John Wells, the author of the Longman Pronunciation Dictionary, uses a definition of stress that treats the primary stressed syllable in a word as the rightmost stressed syllable. This means that in words like digitize and utilize, where the primary stress is on the first syllable, neither of the two following syllables has any stress: ˈdigitize and ˈutilize. (This contrasts with the tradition of American dictionaries which tend to mark the final syllables of these words with a minor stress mark: ˈdigiˌtize and ˈutiˌlize.) For more on the concept of primary stress, secondary stress and their positions, see my posts here:
However, the second and third syllables of utilize and digitize behave differently in terms of certain affects on the adjacent sounds: in many accents of American English, the "t" in utilize may be voiced and "flapped/tapped" (roughly, [ˈjuɾəlaɪz]) while the "t" in digitize may not be (roughly, [ˈdɪdʒətaɪz]).
Wells' explanation of the difference is that t is flapped in American English only before "weak" vowels, not before "strong" vowels.
Incidentally, I'm not sure that Wells is correct in his formulation of the t-flapping rule. Subjectively, flapping seems possible to me (I'm an American English speaker) in the edge case of Latin-style plurals in /aɪ/ in words where the singular ends in [ɾəs]: emeritus [əˈmerəɾəs], emeriti [əˈmerətai], ?[əˈmerəɾai] or stratus [ˈstreɾəs], strati ?[ˈstreɾai]. However, my intuition that [ɾai] is possible here is not supported by dictionaries such as Merriam Webster and the American Heritage Dictionary, which transcribe the last syllable of emeriti with a minor stress mark.
In any case, Wells uses the "strong vowel"/"weak vowel" concept not only to explain t-flapping, but also other things like a vowel's tendency to have a reduced pronunciation.