Sage Pourpre's helpful answer shows how to parse your multiline string line by line into an ordered hashtable (dictionary, of type [ordered]), which preserves the input order, i.e. the order of the key-property values encountered in the input string.
A simpler alternative that does not preserve the input order, however,[1] is to use the ConvertFrom-StringData cmdlet, which requires replacing the : separators in your input with = in order to make the cmdlet recognize the input lines as key-value pairs:
$items = @("Title : 123456
field_1 : User One
field_2 : User
field_3 : One
field_4 : 26
field_5 : J123456")
ConvertFrom-StringData ($items -join "`n" -replace ':', '=')
Note:
-join "`n" joins the elements of the array with newlines to form a single, multiline string to pass to ConvertFrom-StringData; in your case, given that there's only one element, that one element (which is a multiline string itself) is returned as-is.
As zett42 points, out, ConvertFrom-StringData treats \ as the escape character, so if verbatim \ are in your input, you'd have to double them first, with -replace '\\', '\\' (sic).
Resulting output (a [hashtable], with inherently unordered entries):[1]
Name Value
---- -----
field_3 One
field_4 26
field_2 User
Title 123456
field_1 User One
field_5 J123456
[1] GitHub issue #19070 is a feature request to make ConvertFrom-String data output a dictionary type that does preserve the input ordering, using a type derived from [hashtable] so as to preserve backward compatibility, something that has already been successfully implemented in the v7.3+ in the ConvertFrom-Json cmdlet via its -AsHashtable switch. Unfortunately, as of this writing (v7.5.x), the feature request is languishing.