Subira ku makuru
Guhanga udushyaAI Understanding ibisobanuro

Gufungura-isoko ya OCR umuyoboro ugabanya amakuru ya kanseri yihuta

Abashakashatsi bo muri kaminuza ya Minnesota bakoze umuyoboro uhuza isoko ya OCR ikuramo amanota y’ibizamini bya genoside muri raporo z’amavuriro bifite 97%, bikaba bishoboka ko byasimburwa n’intoki zinjira buhoro.

4 min readRead the linked source
Source-provided image accompanying Open-source OCR pipeline reduces cancer registry data latency
InkomokoInkomoko yanditse
Umwanditsi
bioengineer.org
Ihuza ry'inkomoko
bioengineer.orghttps://bioengineer.org/ai-reads-the-charts-how-open-source-ocr-could-slash-cancer-data-delays/
Ubwoko bw'inkomoko
Inkomoko ihujwe - ibanze-isoko yimiterere ntabwo yashizweho.
ImirongoSobanukirwa ibi mumasegonda 60

Tangira hano

Amagambo y'ingenzi

OCR (Kumenyekanisha neza)
Tekinoroji ihindura inyandiko mumashusho cyangwa scan mumashini isomwa.
Umuyoboro
Urutonde rwakazi rwo gutunganyiriza ibintu, intambwe yicyitegererezo, hamwe nibyiciro bitunganijwe.
Ubukererwe
Igihe kiri hagati yo kohereza icyifuzo no kwakira icyitegererezo cyibisubizo.
IsuzumeModeri ya AI Yasobanuwe Ikibazo

Byagenze bite

Itsinda ry’ubushakashatsi riyobowe na kaminuza ya Minnesota ryerekanye ko umuyoboro wa Hybrid ufungura isoko ya optique yerekana imiterere (OCR) ushobora gukuramo neza amanota yisubiramo muri raporo ya kanseri y'ibere ya Oncotype DX. Ubushakashatsi bwasohotse muri Kanseri itera & Igenzura, bwakoresheje moteri ya Tesseract na EasyOCR mu gutunganya raporo 675, bigera ku gipimo cya 97% cy’amasezerano hamwe no gukuramo intoki. Sisitemu yikora yarushije cyane umuvuduko wo kwandikisha intoki gakondo, akenshi ihura nubukererwe bwamezi 12 kugeza 18.

Itsinda ry’ubushakashatsi, barimo Qianyun Luo na Schelomo Marmor, bakoze umuyoboro wa 'Hybrid OCR' (H-OCR) kugira ngo bakemure ubukererwe mu gitabo cya kanseri. Ibisubizo by'ibizamini rusange, nk'amanota ya Oncotype DX yisubiramo, bibikwa kenshi mubitabo byubuzima bwa elegitoronike (EHRs) nkibishusho bitunganijwe neza aho kuba amakuru asomeka imashini, bisaba kwandukura intoki.

Umuyoboro wa H-OCR ukoresha EasyOCR kugirango wige-byimbitse-bishingiye ku nyandiko hamwe na Tesseract nka moteri yo gusubira inyuma kugirango imenyekane. Ubu buryo bwa Hybrid bwagenewe gukoresha ubushobozi bwa EasyOCR bwo gutondekanya amatangazo azenguruka kuri raporo mugihe ukoresha umuvuduko wa Tesseract hamwe nubushobozi bwo kumenya imiterere.

Mu kwemeza raporo 675, umuyoboro wa H-OCR wageze ku masezerano 97% hamwe n’ibipimo ngenderwaho byifashishijwe n’intoki, urenze igipimo cya 91% cy’ukuri kwakuweho. Igikorwa cyikora cyarangije umurimo mumasaha agera kuri 4.9, ugereranije nubukererwe bwamezi asanzwe yibikorwa byintoki.

Ubushakashatsi bwerekanye ko amanota y’icyizere y’umuyoboro yerekanaga neza amakosa ashobora kuba, byerekana ko ishyirwa mu bikorwa ry’ejo hazaza rishobora gukoresha aya manota kugira ngo ryerekane ibicuruzwa bitazwi kugira ngo bisuzumwe n'abantu, bityo bigabanye ingaruka zo kutavugwa nabi.

Ibisobanuro birambuye: bioengineer.org ↗

Impamvu ari ngombwa

Kugeza ubu kwishingikiriza ku ntoki ku mibare ya genomic bitera icyuho gikomeye mu bushakashatsi bwa kanseri na oncology yuzuye. Mugukoresha uburyo bwo kuvana amakuru yimiterere muri PDF zitubatswe, ubu buryo butangirwa isoko butuma sisitemu yubuzima itondekanya kanseri mu gihe gikwiye. Ihinduka ni ingenzi mu kuzamura ireme ryibimenyetso bifatika ku isi, bigafasha ubushakashatsi bwihuse bwo kugereranya, no gutanga amakuru yubatswe akenewe mu guhugura imiterere ya AI muri onkologiya. Kuberako ibikoresho bifunguye-isoko kandi birashobora gukorera mubidukikije byubahiriza HIPAA, ubu buryo butanga igisubizo gito, cyororoka cyibisubizo byinzego zidafite amikoro yo kuvugurura ibikorwa remezo byamakuru bidafite software yihariye.

Ibinyabuzima bya biomark ni ngombwa kugirango oncologiya iboneye kandi ipime ubuziranenge, nyamara akamaro kayo muri iki gihe kugarukira ku gihe bifata kugira ngo biboneke mu bubiko bw'ubushakashatsi. Kugabanya ubukererwe bituma habaho ibimenyetso byinshi mugihe.

Gukoresha ibikoresho bifungura isoko nka Tesseract na EasyOCR byemeza ko igisubizo kiboneka ku bigo bito bya kanseri bito cyangwa biterwa n'umutungo bishobora kubura ingengo yimishinga ya software ikuramo amakuru ahenze kandi yihariye.

Ubushakashatsi bwerekanye ko inyungu z’ifatwa ryihuse zikoreshwa cyane mu baturage batandukanye b’abarwayi, kuko abashakashatsi basanze imibare y’abarwayi n’ibintu by’amavuriro bitavuze neza amakosa yo gukuramo.

Muguhindura ibintu bibiri byubatswe mubintu byubatswe, uyu muyoboro utanga amakuru yo murwego rwohejuru, maremare maremare akenewe mumahugurwa no gutunganya ibizaza bya AI hamwe no kwiga imashini mumwanya wa oncology.

Interactive Mechanism

Uburyo bukoreshwa: Uburyo bukora

Shakisha ikoranabuhanga ryihishe inyuma yiri terambere.

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
Kugenzura Ibitekerezo Byagenzuwe+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

Ibyo kureba

Ubushakashatsi bw'ejo hazaza buzakenera gukemura ubunini bw'uyu muyoboro urenze imiterere ya Oncotype DX isanzwe. Abanditsi bavuze ko raporo zikurikiraho zizakurikiraho zerekana raporo zerekana ubudasa butandukanye ku bacuruzi, zitanga ikibazo gikomeye cyo kuvoma mu buryo bwikora. Byongeye kandi, mugihe ubushakashatsi bwemeje umuyoboro urwanya dataset isubira inyuma, ibizamini byateganijwe mubikorwa byubuvuzi byanditse birasabwa kwemeza ko byizewe mubyukuri, kwisi-nini cyane.

Abashakashatsi bagaragaje mu buryo bweruye ko ari ngombwa kugerageza umuyoboro ku bwoko butandukanye bw’inyandiko, nka raporo zikurikiraho zizakurikiraho, zikaba zitandukanye cyane n'abacuruzi n'imiterere.

Ubushakashatsi bwakorewe muri sisitemu imwe yubuzima; imirimo izaza igomba gusuzuma imikorere yumuyoboro mubigo byinshi kugirango irebe ko ikomeza gukomera muburyo butandukanye bwo gusikana, gukemura fax, hamwe nuburyo butandukanye bwa EHR.

Ibyifuzo byo kwinjiza mubikorwa bya rejisitiri bizima nintambwe ikurikira yo kumenya niba sisitemu igumana ukuri kwayo no gukora neza mugihe itunganya amakuru yubuvuzi yinjira mugihe nyacyo.

Ibijyanye nuyobora & ibibazo

Moderi ya AI YasobanuweAmahugurwa ya AIEjo hazaza ha AIGerageza ibyo uzi - gerageza ikibazo cya AI kubuntuReba ijambo AI mumagambo yacuKurikiza icyerekezo cya AI cyo kurekura
Basanze ari ingirakamaro?