advertisement

Editors Selection IGR 7-1

Visual function: Artificial intelligence

Comment by Christian Mardin on:

11704 Relevance vector machine and support vector machine classifier analysis of scanning laser polarimetry retinal nerve fiber layer measurements, Bowd C; Medeiros FA; Zhang Z et al., Investigative Ophthalmology and Visual Science, 2005; 46: 1322-1329

Find related abstracts

Bowd et al. (151) show the use of machine learning classifiers for the interpretation of the findings of the GDx VCC. The methods and results are well agreement with literature and our own experience on this topic with HRT. We need to be careful in the conclusions concerníng detection of glaucoma. The glaucoma group is already manifest with probably 30% neuroretinal rim loss. Demonstrating that the method is ok, does not necessarily mean it can be used for detecting early glaucoma. A second point of concern is that the classifiers have been trained on 'own' patients and normals. This is fine to classify own patients. And the GDx imanent classifier will always be worse in classifying, because it was trained on another population than the study population. I see the limitation only in the experience, that in other populations these new classifiers need to be adapted. This of course can be done. A third limitation is the relatively small number reported in this study. This is an interesting and novel study applying classifiers to GDX VCC.

• ← Previous
• Next →