lsadump::lsa extracts hashes from memory by asking the LSA server. The patch or inject takes place on the fly. It has the following command line arguments:
/name or /user : the target user account
/id : the RID (relative identifier) for the target account (500 for Administrator)
/patch : Only dumps the LM and NT password hashes
/inject : when run on a workstation, it will dump the LM and NT password hashes. When run on domain controller is will dump LM, NT, Wdigest, Kerberos keys and password history.
LM and NT hashes are used to authenticate accounts using the NTLM protocol. These hashes are often called NTLM hash and many documentations, resources, blogpost and tools mix terms. In this case, "NTLM" refers to the NT hash.