{"id":37209,"date":"2020-04-27T12:04:00","date_gmt":"2021-02-09T12:17:06","guid":{"rendered":"https:\/\/www.globallogic.com\/sk\/insights\/blogs\/umela-inteligencia-a-lekari-spajaju-sily-v-boji-proti-ochoreniam\/"},"modified":"2025-03-08T18:41:35","modified_gmt":"2025-03-08T18:41:35","slug":"umela-inteligencia-a-lekari-spajaju-sily-v-boji-proti-ochoreniam","status":"publish","type":"insightsection","link":"https:\/\/www.globallogic.com\/sk\/insights\/blogs\/umela-inteligencia-a-lekari-spajaju-sily-v-boji-proti-ochoreniam\/","title":{"rendered":"Umel\u00e1 inteligencia a lek\u00e1ri sp\u00e1jaj\u00fa sily v boji proti ochoreniam"},"content":{"rendered":"
Vyu\u017e\u00edvanie algoritmov Deep Learning-u v\u00a0diagnostike ochoren\u00ed m\u00e1 obrovsk\u00fd potenci\u00e1l. Mnoh\u00e9 \u0161t\u00fadie v\u00a0ostatn\u00fdch rokoch dok\u00e1zali, \u017ee umel\u00e1 inteligencia m\u00e1 pri odha\u013eovan\u00ed ochoren\u00ed rovnak\u00fa, ba niekedy aj vy\u0161\u0161iu \u00faspe\u0161nos\u0165 ako kvalifikovan\u00ed lek\u00e1ri.\u00a0<\/b><\/p>\n

Po\u010d\u00edta\u010dov\u00e9 programy vyu\u017e\u00edvaj\u00fa Deep Learning algoritmy na to, aby sa nau\u010dili rozozn\u00e1va\u0165 ochorenia na z\u00e1klade anal\u00fdzy tis\u00edcok zdravotn\u00fdch inform\u00e1ci\u00ed, sn\u00edmok \u010di prejavov.<\/span> Umel\u00e1 inteligencia tak na z\u00e1klade star\u0161\u00edch medic\u00ednskych \u00fadajov a\u00a0z\u00e1znamov dok\u00e1\u017ee identifikova\u0165 spolo\u010dn\u00e9 \u010drty konkr\u00e9tnych ochoren\u00ed a\u00a0ur\u010di\u0165 spr\u00e1vnu diagn\u00f3zu aj bez pomoci \u013eud\u00ed. V\u00a0tomto \u010dl\u00e1nku v\u00e1m predstav\u00edme nieko\u013eko zauj\u00edmav\u00fdch POC projektov, ktor\u00e9 u n\u00e1s v GlobalLogic vznikli.\u00a0<\/span><\/p>\n

Automatick\u00e1 detekcia infarktu z\u00a0EKG z\u00e1znamu<\/b><\/h3>\n

Slovensko patr\u00ed v\u00a0OECD medzi krajiny s\u00a0najvy\u0161\u0161ou mierou \u00famrtnosti zapr\u00ed\u010dinenej kardiovaskul\u00e1rnymi chorobami. V\u00a0pr\u00edpade infarktu myokardu (zn\u00e1meho aj ako z\u00e1stava srdca) je pre \u00faspe\u0161n\u00fa lie\u010dbu potrebn\u00e9 \u010do najviac skr\u00e1ti\u0165 \u010das, ktor\u00fd uplynie od samotn\u00e9ho infarktu po poskytnutie lek\u00e1rskej pomoci. Doslova plat\u00ed, \u017ee ka\u017ed\u00e1 min\u00fata m\u00e1 cenu zlata.\u00a0<\/span><\/p>\n

Zvy\u010dajn\u00fd scen\u00e1r prebieha nasledovne: z\u00e1chranka dost\u00e1va hl\u00e1senie o\u00a0pacientovi so subjekt\u00edvnym pocitom bolesti na hrudi, prev\u00e1\u017ea ho do najbli\u017e\u0161ej nemocnice, kde pr\u00edslu\u0161n\u00fd \u0161pecialista pacienta diagnostikuje (na z\u00e1klade EKG) a\u00a0ur\u010d\u00ed, \u017ee je nutn\u00fd ur\u00fdchlen\u00fd prevoz pacienta na \u0161pecializovan\u00e9 pracovisko, preto\u017ee dan\u00e9 zariadenie neposkytuje potrebn\u00fa starostlivos\u0165. V kone\u010dnom d\u00f4sledku v\u0161ak u\u017e m\u00f4\u017ee by\u0165 pr\u00edli\u0161 neskoro.<\/span><\/p>\n

V s\u00fa\u010dasnosti u\u017e s\u00fa na trhu k dispoz\u00edcii rie\u0161enia, ktor\u00e9 pom\u00e1haj\u00fa bojova\u0165 proti drahocenn\u00e9mu \u010dasu, ktor\u00fd mo\u017eno u\u0161etri\u0165 diagnostikou stavu pacienta na mieste alebo po\u010das prepravy. Tieto \u00fadaje EKG vyhodnocuj\u00fa odborn\u00edci (v tomto konkr\u00e9tnom pr\u00edpade kardiol\u00f3govia), ktor\u00ed sa nenach\u00e1dzaj\u00fa v ter\u00e9ne, ale dostan\u00fa \u00fadaje, ktor\u00e9 po\u0161le z\u00e1chran\u00e1r prostredn\u00edctvom mobilnej aplik\u00e1cie. Prototyp, ktor\u00fd sme vyvinuli, mal za cie\u013e roz\u0161\u00edri\u0165 u\u017e existuj\u00face rie\u0161enie a z\u00edska\u0165 v\u00fdsledky z d\u00e1t bez potreby kontaktova\u0165 in\u00fa osobu alebo \u0161pecialistu.<\/span><\/p>\n

Aplik\u00e1cia najprv zo z\u00e1znamu pomocou konvolu\u010dnej neur\u00f3novej siete tr\u00e9novanej na manu\u00e1lne ozna\u010den\u00fdch sn\u00edmkach automaticky vyextrahuje jednotliv\u00e9 zvody, a\u00a0n\u00e1sledne ich analyzuje \u010fal\u0161ia neur\u00f3nov\u00e1 sie\u0165, ktor\u00e1 analyzuje tvar z\u00e1znamu sign\u00e1lov z\u00a0elektr\u00f3d a\u00a0vyhodnot\u00ed pravdepodobnos\u0165 infarktu. Pos\u00e1dka sanitky sa tak m\u00f4\u017ee rozhodn\u00fa\u0165, \u010di pacienta previez\u0165 priamo na \u0161pecializovan\u00e9 pracovisko, kde mu bude poskytnut\u00e1 vhodn\u00e1 lie\u010dba.\u00a0<\/span><\/p>\n

Predikcia rozmerov endoprot\u00e9zy kolenn\u00e9ho k\u013abu<\/b><\/h3>\n

Predstavte si, \u017ee m\u00e1te probl\u00e9m s\u00a0kolenom a\u00a0ortop\u00e9d v\u00e1m doporu\u010d\u00ed implant\u00e1ciu umel\u00e9ho k\u013abu. Ka\u017ed\u00fd \u010dlovek m\u00e1 in\u00e9 rozmery a\u00a0tvar kolenn\u00e9ho k\u013abu. Vyvst\u00e1va teda ot\u00e1zka, ako ur\u010di\u0165 vhodn\u00fa ve\u013ekos\u0165 tej va\u0161ej n\u00e1hrady. V\u00a0s\u00fa\u010dasnosti sa r\u00f4zne \u0161tandardizovan\u00e9 ve\u013ekosti endoprot\u00e9z sk\u00fa\u0161aj\u00fa priamo po\u010das oper\u00e1cie.\u00a0<\/span><\/p>\n

Tak\u00fdto postup v\u0161ak z\u010faleka nie je optim\u00e1lny. Ide\u00e1lne by bolo pozna\u0165 presn\u00fd rozmer n\u00e1hrady e\u0161te pred samotnou oper\u00e1ciou, \u010do by umo\u017enilo efekt\u00edvnej\u0161iu pr\u00edpravu na z\u00e1krok. Pon\u00faka sa rie\u0161enie zalo\u017een\u00e9 na vyu\u017eit\u00ed CT sn\u00edmku kolena, ktor\u00fd by vytvoril trojrozmern\u00fd model pacientovho kolenn\u00e9ho k\u013abu, na z\u00e1klade ktor\u00e9ho by bolo mo\u017en\u00e9 ur\u010di\u0165 najvhodnej\u0161\u00ed \u0161tandardizovan\u00fd model endoprot\u00e9zy, alebo vyrobi\u0165 personalizovan\u00fa verziu umel\u00e9ho kolenn\u00e9ho k\u013abu pomocou 3D tla\u010de.\u00a0<\/span><\/p>\n

Po\u010d\u00edta\u010dov\u00e1 tomografia (CT) v\u0161ak nie je \u00faplne ne\u0161kodn\u00e1, preto\u017ee pacienti s\u00fa pri nej vystaven\u00ed istej nezanedbate\u013enej d\u00e1vke \u017eiarenia. Napadlo n\u00e1m preto, \u017ee by sa na ur\u010denie ve\u013ekosti kolenn\u00e9ho implant\u00e1tu mohli pou\u017ei\u0165 v\u00fdlu\u010dne r\u00f6ntgenov\u00e9 sn\u00edmky, zhotovenie ktor\u00fdch pre pacientov znamen\u00e1 cca. 100x men\u0161iu radi\u00e1ciu v\u00a0porovnan\u00ed s\u00a0CT. Podobne, ako pri predo\u0161lom projekte, aj v\u00a0tomto pr\u00edpade sme na extrakciu zobrazenia kolenn\u00e9ho k\u013abu z\u00a0RTG sn\u00edmky \u00faspe\u0161ne pou\u017eili viacero konvolu\u010dn\u00fdch neur\u00f3nov\u00fdch siet\u00ed a\u00a0n\u00e1sledne sme vytvorili jeho kompletn\u00fd 3D model. V\u00fdsledkom je mo\u017enos\u0165 predikcie ve\u013ekosti endoprot\u00e9zy kolenn\u00e9ho k\u013abu s\u00a0prijate\u013enou mierou presnosti len na z\u00e1klade RTG sn\u00edmkov.<\/span><\/p>\n

Automatick\u00e9 meranie r\u00e1n<\/b><\/h3>\n

Po\u010det pacientov s\u00a0diabetom v\u00a0ostatn\u00fdch desa\u0165ro\u010diach prudko vzr\u00e1stol. Kombin\u00e1cia zn\u00ed\u017een\u00e9ho prietoku krvi a\u00a0diabetickej neuropatie zvy\u0161uje pravdepodobnos\u0165 tvorby vredov na doln\u00fdch kon\u010datin\u00e1ch, ktor\u00e9 m\u00f4\u017eu vies\u0165 a\u017e k\u00a0amput\u00e1cii. Lie\u010dba vredov je ve\u013emi n\u00e1ro\u010dn\u00e1 a\u00a0vy\u017eaduje si ka\u017edodenn\u00fa starostlivos\u0165, pri\u010dom pre jej dobr\u00e9 nastavenie a\u00a0koordin\u00e1ciu je nutn\u00e9 pravidelne sledova\u0165 zmeny morfol\u00f3gie r\u00e1n. Pacienti tak musia \u010dasto nav\u0161tevova\u0165 lek\u00e1ra, alebo si meraj\u00fa ve\u013ekos\u0165 rany pomocou prav\u00edtka, \u010do nie je ani pr\u00edjemn\u00e9 a\u00a0nemus\u00ed by\u0165 ani presn\u00e9.\u00a0<\/span><\/p>\n

Nedalo by sa to v\u00a0dne\u0161nej digit\u00e1lnej dobe robi\u0165 lep\u0161ie? S\u00fa smartf\u00f3ny dostato\u010dne inteligentn\u00e9 na to, aby dok\u00e1zali poskytn\u00fa\u0165 potrebn\u00e9 inform\u00e1cie pomocou fotografie? Telef\u00f3ny u\u017e dnes vyu\u017e\u00edvaj\u00fa mnoho sn\u00edmac\u00edch technol\u00f3gi\u00ed, napr. fotoapar\u00e1ty \u010di sn\u00edma\u010de h\u013abky, a\u00a0dok\u00e1\u017eu vytvori\u0165 pomerne presn\u00fa priestorov\u00fa mapu skenovan\u00e9ho prostredia.\u00a0<\/span><\/p>\n

V\u00a0kombin\u00e1cii s\u00a0algoritmami strojov\u00e9ho u\u010denia tak s\u00a0ich pomocou dok\u00e1\u017eeme ka\u017edodenne z\u00edska\u0165 3D model vredu a\u00a0monitorova\u0165 postup jeho hojenia. Pacienti si m\u00f4\u017eu ranu zmera\u0165 r\u00fdchlo, presne a\u00a0bezkontaktne v\u00a0pohodl\u00ed svojho domova. Navy\u0161e maj\u00fa pr\u00edstup aj k\u00a0odborn\u00e9mu poradenstvu, preto\u017ee lek\u00e1r monitoruje \u00fadaje zad\u00e1van\u00e9 pacientmi pomocou online port\u00e1lu, v\u010faka \u010domu sa m\u00f4\u017ee sk\u00fasen\u00fd odborn\u00edk pomocou 3D modelu \u013eahko dosta\u0165 aj k\u00a0detailnej\u0161iemu rozboru rany bez toho, aby pacient musel osobne nav\u0161t\u00edvi\u0165 zdravotn\u00edcke zariadenie. V\u00a0praxi to teda znamen\u00e1, \u017ee telemedic\u00edna u\u017e nie je len hudbou bud\u00facnosti.<\/span><\/div>\n","protected":false},"excerpt":{"rendered":"

Vyu\u017e\u00edvanie algoritmov Deep Learning-u v\u00a0diagnostike ochoren\u00ed m\u00e1 obrovsk\u00fd potenci\u00e1l. Mnoh\u00e9 \u0161t\u00fadie v\u00a0ostatn\u00fdch rokoch dok\u00e1zali, \u017ee umel\u00e1 inteligencia m\u00e1 pri odha\u013eovan\u00ed ochoren\u00ed rovnak\u00fa, ba niekedy aj vy\u0161\u0161iu \u00faspe\u0161nos\u0165 ako kvalifikovan\u00ed lek\u00e1ri.\u00a0V tomto \u010dl\u00e1nku v\u00e1m predstav\u00edme nieko\u013eko zauj\u00edmav\u00fdch POC projektov, ktor\u00e9 u n\u00e1s na t\u00fato t\u00e9mu v GlobalLogic vznikli. <\/p>\n","protected":false},"author":12,"featured_media":30459,"parent":0,"menu_order":0,"template":"","insight":[41],"insight-subcats":[809,810,811,808],"insight-industry":[747],"insight-services":[],"insight-partners":[],"class_list":["post-37209","insightsection","type-insightsection","status-publish","has-post-thumbnail","hentry","insight-blogs","insight-subcats-ai-governance","insight-subcats-analytics","insight-subcats-data-engineering","insight-subcats-mlops","insight-industry-zdravotnictvo-a-life-sciences"],"acf":[],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insightsection\/37209","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insightsection"}],"about":[{"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/types\/insightsection"}],"author":[{"embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/users\/12"}],"version-history":[{"count":2,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insightsection\/37209\/revisions"}],"predecessor-version":[{"id":102454,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insightsection\/37209\/revisions\/102454"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/media\/30459"}],"wp:attachment":[{"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/media?parent=37209"}],"wp:term":[{"taxonomy":"insight","embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insight?post=37209"},{"taxonomy":"insight-subcats","embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insight-subcats?post=37209"},{"taxonomy":"insight-industry","embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insight-industry?post=37209"},{"taxonomy":"insight-services","embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insight-services?post=37209"},{"taxonomy":"insight-partners","embeddable":true,"href":"https:\/\/www.globallogic.com\/sk\/wp-json\/wp\/v2\/insight-partners?post=37209"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}