ABSTRACT
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 4.A Appendix: Measures of Binary Test Performance . . . . . . . . . . . . . . . . . . . . . . . . . . 120
There are strong reasons why well-designed retinopathy risk reduction programs need to be implemented on a large scale. The importance of such programs, and the difficulty of achieving reliable early diagnosis of diabetic retinopathy (DR) at reasonable cost, merit considerable efforts in order to develop and evaluate computer-aided diagnosis (CAD). So why is CAD software not yet widely used in this setting? Problems and diverse application possibilities encountered by people responsible for such programs need to be considered in conjunction with the opportunities and threats facing anyone introducing computer-aided diagnosis in this field.
