A different view on Casson-Gordon invariants for knots