Chii chaitika
Vatsvagiri vanotsanangura LUCAID, an agentic multimodal AI system yechokwadi chemapapu-cancer pathology. Iyo sisitimu inosanganisa mapfumbamwe mamodule anovhara maitiro ekushanda, kubva pamhando yekudzora uye kuona bundu kune biomarker scoring uye yakarongeka rondedzero yechizvarwa. Iro bepa rinoshuma kushanda kwakasimba kunopesana neanonongedzerwa nenyanzvi uye chiyero chepamusoro checoncordance kupfuura vashanu vane ruzivo rwethoracic pathologists mune zvinotarisirwa kusimbiswa.
Kwazvakabva ndeye arXiv preprint yakaunzwa muna Aug. 24, 2026. Inopa LUCAID seagentic AI system yakagadzirirwa kunyatsoita chirwere chegomarara remapapu, kwete sechiratidziro-chinangwa chemufananidzo chinoshandiswa kune rimwe basa rakatetepa. Iro bepa rinoti diagnostics yekenza yemapapu inoda kubatanidzwa kwehistomorphological, immunohistochemical uye mamorekuru maficha, nepo ikozvino AI maturusi anowanzo gadzirisa zvikamu zvakasarudzwa zvemaitiro iwayo. LUCAID inoitirwa kubatanidza izvo zvikamu kuburikidza neanobatanidza mumiriri anogona kufunga pane zvese zvinobuda module uye kutsigira inopindirana kubvunza.
Iro bepa rinoratidza mapfumbamwe mamodule anovhara maitiro ekufamba kwekushanda: kutonga kwemhando, kuona bundu uye kupatsanurwa, histological subtyping, tumor-microenvironment profiling, tumor-cellularity quantification, kufungidzira biomarker scoring yePD-L1, MET uye TROP-2, uye otomatiki yakarongeka rondedzero yechizvarwa. Iwo abstract mapoka mamwe anoshanda pamwechete, saka haaperengere zita rakasiyana kune yega module, asi inojekesa kuti sisitimu inotambanudzira ongororo yemifananidzo, ongororo yehuwandu, kududzira biomarker uye kushuma. Vashandisi vanogona kubvunza zvabuda uye kuburitsa mishumo inoenderana nemhedzisiro, maererano nevanyori.
Zvichipesana nezvinodaidzwa nesosi yakakura-yepasi-yechokwadi zvirevo, mamodule ekuongorora akawana F1 mamakisi kubva pa0.82 kusvika 0.95. F1 chiyero chinosanganisa kurongeka uye kuyeuka, asi chirevo hachipi basa-ne-basa zvibodzwa, nguva dzekuvimba, nhamba dzekesi kana maumbirwo ezvinyorwa. Izvo zvakasiiwa zvine basa nekuti mutsara umwechete unogona kuvanza misiyano yakakura pakati pekuonekwa kwebundu, kupatsanurwa, subtyping, microenvironment kuongororwa uye biomarker scoring. Tsanangudzo yakapihwa inosiya huwandu hwakaunganidzwa pamamodule aya.
Iyo bepa zvakare inoshuma inotarisirwa kukiriniki kusimbiswa. LUCAID yakasvika 93.0% concordance nenyanzvi-panel yakatongerwa referensi chiyero pane zvese zvinogona kuitwa nekiriniki sarudzo. Iyo abstract inoenzanisa mhedzisiro yacho neconcordance rates ye68.3% kusvika 81.1% yevashanu vane ruzivo rwethoracic pathologists. Iko kunobva hakuzivise huwandu hwezviitiko, nzvimbo dzekiriniki, izvo chaizvo sarudzo dzakaongororwa, kusawirirana kwakagadziriswa sei, kana kuti kuenzanisa kwakaitwa pasi pemamiriro akafanana. Iwo manhamba saka zvichemo zvakaitwa nevanyori vepreprint uye hazvifanirwe kubatwa seyakazvimiririra yakasimbiswa kuita kwekiriniki.
Nei zvichikosha
Sarudzo dzePathology dzekunyatso oncology zvinoenderana nekusanganisa zvinoonekwa zvinyama zvinowanikwa, immunohistochemical mhedzisiro uye mamorekuru maficha. Kana zvakashumwa zvikaramba zviripo pasi pekuongororwa kwakazvimiririra, hurongwa hunobatanidza mabasa aya hunogona kutsigira kuenderana kwekuita sarudzo dzekiriniki uye kuderedza kukosha kwekufamba pakati pezvishandiso zvakabviswa. Humbowo hunoramba huri mushumo wekutsvagisa, kwete humbowo hwekubvumidzwa kwemutemo kana kujaira kuendeswa kwekiriniki.
Kukosha kunoshanda ndiko hupamhi hwegadziriro yakarongwa. Pathology workflows inowanzoda mafomu akati wandei ehumbowo kuti adudzirwe pamwechete, uye bviro inopokana kuti ongororo yekuona ihafu-yehuwandu uye iri pasi pekusiyana kweinterobserver. Sisitimu inobatanidza tishu morphology, cellularity, tumor microenvironment uye inofanotaura biomarkers inogona kupa yakajairwa analytical layer yemakesi ayo kana achida kuongororwa kwakasiyana. Icho chirevo chakanyanya kukosha pane kuvandudza chibodzwa chega chega chemufananidzo nekuti chinangwa chinobuda idudziro yakarongeka.
Iyo yakashumwa 93.0% concordance inocherechedzwa nekuti yakasungirirwa kune inotarisirwa kusimbiswa kwekiriniki uye nekiriniki sarudzo dzinoitwa, pane kungodzokororwa kuongororwa. Kufananidzwa nevashanu vane ruzivo rwethoracic pathologists zvakare inoratidza kuti vanyori vari kuongorora sisitimu maererano nehunyanzvi hwekuita, kwete kungopesana neiyo otomatiki bhenji. Nekudaro, iyo sosi hairatidze kuti LUCAID inovandudza mhedzisiro yevarwere, inoshandura sarudzo dzekurapa zvakachengeteka, inopfupisa nguva yekutendeuka kana kuderedza mutengo. Concordance ine referensi chiyero haina kufanana nekubatsirwa kwekiriniki.
Iyo system's agent dhizaini inogona kuita kuti zvazvinobuda zvishandiswe zvakanyanya nekubvumira varapi kubvunza mhinduro dzemodule uye kugamuchira mushumo wakarongeka unoisa zvakawanikwa mumamiriro ezvinhu. Inogonawo kugadzira zvitsva zvekutarisa zvinodiwa. Kana mumiririri mumwe chete achironga ongororo dzakawanda, kukanganisa kukwidza kwemhando-yekudzora, kupatsanura kana danho rekuisa rinogona kukanganisa zvinozoitika uye mushumo wekupedzisira. Iyo abstract haitsanangure matanda ekuongorora, kusava nechokwadi kuratidzwa, kusazvibata, chengetedzo kubva kune isingatsigirwe mhedziso kana maitiro ekugadzirisa zvinopokana module zvinobuda.
Tsvagiridzo iyi inogona kubatsira kupfuura iyi chaiyo sisitimu nekuti inomisa yekurapa AI seyakabatanidzwa mafambiro ebasa kwete muunganidzwa wevanodimburwa classifiers. Zvakadaro humbowo huchiri kushomekerwa nekwakabva pano: an arXiv abstract ye preprint. Iko hakuna chiratidzo mune zvakapihwa zvekuongorora nevezera, mvumo yekutonga, kuwanikwa kwekutengesa kana kudzokorora kwakazvimirira. Masangano anofunga nezvehurongwa hwakadaro angada humbowo pamusoro pekuenzanisa, kuita kweboka diki, kutonga kwedata, kudyidzana uye basa revanhu vasati vabata mhedzisiro yakataurwa seyakagadzirira kutarisirwa kwevarwere.
Interactive Mechanism: Iyo Inonyatsoshanda
Ongorora ari pasi tekinoroji kuseri kwekusimudzira uku uchipindirana.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
Zvekutarisa zvinotevera
Mibvunzo mikuru ndeyekuti LUCAID inoita sei muzvipatara, scanner, huwandu hwevarwere uye nyaya dzakaoma; kangani varapi vanokunda kana kugamuchira zvazvinobuda; uye kana konikodhenzi yakashumwa inoshandura kuita sarudzo dzekurapa dziri nani. Iyo sosi haipe inotarisirwa saizi yeboka, nzvimbo dzinotora chikamu, yekunze-yekusimbisa dhizaini, mamiriro ekutonga, kuwanikwa, kufambiswa kwebasa kana kukanganisa kukanganisa.
Chekutanga chakanyanya kukoshesa ruzivo rwakazara rwekusimbisa. Vaverengi vanofanirwa kutarisa vangangove saizi yeboka, chipatara nerabhoritari marongero, scanner kana kusvibisa kusiyanisa, huwandu hwevarwere, chirwere chidiki uye tsananguro chaiyo yesarudzo inogona kuitwa nekiriniki. Kuyedzwa kwekunze kunzvimbo dzisingabatanidzwe mukusimudzira kwaizobatsira kuratidza kana mashandiro akashumwa achiwedzera kupfuura data revanyori uye mafambiro ebasa.
Chechipiri chinonyanya kukosha ndechekuita kukanganisa. Mishumo yeramangwana inofanira kuratidza apo LUCAID isingabvumirani nenyanzvi-panel referensi chiyero, ingave zvikanganiso zvakaungana muzviitiko zvisingawanzo kana zvisinganzwisisike, uye kuti mashandiro anochinja sei kana mhando yemufananidzo yakashata kana mamorekuru uye immunohistochemical humbowo hunopesana. Mibairo yetask-level, nguva dzekusavimbika, zviyero zvekuenzanisa uye mwero wekupfuura kwevanhu zvingave zvinodzidzisa pane aggregate concordance nhamba chete.
Chechitatu chinonyanya kukosha kiriniki uye maitiro ekushanda. Sosi yacho haitaure kana LUCAID iripo kuzvipatara, ingave ine mvumo yekutonga, kuti inopinda sei mumarabhoritari aripo kana kuti idanho ripi rekuongorora kwepathologist. Humbowo hwekuti sisitimu inovandudza kusarudzwa kwekurapa, kuenderana, nguva yekuchinja kana mhedzisiro yemurwere ingazodikanwa kusiyanisa inovimbisa yetsvakiridzo prototype kubva kune yakavimbika kiriniki chishandiso. Kusvika panguva iyoyo, iyo 93.0% nhamba inofanirwa kunzwisiswa seyakashumwa preprint mhedzisiro ine zvakakosha zvisingazivikanwe.