Archetypal analysis for dominant optic atrophy visual field loss

Abstract

Aims/Purpose: Identification of the characteristic visual field loss patterns of dominant optic atrophy (DOA) employing the Archetypal Analysis (AA) machine learning algorithm and further relation with patient‐specific parameters.Methods: For 64 patients affected by molecularly confirmed DOA with OPA1 heterozygous mutation, binocular visual field (VF) tests performed by SITA standard 30–2 or 24–2 Humphrey VF analyser (Carl Zeiss Meditec, Dublin, CA, USA) were collected. For patients that had multiple VF tests, several VFs were collected to enlarge the dataset. Resorting to Python 3.8 (Anaconda) an AA model was developed to determine and quantify the DOA underlying patterns of visual loss. After a careful analysis involving different linear and non‐linear machine Learning models and algorithms, the different archetypes (AT) were related to parameters such as age and visual acuity.Results: Considering 229 VF test, employing the AA, twelve archetypes were detected for the characterization and quantification of visual loss in DOA. Central, ceco‐central, and para‐central ATs revealed to be the most significant ones, matching the common characteristic scotomas for DOA patients, followed by quadrantanopia, altitudinal, and nasal step ATs.Younger patients were related to less severe visual loss representative ATs, while patients with lower visual acuity were related to more severe visual loss representative ATs.Conclusions: Archetypal analysis shows to be a potential powerful tool to support clinicians in the identification of DOA patients by identifying and quantifying visual loss patterns characteristic of this disease. Relation with other patient‐specific parameters might strengthen the differential DOA diagnosis.