha chekuongorora zera revana nevechiri kuyaruka1,2,3.Nekudaro, kune vechidiki, kufungidzira zera remazino kunoenderana nekukura kwemazino kune zvarisingakwanisi nekuti kukura kwemazino kunenge kwapera, kunze kwechitatu molars.Chinangwa chepamutemo chekusarudza zera revechidiki nevechiri kuyaruka ndechekupa fungidziro dzakarurama uye humbowo hwesainzi hwekuti vasvika zera reruzhinji.Mune medico-mutemo tsika yevechiri kuyaruka uye vechidiki vechidiki muKorea, zera raifungidzirwa kushandisa nzira yaLee, uye chikumbaridzo chepamutemo chemakore 18 chakafanotaurwa maererano nemashoko akataurwa naOh et al 5.
Kudzidza kwemichina imhando yehungwaru hwekugadzira (AI) iyo inodzokorora kudzidza uye kurongedza huwandu hwe data, inogadzirisa matambudziko pachayo, uye inotyaira hurongwa hwedata.Kudzidza kwemichina kunogona kuwana anobatsira akavanzwa mapatani mumavhoriyamu makuru e data6.Kusiyana neizvi, nzira dzechinyakare, dzinoshanda zvakanyanya uye dzinotora nguva, dzinogona kunge dziine painogumira pakubata nemavhoriyamu makuru e data yakaoma iyo yakaoma kugadzirisa nemaoko7.Naizvozvo, zvidzidzo zvakawanda zvakaitwa nguva pfupi yadarika pachishandiswa tekinoroji yekombuta yazvino kudzikisa kukanganisa kwevanhu uye kunyatsogadzira multidimensional data8,9,10,11,12.Kunyanya, kudzidza kwakadzama kwakashandiswa zvakanyanya mukuongorora mifananidzo yekurapa, uye nzira dzakasiyana-siyana dzekufungidzira zera kuburikidza nekuongorora otomatiki maredhiyographs akashumwa kuti avandudze huchokwadi uye nehunyanzvi hwekufungidzira zera13,14,15,16,17,18,19,20 .Semuyenzaniso, Halabi et al 13 vakagadzira muchina wekudzidza algorithm yakavakirwa paconvolutional neural network (CNN) kufungidzira zera reskeletal vachishandisa radiographs emaoko evana.Ichi chidzidzo chinopa modhi inoshandisa muchina kudzidza kumifananidzo yezvekurapa uye inoratidza kuti nzira idzi dzinogona kuvandudza kurongeka kwekuongorora.Li et al14 vanofungidzirwa zera kubva papilvic X-ray mifananidzo vachishandisa yakadzika yekudzidza CNN uye vakaienzanisa neregression mhedzisiro vachishandisa ossification nhanho fungidziro.Vakaona kuti kudzidza kwakadzama CNN modhi yairatidza fungidziro yezera rimwe chete seyechinyakare regression modhi.Ongororo yaGuo et al.
Zvidzidzo zvizhinji pafungidziro yezera uchishandisa muchina kudzidza shandisa nzira dzekudzidza dzakadzama13,14,15,16,17,18,19,20.Kufungidzira zera kwakavakirwa pakudzidza kwakadzama kunonzi kwakarurama kupfuura nzira dzechinyakare.Nekudaro, nzira iyi inopa mukana mudiki wekupa hwaro hwesainzi hwekufungidzira kwezera, senge zviratidzo zvezera zvinoshandiswa mukufungidzira.Panewo makakatanwa ari pamutemo panyaya yekuti ndiani anoita ongororo iyi.Naizvozvo, fungidziro yezera yakavakirwa pakudzidza kwakadzama kwakaoma kugamuchirwa nevakuru vekutonga nevatongi.Data mining (DM) inyanzvi inogona kuwana kwete chete inotarisirwa asiwo ruzivo rusingatarisirwi senzira yekuwana hukama hunobatsira pakati pehuwandu hwe data6,21,22.Kudzidza kwemichina kunowanzo shandiswa mukuchera data, uye zvese zviri zviviri kuchera data uye kudzidza muchina zvinoshandisa makiyi algorithms kuwana mapatani mune data.Kufungidzira zera uchishandisa mazino kukura kunobva pakuongorora kwemuongorori kwekukura kwemazino anonangwa, uye kuongorora uku kunoratidzwa sedanho rezino rega rega.DM inogona kushandiswa kuongorora kuwirirana pakati pedanho rekuongorora mazino uye zera chairo uye ine mukana wekutsiva echinyakare ongororo yenhamba.Naizvozvo, kana tikashandisa matekiniki eDM kune fungidziro yezera, tinogona kushandisa muchina kudzidza mufungidziro yezera tisinganetseki nezvemhosva yepamutemo.Zvidzidzo zvinoverengeka zvekufananidza zvakaburitswa pane dzingangove dzimwe nzira dzechinyakare dzemanyorero nzira dzinoshandiswa mune forensic tsika uye EBM-yakavakirwa nzira dzekutarisa zera remazino.Shen et al23 vakaratidza kuti DM modhi yakanyatsojeka kupfuura yechinyakare Camerer formula.Galibourg et al24 vakashandisa nzira dzakasiyana dzeDM kufanotaura zera maererano neDemirdjian criterion25 uye zvigumisiro zvakaratidza kuti nzira yeDM yakakunda nzira dzeDemirdjian neWillems mukufungidzira zera rehuwandu hweFrance.
Kufungidzira zera remazino evachiri kuyaruka vekuKorea uye vechidiki, nzira yaLee 4 inoshandiswa zvakanyanya muKorea forensic tsika.Iyi nzira inoshandisa ongororo yenhamba dzechinyakare (sekudzokorodza kakawanda) kuongorora hukama huripo pakati pezvidzidzo zvekuKorea nezera rekurongeka.Muchidzidzo ichi, nzira dzekufungidzira zera dzakawanikwa pachishandiswa nzira dzechinyakare dzenhamba dzinotsanangurwa se "nzira dzechinyakare."Nzira yaLee inzira yechinyakare, uye kururama kwayo kwakasimbiswa naOh et al.5;zvisinei, kushanda kwekufungidzirwa kwezera kunoenderana neDM modhi muKorea forensic tsika kuchiri kukahadzika.Chinangwa chedu chaive chekusimbisa nesainzi kugona kubatsira kwekufungidzira kwezera kunoenderana neDM modhi.Chinangwa chechidzidzo ichi chaive (1) kuenzanisa kurongeka kwemhando mbiri dzeDM mukufungidzira zera remazino uye (2) kuenzanisa kurongwa kwemaitiro e7 DM modhi pazera re 18 makore neaya akawanikwa uchishandisa echinyakare manhamba nzira Kukura kwechipiri. uye yechitatu molars mushaya mbiri.
Zvinoreva uye kutsauka kwakajairwa kwezera rezera nedanho uye rudzi rwezino zvinoratidzwa online muSupplementary Tafura S1 (yekudzidzira seti), Supplementary Tafura S2 (yemukati bvunzo seti), uye Supplementary Tafura S3 (yekunze bvunzo seti).Iwo kappa kukosha kwe intra- uye interobserver kuvimbika kwakawanikwa kubva mukudzidziswa seti yaive 0.951 uye 0.947, zvichiteerana.P kukosha uye 95% nguva yekuvimba yekappa kukosha inoratidzwa mune yepamhepo yekuwedzera tafura S4.Iko kukosha kwekappa kwakadudzirwa se "yakakwana", inowirirana nemaitiro eLandis neKoch26.
Kana tichienzanisa zvinoreva kukanganisa zvachose (MAE), nzira yechinyakare zvishoma kudarika iyo DM modhi kune vese varume uye mune yekunze murume test set, kunze kweiyo multilayer perceptron (MLP).Musiyano pakati pechinyakare modhi neDM modhi pane yemukati MAE test set yaive 0.12-0.19 makore evarume uye 0.17-0.21 makore evakadzi.Kune bhatiri rekuyedza rekunze, misiyano idiki (0.001–0.05 makore kuvarume uye 0.05–0.09 makore kuvakadzi).Pamusoro pezvo, mudzi unoreva square mhosho (RMSE) wakadzikira zvishoma pane yechinyakare nzira, ine misiyano midiki (0.17-0.24, 0.2-0.24 yemurume yemukati test set, uye 0.03-0.07, 0.04-0.08 yekunze bvunzo set).)MLP inoratidza kuita kuri nani zvishoma pane Single Layer Perceptron (SLP), kunze kwekunge iri yemukadzi wekunze test set.Kune MAE neRMSE, bvunzo dzekunze dzakaisa zvibodzwa zvakakwirira kupfuura zvemukati bvunzo dzakasetwa kune ese varume nemhando.Yese MAE neRMSE inoratidzwa muTafura 1 uye Mufananidzo 1.
MAE uye RMSE yechinyakare uye data mugodhi regression modhi.Kureva kukanganisa kwakakwana MAE, mudzi zvinoreva square error RMSE, single layer perceptron SLP, multilayer perceptron MLP, yetsika CM nzira.
Classification performance (ine cutoff yemakore 18) echinyakare uye DM modhi yakaratidzwa maererano nekunzwa, chaiyo, yakanaka yekufungidzira kukosha (PPV), yakaipa yekufungidzira kukosha (NPV), uye nzvimbo iri pasi peanogamuchira anoshanda hunhu curve (AUROC) 27 (Tafura 2, Mufananidzo 2 uye Supplementary Figure 1 paIndaneti).Panyaya yekunzwisiswa kwebhatiri rekuyedza remukati, nzira dzechinyakare dzakaita zvakanaka pakati pevarume uye zvakanyanya pakati pevakadzi.Nekudaro, mutsauko mukuita kwechikamu pakati pemaitiro echinyakare uye SD ndeye 9.7% yevarume (MLP) uye chete 2.4% yevakadzi (XGBoost).Pakati pemhando dzeDM, logistic regression (LR) yakaratidza kunzwa kuri nani muvanhurume vese.Nezvekujeka kweyemukati bvunzo seti, zvakaonekwa kuti mana eSD modhi aita zvakanaka muvarume, nepo yechinyakare modhi yaiita zvirinani muvakadzi.Misiyano muchikamu chekuita kwevarume nevakadzi ndeye 13.3% (MLP) uye 13.1% (MLP), zvichiteerana, zvichiratidza kuti mutsauko mukuita kwechikamu pakati pemhando unopfuura kunzwa.Pakati pemhando dzeDM, muchina wekutsigira vector (SVM), muti wesarudzo (DT), uye musango (RF) mhando dzakaita zvakanakisa pakati pevarume, nepo LR modhi yaiita zvakanyanya pakati pevakadzi.Iyo AUROC yechinyakare modhi uye ese maSD modhi yaive yakakura kupfuura 0.925 (k-yepedyo muvakidzani (KNN) muvarume), ichiratidza yakanakisa kupatsanura kuita mukusarura vane makore gumi nemasere samples28.Kune yekunze bvunzo seti, pakanga paine kudzikira kwechikamu chekuita maererano nekunzwa, chaiyo uye AUROC zvichienzaniswa neyemukati bvunzo seti.Uyezve, mutsauko wekunzwa uye kujeka pakati pekuita kwechikamu chemhando dzakanakisa uye dzakaipisisa kubva pa10% kusvika ku25% uye yaive yakakura kupfuura musiyano muyedzo yemukati seti.
Sensitivity uye kujeka kweiyo data migodhi yemhando yemhando zvichienzaniswa nemaitiro echinyakare ane cutoff yemakore gumi nemasere.KNN k muvakidzani wepedyo, SVM inotsigira vector muchina, LR logistic regression, DT sarudzo muti, RF random sango, XGB XGBoost, MLP multilayer perceptron, yechinyakare CM nzira.
Nhanho yekutanga muchidzidzo ichi yaive yekuenzanisa kurongeka kwekufungidzira kwezera remazino kunowanikwa kubva kune manomwe eDM modhi neaya akawanikwa pachishandiswa tsika dzechinyakare.MAE uye RMSE zvakaongororwa mukati mekuedza seti yevarume vose, uye musiyano pakati pemaitiro echinyakare uye DM modhi yakabva ku44 kusvika ku77 mazuva eMAE uye kubva ku62 kusvika ku88 mazuva eRMSE.Kunyange zvazvo nzira yechinyakare yakanga yakarurama zvishoma muchidzidzo ichi, zvakaoma kugumisa kuti mutsauko mudiki wakadaro une kukosha kwekiriniki here kana kuti kunoshanda.Mhedzisiro iyi inoratidza kuti huchokwadi hwekufungidzira zera remazino uchishandisa iyo DM modhi yakada kufanana neyechinyakare nzira.Kuenzanisa kwakananga nemigumisiro kubva kuzvidzidzo zvekare kwakaoma nokuti hapana tsvakurudzo yakaenzanisa huchokwadi hwemhando dzeDM nenzira dzechinyakare dzenhamba dzichishandisa nzira imwechete yekurekodha mazino muzera rimwechete sechidzidzo ichi.Galibourg et al24 akaenzanisa MAE neRMSE pakati penzira mbiri dzechinyakare (Demirjian method25 uye Willems method29) uye gumi DM modhi muvanhu vekuFrance vane makore maviri kusvika makumi maviri nemana.Vakashuma kuti mhando dzose dzeDM dzaive dzakarurama kupfuura nzira dzechinyakare, nekusiyana kwe0.20 uye 0.38 makore muMAE uye 0.25 uye 0.47 makore muRMSE zvichienzaniswa neWillems neDemirdjian nzira, zvichiteerana.Kusiyana pakati peiyo SD modhi uye dzechinyakare nzira dzakaratidzwa muchidzidzo cheHalibourg dzinofunga dzakawanda mishumo30,31,32,33 yekuti nzira yeDemirdjian haifungidzire zera remazino muhuwandu kunze kwevaFrench vekuCanada kwakavakirwa chidzidzo.muchidzidzo ichi.Tai et al 34 vakashandisa iyo MLP algorithm kufanotaura zera rezino kubva ku1636 Chinese orthodontic mafoto uye vakaenzanisa kurongeka kwayo nemhedzisiro yeDemirjian uye Willems nzira.Vakataura kuti MLP ine chokwadi chepamusoro kupfuura nzira dzechinyakare.Musiyano uripo pakati penzira yeDemirdjian neyechinyakare ndeye <0.32 makore, uye nzira yeWillems makore 0.28, inofanana nemhedzisiro yechidzidzo chazvino.Migumisiro yezvidzidzo izvi zvakapfuura24,34 zvakare inopindirana nemigumisiro yechidzidzo chemazuva ano, uye nguva yekufungidzira yezera remuenzaniso weDM uye nzira yekare yakafanana.Zvisinei, zvichibva pamigumisiro yakaratidzwa, tinogona kungogumisa nekuchenjerera kuti kushandiswa kweDM modhi yekufungidzira zera kunogona kutsiva nzira dziripo nekuda kwekushaikwa kwekuenzanisa uye kunongedzera zvidzidzo zvekare.Zvidzidzo zvekutevera pachishandiswa zvienzaniso zvakakura zvinodikanwa kusimbisa zvawanikwa muchidzidzo ichi.
Pakati pezvidzidzo zvinoedza huchokwadi hweSD mukufungidzira zera remazino, zvimwe zvakaratidza huchokwadi hwepamusoro kupfuura kudzidza kwedu.Stepanovsky et al 35 vakashandisa 22 SD modhi kune panoramic radiographs ye976 Czech vagari vane makore 2.7 kusvika 20.5 makore uye vakaedza chokwadi chemuenzaniso wega wega.Vakaongorora kukurudzirwa kwemazino e16 epamusoro uye ezasi kuruboshwe asingagumi vachishandisa maitiro ekugadzirisa akarongwa naMoorrees et al 36.Iyo MAE inotangira kubva ku0.64 kusvika ku0.94 makore uye RMSE inotangira kubva ku0.85 kusvika 1.27 makore, ayo ari akarurama kupfuura maviri eDM modhi akashandiswa muchidzidzo ichi.Shen et al23 vakashandisa nzira yeCameriere kufungidzira zera remazino emazino manomwe echigarire kuruboshwe muvagari vekuChina vane makore 5 kusvika 13 makore uye vakaienzanisa nemazera anofungidzirwa kushandisa mutsara regression, SVM neRF.Vakaratidza kuti ese matatu maDM modhi ane hunyanzvi hwepamusoro zvichienzaniswa neyechinyakare Cameriere formula.Iwo MAE neRMSE muchidzidzo cheShen aive akaderera pane ayo ari muDM modhi muchidzidzo ichi.Kuwedzera kwakanyatsojeka kwezvidzidzo naStepanovsky et al.35 uye Shen et al.23 inogona kunge yakakonzerwa nekubatanidzwa kwevadiki zvidzidzo mumasampuli avo ekudzidza.Nekuda kwekuti fungidziro yezera yevatori vechikamu vane mazino ari kukura inova yakanyatsojeka sezvo huwandu hwemazino hunowedzera panguva yekukura kwemazino, huchokwadi hwezvinokonzeresa nzira yekufungidzira zera inogona kukanganisika kana vatori vechikamu muzvidzidzo vachiri vadiki.Pamusoro pezvo, kukanganisa kweMLP mufungidziro yezera idiki zvishoma pane SLP, zvichireva kuti MLP yakarurama kupfuura SLP.MLP inoonekwa zvirinani zvishoma pafungidziro yezera, pamwe nekuda kweakavanzika akaturikidzana muMLP38.Nekudaro, pane kusarudzika kune yekunze sampuli yevakadzi (SLP 1.45, MLP 1.49).Kuwanikwa kwekuti MLP ndeyechokwadi kupfuura iyo SLP mukuongorora zera inoda zvimwe zvidzidzo zvekare.
Kuitwa kwemhando yeDM modhi uye nzira yechinyakare pagumi nemasere-makore yakaenzaniswa zvakare.Ese akaedzwa maSD modhi uye echinyakare nzira pane yemukati bvunzo seti yakaratidza zvinogamuchirika mazinga erusarura kune ane makore gumi nemasere emuenzaniso.Sensitivity yevarume nevakadzi yaive yakakura kupfuura 87.7% uye 94.9%, zvichiteerana, uye chaiyo yaive yakakura kupfuura 89.3% uye 84.7%.Iyo AUROC yemamodhi ese akaedzwa zvakare inodarika 0.925.Sekuziva kwedu, hapana ongororo yakaedza mashandiro eDM modhi ye18-year classification zvichienderana nekukura kwemazino.Tinogona kuenzanisa mhedzisiro yechidzidzo ichi nekuita kwechikamu chemhando dzekudzidza kwakadzama papanoramic radiographs.Guo et al.15 vakaverenga mashandiro echikamu cheCNN-based deep learning model uye manyorerwo anoitwa nzira yaDemirjian yeimwe zera.Kunzwa uye kujeka kweiyo nzira yemanyorero yaive 87.7% uye 95.5%, zvichiteerana, uye kunzwisiswa uye kujeka kweiyo CNN modhi yakapfuura 89.2% uye 86.6%, zvichiteerana.Vakagumisa kuti modhi yekudzidza yakadzama inogona kutsiva kana kupfuura ongororo yemanyorero mukuronga zvikumbaridzo zvezera.Migumisiro yechidzidzo ichi yakaratidza maitiro akafanana ekugadzirisa;Zvinotendwa kuti kupatsanura uchishandisa DM modhi kunogona kutsiva echinyakare nzira dzehuwandu hwekufungidzira zera.Pakati pemamodheru, DM LR yaive yakanakisa modhi maererano nekunzwa kwechirume sampuli uye kunzwisiswa uye chaiyo kune yevakadzi sampuli.LR iri pachinzvimbo chechipiri mukunangana nevarume.Uyezve, LR inoonekwa seimwe yeanonyanya kushandisa-mushandisi DM35 modhi uye haina kuoma uye yakaoma kugadzirisa.Zvichienderana nemhedzisiro iyi, LR yaionekwa seyakanakisa cutoff classification modhi yevane makore gumi nemasere muhuwandu hweKorea.
Pakazere, huchokwadi hwekufungidzira zera kana kuita kwechikamu pabvunzo dzekunze kwaive kwakashata kana kuderera kana zvichienzaniswa nezvabuda paseti yebvunzo yemukati.Mimwe mishumo inoratidza kuti kurongeka kana kushanda zvakanaka kunodzikira kana fungidziro yezera yakavakirwa pahuwandu hweKorea ichishandiswa kuhuwandu hweJapan5,39, uye patani yakafanana yakawanikwa muchidzidzo chazvino.Uku kuderera kwemaitiro kwakaonekwa zvakare muDM modhi.Nokudaro, kunyatsofungidzira zera, kunyange kana uchishandisa DM mukugadzirisa maitiro, nzira dzakatorwa kubva kuhuwandu hwehuwandu hwevanhu, dzakadai semaitiro echinyakare, dzinofanirwa kusarudzwa5,39,40,41,42.Sezvo zvisiri kujeka kana modhi yekudzidza yakadzama inogona kuratidza mafambiro akafanana, zvidzidzo zvinoenzanisa kurongeka kwechokwadi uye kushanda nesimba uchishandisa nzira dzechinyakare, maDM modhi, uye yakadzama yekudzidza modhi pamasample mamwe chete inodiwa kuratidza kana hungwaru hwekugadzira hunogona kukunda uku kusaenzana kwemarudzi muzera rishoma.kuongorora.
Isu tinoratidza kuti nzira dzechinyakare dzinogona kutsiviwa nefungidziro yezera zvichibva pane yeDM modhi mune forensic zera rekufungidzira maitiro muKorea.Isu takawana zvakare mukana wekushandisa muchina kudzidza kune forensic zera rekuongorora.Zvisinei, pane zvipingamupinyi zvakajeka, zvakadai sehuwandu husina kukwana hwevatori vechikamu muchidzidzo ichi kuti vanyatsoona migumisiro, uye kushayikwa kwezvidzidzo zvekare kuti zvienzanise uye zvisimbise migumisiro yechidzidzo ichi.Mune ramangwana, zvidzidzo zveDM zvinofanirwa kuitwa nehuwandu hukuru hwemasamples uye huwandu hwakasiyana-siyana kuti ivandudze kushanda kwayo kana zvichienzaniswa nemaitiro echinyakare.Kusimbisa kugona kwekushandisa hungwaru hwekugadzira kufungidzira zera muhuwandu hwevanhu vakawanda, zvidzidzo zvenguva yemberi zvinodikanwa kuenzanisa kurongeka nekubudirira kweDM uye modhi yekudzidza yakadzama nemaitiro echinyakare mumasampuli mamwe chete.
Chidzidzo ichi chakashandisa 2,657 orthographic mafoto akaunganidzwa kubva kuKorea neJapan vakuru vane makore gumi nemashanu kusvika makumi maviri nematatu.MaKorea radiographs akakamurwa kuita mazana mapfumbamwe ekudzidziswa seti (19.42 ± 2.65 makore) uye mazana mapfumbamwe emukati bvunzo seti (19.52 ± 2.59 makore).Iyo yekudzidziswa seti yakaunganidzwa pane imwe institution (Seoul St. Mary's Hospital), uye iyo yega bvunzo yakatarwa yakaunganidzwa pamasangano maviri (Seoul National University Dental Hospital uye Yonsei University Dental Hospital).Takaunganidzawo mazana masere nemakumi mashanu nenomwe maredhiyogirafu kubva kune imwe dhata-yakavakirwa vanhu (Iwate Medical University, Japan) yekuongororwa kwekunze.Radiographs yezvidzidzo zveJapan (19.31 ± 2.60 makore) yakasarudzwa seyekunze bvunzo seti.Dhiyabhorosi yakaunganidzwa retrospectively kuti iongorore matanho ekukura kwemazino pane panoramic radiographs yakatorwa panguva yekurapa mazino.Yese data yakaunganidzwa yakanga isingazivikanwe kunze kwechikadzi, zuva rekuzvarwa uye zuva reradiograph.Kubatanidzwa uye kusabvisa maitiro aive akafanana neakambodhindwa zvidzidzo 4, 5.Nguva chaiyo yemuenzaniso yakaverengerwa nekubvisa zuva rekuzvarwa kubva pazuva rakatorwa radiograph.Boka remuenzaniso rakakamurwa kuva mapoka mapfumbamwe ezera.Zera uye kugoverwa kwepabonde kunoratidzwa muTebhu 3 Ichi chidzidzo chakaitwa maererano neDeclaration of Helsinki uye yakagamuchirwa neInstitutional Review Board (IRB) yeSeoul St. Mary's Hospital yeCatholic University of Korea (KC22WISI0328).Pamusana pekugadzirisa kwechidzidzo ichi, kubvumirwa kweruzivo hakuna kukwanisa kuwanikwa kubva kuvarwere vose vari kuongororwa neradiographic nokuda kwezvinangwa zvekurapa.Seoul Korea University St. Mary's Hospital (IRB) yakasiya zvaidiwa kuti munhu abvume.
Matanho ekusimudzira echipiri uye chechitatu molars akaongororwa maererano neDemircan maitiro25.Zino rimwe chete raisarudzwa kana rudzi rumwe chete rwezino rwakawanikwa kuruboshwe nekurudyi kweshaya imwe neimwe.Kana mazino homologous kumativi ose akanga ari akasiyana kukura nhanho, zino pamwe yakaderera kukura padanho rakasarudzwa kuti chizvidavirire kusavimbika mu inofungidzirwa zera.Mazana zana eredhiyogirafu akasarudzwa kubva pakudzidziswa akaiswa nevaviri vane ruzivo vacherechedzi kuti vaedze kuvimbika kwe-interobserver mushure mekugadzirisa kuti vaone danho rekukura kwemazino.Kuvimbika kweIntraobserver kwakaongororwa kaviri panguva dzemwedzi mitatu nemucherechedzi wekutanga.
Nhanho yepabonde uye yekukura kwechipiri uye chechitatu molars yeshaya imwe neimwe mugadziriro yekudzidzira yakafungidzirwa nemukuru wekucherechedza akadzidziswa nemhando dzakasiyana dzeDM, uye zera chairo rakagadzirirwa sechinangwa chakakosha.Mamodheru eSLP neMLP, ayo anoshandiswa zvakanyanya mukudzidza kwemuchina, akaedzwa achipesana neregression algorithms.Iyo DM modhi inosanganisa mutsara mabasa uchishandisa nhanho dzekuvandudza mazino mana uye inosanganisa data iri kufungidzira zera.SLP ndiyo yakapusa neural network uye haina akavigwa akaturikidzana.SLP inoshanda yakavakirwa pachikumbaridzo kutapurirana pakati node.Iyo SLP modhi mukudzoreredza yakafanana nemasvomhu yakafanana neyakawanda mutsara regression.Kusiyana neiyo SLP modhi, iyo MLP modhi ine akawanda akavanzika akaturikidzana ane asina mutsara activation mabasa.Maedzo edu akashandisa yakavanzika layer ine makumi maviri chete yakavanzika node ine nonlinear activation mabasa.Shandisa gradient descent senzira yekuwedzera uye MAE neRMSE sebasa rekurasikirwa kudzidzisa modhi yedu yekudzidza muchina.Iyo yakanakisa yakawanikwa yekudzoreredza modhi yakashandiswa kune yemukati uye yekunze bvunzo seti uye zera remeno raifungidzirwa.
A classification algorithm yakagadziridzwa inoshandisa kukura kwemazino mana pane yekudzidziswa seti kufanotaura kana sampuli ine makore gumi nemasere kana kuti kwete.Kuvaka modhi, takatora manomwe anomiririra muchina kudzidza algorithms6,43: (1) LR, (2) KNN, (3) SVM, (4) DT, (5) RF, (6) XGBoost, uye (7) MLP. .LR ndeimwe yeanonyanya kushandiswa classification algorithms44.Inyanzvi yekudzidza yakatariswa inoshandisa kudzokororwa kufanotaura mukana wedata rechimwe chikamu kubva pa0 kusvika ku1 uye inoisa data serimwe reboka ringangove richibva pane izvi zvingangoitika;inonyanya kushandiswa pakuronga kwebhinari.KNN ndeimwe yeakareruka muchina kudzidza algorithms45.Kana yapihwa data nyowani yekupinza, inowana k data padyo neiyo iripo seti yobva yaisa mukirasi neiyo yakanyanya frequency.Takaisa 3 yehuwandu hwevavakidzani vanotariswa (k).SVM is algorithm inokwirisa chinhambwe pakati pemakirasi maviri nekushandisa kernel basa kuwedzera nzvimbo yemutsara munzvimbo isiri-mutsara inonzi fields46.Kune iyi modhi, tinoshandisa bias = 1, simba = 1, uye gamma = 1 se hyperparameter yepolynomial kernel.DT yakashandiswa munzvimbo dzakasiyana siyana sealgorithm yekukamura data rese rakaiswa muzvikamu zvidiki nekumiririra mitemo yesarudzo muchimiro chemuti47.Iyo modhi inogadziriswa nehuwandu hushoma hwemarekodhi pane node ye2 uye inoshandisa iyo Gini index seyero yemhando.RF inzira yekubatanidza inosanganisa akawanda maDT kuti avandudze mashandiro achishandisa nzira yebootstrap yekuunganidza iyo inoburitsa isina kusimba classifier yesample yega yega nekudhirowa masampula ehukuru hwakafanana kakawanda kubva kune yekutanga dataset48.Takashandisa 100 miti, 10 miti yakadzika, 1 yakaderera node saizi, uye Gini admixture index senzira yekuparadzanisa node.Kurongeka kwedata idzva kunotemwa nevhoti yakawanda.XGBoost is algorithm inosanganisa hunyanzvi hwekusimudzira uchishandisa nzira inotora se data rekudzidzisa kukanganisa pakati peiyo chaiyo uye yakafanotaurwa kukosha yeiyo yapfuura modhi uye inowedzera chikanganiso uchishandisa gradients49.Iri algorithm inoshandiswa zvakanyanya nekuda kwekuita kwayo kwakanaka uye kugona kwezviwanikwa, pamwe nekuvimbika kwakanyanya seyakanyanya kururamisa basa.Iyo modhi ine 400 mavhiri ekutsigira.MLP inonzi neural network umo imwe kana akawanda maperceptrons anoumba akawanda layers ane one kana anopfuura akavanzika akaturikidzana pakati peyekupinda nekubuda layer38.Uchishandisa izvi, iwe unogona kuita isiri-mutsara classification apo iwe paunowedzera yekuisa layer uye wowana kukosha kwemhedzisiro, iyo yakafanotaurwa mhedzisiro kukosha inofananidzwa neiyo chaiyo mhedzisiro kukosha uye kukanganisa kunoparidzirwa kumashure.Isu takagadzira dhizaini yakavanzika ine makumi maviri akavanzika neurons mune yega yega.Imwe neimwe modhi yatakagadzira yakashandiswa kune emukati uye ekunze seti kuyedza kupatsanura maitiro nekuverenga senitivity, chaiyo, PPV, NPV, uye AUROC.Sensitivity inotsanangurwa sereshiyo yemuenzaniso unofungidzirwa kuva wemakore gumi nemasere ekuzvarwa kana kupfuura kumuenzaniso unofungidzirwa kuva wemakore gumi nemasere kana kupfuura.Kunyatsozivikanwa chikamu chemasamples ari pasi pemakore gumi nemasere ekuzvarwa uye ayo anofungidzirwa kuve ari pasi pemakore gumi nemasere.
Matanho emazino akaongororwa mugadziriro yekudzidzira akashandurwa kuva nhamba dzenhamba dzekuongorora nhamba.Multivariate linear uye logistic regression yakaitwa kugadzira mamodheru ekufungidzira ega ega pabonde uye kuwana regression mafomula anogona kushandiswa kufungidzira zera.Takashandisa mafomula aya kufungidzira zera remazino kune ese emukati neekunze ebvunzo seti.Tafura 4 inoratidza kudzokororwa uye kupatsanura mhando dzakashandiswa muchidzidzo ichi.
Kuvimbika kweIntra- uye interobserver kwakaverengerwa uchishandisa Cohen's kappa statistic.Kuti tiedze huchokwadi hweDM uye echinyakare regression modhi, takaverenga MAE neRMSE tichishandisa inofungidzirwa uye chaiyo mazera emukati nekunze bvunzo seti.Mhosho idzi dzinowanzo shandiswa kuongorora huchokwadi hwekufanotaura kwemuenzaniso.Iko kukanganisa kudiki, ndiko kunowedzera kurongeka kwekufanotaura24.Enzanisa iyo MAE neRMSE yemukati neyekunze bvunzo seti yakaverengerwa uchishandisa DM uye yechinyakare kudzoreredza.Classification performance ye18-year cutoff muzvinyorwa zvechinyakare zvakaongororwa pachishandiswa 2 × 2 contingency table.Iyo yakaverengerwa senitivity, chaiyo, PPV, NPV, uye AUROC yebvunzo seti yakaenzaniswa neyakayerwa kukosha kweiyo DM yemhando modhi.Data inoratidzwa sezvinoreva ± yakajairwa kutsauka kana nhamba (%) zvichienderana nehunhu hwe data.Maviri-mativi P maitiro <0.05 aionekwa seakakosha.Zvose zvinongororwa zvinongororwa zvakaitwa uchishandisa SAS version 9.4 (SAS Institute, Cary, NC).Iyo DM regression modhi yakashandiswa muPython pachishandiswa Keras50 2.2.4 backend uye Tensorflow51 1.8.0 zvakanangana nemasvomhu.Iyo DM classification model yakashandiswa muWaikato Knowledge Analysis Environment uye Konstanz Information Miner (KNIME) 4.6.152 yekuongorora nzvimbo.
Vanyori vanobvuma kuti data inotsigira mhedzisiro yekudzidza inogona kuwanikwa muchinyorwa uye zvekuwedzera zvinhu.Maseti akagadzirwa uye/kana akaongororwa panguva yechidzidzo anowanikwa kubva kumunyori anowirirana nechikumbiro chine musoro.
Ritz-Timme, S. et al.Zera rekuongorora: mamiriro ehunyanzvi kusangana nezvinodiwa zve forensic tsika.internationality.J. Mushonga wepamutemo.113, 129–136 (2000).
Schmeling, A., Reisinger, W., Geserik, G., naOlze, A. Chimiro chazvino chekuongorora zera rezvidzidzo zvevapenyu nechinangwa chekupomera mhosva.Forensics.mushonga.Pathology.1, 239–246 (2005).
Pan, J. nevamwe.Nzira yakagadziridzwa yekuongorora zera remazino evana vane makore mashanu kusvika kumakore gumi nematanhatu kumabvazuva kweChina.kliniki.Oral survey.25, 3463–3474 (2021).
Lee, SS etc. Chronology yekuvandudzwa kwechipiri uye chechitatu molars muKorea uye kushandiswa kwayo kwe forensic age assessment.internationality.J. Mushonga wepamutemo.124, 659–665 (2010).
Oh, S., Kumagai, A., Kim, SY naLee, SS Kururama kwefungidziro yezera uye fungidziro ye18-gore pachikumbaridzo zvichienderana nekukura kwechipiri nechitatu molars mumaKorea neJapan.PLoS ONE 17, e0271247 (2022).
Kim, JY, nevamwe.Preoperative muchina kudzidza-based data analysis inogona kufanotaura kurara kwekuvhiyiwa kurapwa muvarwere vane OSA.sainzi.Chirevo 11, 14911 (2021).
Han, M. nevamwe.Yechokwadi fungidziro yezera kubva muchina kudzidza kana pasina kupindira kwevanhu?internationality.J. Mushonga wepamutemo.136, 821–831 (2022).
Khan, S. naShaheen, M. Kubva kuData Mining kuenda kuData Mining.J. Information.sainzi.https://doi.org/10.1177/01655515211030872 (2021).
Khan, S. naShaheen, M. WisRule: The First Cognitive Algorithm for Association Rule Mining.J. Information.sainzi.https://doi.org/10.1177/01655515221108695 (2022).
Shaheen M. naAbdullah U. Karm: Kucherwa kwedhata kwechinyakare kwakavakirwa pamitemo yemubatanidzwa.kuverenga.Mat.enderera.68, 3305–3322 (2021).
Muhammad M., Rehman Z., Shaheen M., Khan M. naHabib M. Deep based based semantic kufanana kwekutsvaga uchishandisa data yemavara.zivisa.michina.control.https://doi.org/10.5755/j01.itc.49.4.27118 (2020).
Tabish, M., Tanoli, Z., uye Shahin, M. A system yekucherechedza chiitiko mumavhidhiyo emitambo.multimedia.Zvishandiso Zvishandiso https://doi.org/10.1007/s11042-021-10519-6 (2021).
Halabi, SS et al.RSNA Machine Kudzidza Dambudziko muPediatric Bone Age.Radiology 290, 498–503 (2019).
Li, Y. nevamwe.Forensic zera fungidziro kubva pelvic X-rays uchishandisa kudzidza kwakadzama.EURO.mwaranzi.29, 2322–2329 (2019).
Guo, YC, nevamwe.Kurongeka kwakaringana kwezera uchishandisa nzira dzemanyorero uye yakadzika convolutional neural network kubva ku orthographic projection mifananidzo.internationality.J. Mushonga wepamutemo.135, 1589–1597 (2021).
Alabama Dalora et al.Kufungidzira zera remapfupa uchishandisa nzira dzakasiyana dzekudzidza muchina: ongororo yakarongeka yemabhuku uye meta-analysis.PLoS YEMWE 14, e0220242 (2019).
Du, H., Li, G., Cheng, K., naYang, J. Mutengo wezera chairo revanhu vemuAfrica vekuAmerica nekuChina zvichibva pazvikamu zvekamuri rekutanga molar vachishandisa cone-beam computed tomography.internationality.J. Mushonga wepamutemo.136, 811–819 (2022).
Kim S., Lee YH, Noh YK, Park FK naOh KS Kusarudza mapoka ezera revanhu vanorarama vachishandisa mifananidzo yehungwaru-yakavakirwa pekutanga molars.sainzi.Chirevo 11, 1073 (2021).
Stern, D., Payer, C., Giuliani, N., uye Urschler, M. Automatic age estimation and most age classification from multivariate MRI data.IEEE J. Biomed.Health Alerts.23, 1392–1403 (2019).
Cheng, Q., Ge, Z., Du, H. naLi, G. Estimation yezera yakavakirwa pa3D pulp chamber segmentation yekutanga molars kubva kucone beam computed tomography nekubatanidza kudzidza kwakadzama nemaseti emazinga.internationality.J. Mushonga wepamutemo.135, 365–373 (2021).
Wu, WT, nevamwe.Kuchera data mukiriniki yakakura data: zvakajairika dhatabhesi, nhanho, uye nzira modhi.Nyika.mushonga.resource.8, 44 (2021).
Yang, J. nevamwe.Nhanganyaya kune Medical Databases uye Data Mining Technologies muBig Data Era.J. Avid.Mushonga wekutanga.13, 57–69 (2020).
Shen, S. et al.Camerer's nzira yekufungidzira zera rezino uchishandisa muchina kudzidza.BMC Oral Health 21, 641 (2021).
Galliburg A. et al.Kuenzanisa nzira dzakasiyana dzekudzidza muchina dzekufanotaura zera remazino uchishandisa Demirdjian staging nzira.internationality.J. Mushonga wepamutemo.135, 665–675 (2021).
Demirdjian, A., Goldstein, H. naTanner, JM Nzira itsva yekuongorora zera remazino.ngonono.biology.45, 211–227 (1973).
Landis, JR, uye Koch, GG Matanho evacherechedzi chibvumirano pane categorical data.Biometrics 33, 159–174 (1977).
Bhattacharjee S, Prakash D, Kim C, Kim HK naChoi HK.Textural, morphological and statistical analysis of two-dimensional magnetic resonance imaging uchishandisa artificial intelligence matekiniki ekusiyanisa mamota ekutanga ehuropi.Mashoko ehutano.resource.https://doi.org/10.4258/hir.2022.28.1.46 (2022).
Nguva yekutumira: Jan-04-2024