Epilepsy　is　a　chronic　disease　with　high　prevalence　and　high　disease　burden.　Electroencephalogram　(EEG)　is　one　of　the　most　important　tools　for　diagnosing　epilepsy.　Detection　of　epileptic　spikes　from　EEG　can　effectively　aid　the　diagnosis　of　epilepsy　but　it　is　also　a　time-consuming　and　laborious　process　for　neurologists,　which　can　only　be　addressed　by　automatic　spike　detection　methods.　However,　existing　approaches　are　limited　by　the　simplistic　nature　of　conventional　machine　learning　algorithms.　The　work　researched　about　how　human　experts　identify　spikes　and　over-parameterized　the　process　into　a　deep　neural　network　named　temporal-frequential　and　multi-order　difference　fusion　(TeFreDiA)　model　which　can　accurately　detect　epileptic　spikes.The　model　was　evaluated　on　data　collected　from　7　subjects　with　ten-fold　cross-validation.　Given　an　average　accuracy　and　sensitivity　of　95.64%　and　96.89%,　the　model　can　be　utilized　to　aid　the　diagnosis　of　epilepsy.　©　2023　Copyright　held　by　the　owner/author(s).