Izithombe zamaseli ahlukaniswe zibalulekile futhi zibiza kakhulu ocwaningweni lwezokwelapha kanye nokuxilonga. Nakuba ubuhlakani bokwenziwa (AI) manje bungenza ukuhlukaniswa kwamaseli, kudinga ukuqeqeshwa ngenani elikhulu ledatha enelebula: idatha okumele ilebula ngesandla ngochwepheshe. U-Eiram Mahera Sheikh oqokiwe njengodokotela e-Bielefeld University of Applied Sciences uphenya ukuthi le nqubo ingasheshiswa kanjani ngobuhlakani ngaphakathi kohlaka lwenethiwekhi yocwaningo lwe-SAIL. Usevele esebenzisa i-AI, ikakhulukazi ukufunda okujulile okusebenzayo , ekulungiseleleni idatha yokuqeqesha.
Izithombe ezimbili ezibizwa kukhompuyutha (isithombe) ziyizithombe zamaseli ezinto eziphilayo ezifanayo. Kwesokunxele, isithombe esincane kakhulu samanye amaseli e-pancreatic stem akwi-substrate enezimbotshana: Amachashaza amancane amhlophe, anobukhulu obuhlukahlukene, abude ngokuma, afiphalisiwe kancane kungemuva elimpunga, amaseli ngamanye. Lapha, akulula kangako ukuhlukanisa ukuthi iphikseli isengeyeseli noma isivele ingemuva. U-Eiram Mahera Sheikh udonsela ukunaka kwezivakashi kwesokudla: ukuhlelwa okufanayo kwamaseli, kodwa kulokhu anombala ohlukile futhi aklanywe ngokucacile kusukela kungemuva elimnyama manje. "Ngale maski, isithombe siwusizo kakhulu ekuhlaziyeni. Lapha, amaseli ahlukaniswe ngokucacile, esebenzisa i-AI, okungukuthi, ubuhlakani bokwenziwa." USheikh ufuna ukuqinisekisa ukuthi le nqubo ingenziwa ngokuphumelelayo nakakhulu esikhathini esizayo - ngokwenza i-AI ihlakaniphe ngokusebenzisa i-AI. Ngamanye amazwi, ufuna ukunika amandla ukuqeqeshwa okudingekayo kwe-cell segmentation AI ukuthi kuqhubeke ngokushesha kunangaphambili, futhi ngale njongo, uthembele enqubweni ye-AI.
Ngemva kweminyaka eminingana ekulo mkhakha, usosayensi wamakhompiyutha oneminyaka engu-33 ubudala onezimpande zaseNdiya okhethekile kwisayensi yedatha kanye ne-AI ngesikhathi seziqu zakhe ze-master. Manje usejoyine i-Bielefeld University (HSBI) njengomfundi wobudokotela. Ngokuhlukaniswa kwamaseli, ukhethe isihloko esibaluleke kakhulu ocwaningweni lwe-biomedical kanye nokuxilongwa: "Ukuhlaziywa kwezithombe zamaseli ahlukene kunikeza ukuqonda okubalulekile ezicini zamaseli, izinguquko ze-pathological, noma ukuthi amangqamuzana asabela kanjani ezintweni ezithile ezisebenzayo," kuchaza u-Eiram Mahera Sheikh. "Lolu ulwazi olubalulekile, isibonelo, ekwelapheni umdlavuza noma ucwaningo lwemithi." Ucwaningo lukaSheikh, nokho, luqala izinyathelo ezimbili ngaphambili-ezingeni le-meta, ngomqondo ongokomfanekiso. "Ukuhlukaniswa kwamaseli manje kungenziwa futhi ama-AI. Ngithanda ukuqeqesha le AI."
Le phrojekthi iyingxenye yenethiwekhi yocwaningo lwezinhlaka ezahlukene i-SAIL, emele i-SustAInable Life-Cicle of Intelligent Socio-Technical Systems. Ixhaswe yisifunda saseNyakatho Rhine-Westphalia ngemali efinyelela ezigidini ezingu-14,8 zama-euro, le nhlangano ihlanganisa i-Bielefeld University njengomholi wenhlangano, i-HSBI, i-TH OWL, neNyuvesi yase-Paderborn ukuze kuqinisekiswe ukuthi amasistimu e-AI asebenza ngokusobala, ngokuvikelekile, ngokuqhubekayo, nangokuqina kuwo wonke umjikelezo wabo wokuphila. Cishe ososayensi abangama-90 abathintekayo kunethiwekhi benza ucwaningo oluyisisekelo emkhakheni we-AI, umthelela wayo kubantu nasemphakathini, kanye nokusebenza okuphathekayo ezindaweni ze-Industrial 4.0 kanye nokunakekelwa kwezempilo okuhlakaniphile. UWolfram Schenck, uSolwazi Wezobunjiniyela Informatics, Umqondisi Wesayensi we-Data Analytics Cluster ye-Centre for Applied Data Science (CfADS) ekhempasini yase-Gütersloh kanye nomkhulumeli we-HSBI kuphrojekthi ye-SAIL, uqondise iziqu zobudokotela zika-Eiram Mahera Sheik, ezitholakala endaweni esibonana kuyo phakathi kocwaningo oluyisisekelo nokunakekelwa kwezempilo, kodwa futhi inikeza usizo lwezempilo.
U-Schenck uchaza izinselele zomsebenzi ka-Sheik: "Ukuze ufunde ukuhlukaniswa kwamaseli, i-AI ye-cell segmentation idinga inani elikhulu ledatha enelebula yokuqeqeshwa. Lena idatha yesithombe esezingeni lamaphikseli, enelebula futhi ihlukaniswe ngolwazi olwengeziwe." Lena inqubo eyinkimbinkimbi, ngoba ngokwemibandela ephathekayo, lokhu kusho ukuthi ngephikseli ngayinye emfanekisweni wamaseli ezinto eziphilayo, kufanele kunqunywe ukuthi ingabe ingeyesigaba "sasemuva", isigaba "se-membrane", noma isigaba "somzimba weseli", isibonelo. U-Schenck uyachaza: "Lokhu okuthiwa ukulebula okwamanje kungenziwa ngesandla kuphela ochwepheshe abazi kahle ukuthi yini ebonwayo." Ulwelwesi lwamaseli? Iseli elifile? Ingemuva? – Iphikseli ngephikseli, umuntu owazi okuthile ngendaba kufanele enze umsebenzi oyisicefe, othize oyisicefe ngaphambi kokuba i-AI iqeqeshelwe ukuhlukanisa amaseli athile.
“Ngenxa yokuthi lo msebenzi udla isikhathi esiningi futhi uyisicefe, ochwepheshe bayanqikaza ukuwenza,” kusho uDkt. Constanze Schwan. "Ngaphandle kwalokho, ukulebula kubiza kakhulu ngenxa yobude bakho kanye nokubandakanyeka kochwepheshe abaqeqeshwe kakhulu." U-Schwan ujoyine ingxoxo. Usosayensi wamakhompyutha ohlelweni lwe-Career@BI kwa-HSBI akagcini nje ngokufundisa eMnyangweni Wezobunjiniyela Nezibalo, kodwa futhi usebenza emnyangweni wocwaningo wenkampani ye-biotech i-Miltenyi Biotec e-Göttingen. Ugqamisa izidingo ezingokoqobo: "Kungaba isinyathelo esikhulu esiya phambili uma bekukhona imodeli noma i-algorithm enganciphisa umzamo odingekayo wokulebula." Izindlela ezikhona zokuhlukanisa, ezifanele ukushayela okuzenzakalelayo, isibonelo, zifinyelela umkhawulo wazo lapho usebenza nezithombe ezincane kakhulu. Lokhu kungenxa, phakathi kwezinye izinto, izakhiwo eziyinkimbinkimbi nezingcono kakhulu ezitholakala ezithombeni. Lezi ngokuvamile izitaki zezithombe emugqeni wesikhathi wezakhiwo ze-3D. Lezi zithombe noma izitaki zesithombe zivame ukuqukatha izakhiwo eziyinkimbinkimbi okunzima ukuzehlukanisa.
Ukuze kugwenywe inkinga yezakhiwo eziyinkimbinkimbi kanye nokunciphisa umsebenzi wamaqembu ochwepheshe okufanele baqeqeshe i-AI yokuhlukaniswa kwamaseli, u-Eiram Mahera Sheikh uthembela ku-AI. "Inhloso yami ukukhomba izifunda zesithombe ezibaluleke kakhulu ekufundeni ukuhlukaniswa kwamaseli. Lokhu kuzoqeda isidingo sokulebula mathupha wonke amaphikseli esithombe, kodwa lezo ezibaluleke kakhulu njengedatha yokuqeqeshwa." Lokhu kukhetha okuhlosiwe kwedatha yokuqeqeshwa okungaba khona kubizwa ngokuthi "ukufunda okusebenzayo." Ngenxa yokuthi ama-algorithms e-AI okuhlukaniswa kwamaseli asuka kulokho okubizwa ngokuthi "ukufunda okujulile," lokhu kubuye kubizwe ngokuthi "ukufunda okujulile okusebenzayo."
U-Sheikh manje usehlonze idatha yokuqeqeshwa ebaluleke kakhulu: "Lezi yizo ezingaqinisekile kakhulu, lezo i-algorithm ye-AI enobunzima kuzo, isibonelo, ngenxa yokungafani kwesithombe esibi noma amaseli agqagqene noma anomumo ongavamile." Ngakho-ke, yilawa kuphela okufanele ahlolwe futhi alebulwe ngochwepheshe esikhathini esizayo, kuyilapho umzamo ezimweni ezicacile ungase usindiswe ngoba i-AI ka-Sheik isivele ikwenzile ukulebula. Empeleni, i-cell segmentation AI ifunda ngale ndlela, ngedatha encane kodwa ngempumelelo kangangokuthi isengakwazi ukwenza izigaba zamaseli ezinembile ngokuhamba kwesikhathi.
Funda futhi
Ukusebenzisa i-AI ukusheshisa ukuthuthukiswa kwezinto ezingcono ze-photovoltaic | I-Pugnalom
I-AI inika amandla ukuhlola ukongiwa kwazo zonke izinhlobo zezilwane emhlabeni wonke | I-Pugnalom
