A method using artificial neural networks to morphologically assess mouse blastocyst quality

Abstract

See, stats, and : https : / / www . researchgate . net / publication / 265706428 A morphologically quality Article DOI : 10 . 1186 / 2055 - 0391 - 56 - 15 READS 31 3 : Felipe Ecole 6 SEE José São 13 SEE Marcelo São 64 SEE Available : Marcelo Retrieved : 06 Abstract Background :Morphologicallyclassifyingembryosisimportantfornumerouslaboratorytechniques,whichrangefrombasicmethodstomethodsforassistedreproduction.However,thestandardmethodcurrentlyusedforclassificationissubjectiveanddependsonanembryologist'spriortraining.Thus,ourworkwasaimedatdevelopingsoftwaretoclassifymorphologicalqualityforblastocystsbasedondigitalimages.Methods:Thedevelopedmethodologyissuitablefortheassistanceoftheembryologistonthetaskofanalyzingblastocysts.Thesoftwareusesartificialneuralnetworktechniquesasamachinelearningtechnique.Thesenetworksanalyzebothvisualvariablesextractedfromanimageandbiologicalfeaturesforanembryo.Results:Afterthetrainingprocessthefinalaccuracyofthesystemusingthismethodwas95%.Toaidtheend-usersinoperatingthissystem,wedevelopedagraphicaluserinterfacethatcanbeusedtoproduceaqualityassessmentbasedonapreviouslytrainedartificialneuralnetwork.Conclusions:Thisprocesshasahighpotentialforapplicabilitybecauseitcanbeadaptedtoadditionalspecieswithgreatereconomicappeal(humanbeingsandcattle).Basedonanobjectiveassessment(withoutpersonalbiasfromtheembryologist)andwithhighreproducibilitybetweensamplesordifferentclinicsandlaboratories,thismethodwillfacilitatesuchclassificationinthefutureasanalternativepracticeforassessingembryomorphologies.