Rapid measurement and machine learning classification of colour vision deficiency

Area of Science:

• Ophthalmology
• Genetics
• Computer Science

Background:

• Colour vision deficiencies (CVDs) are linked to genetic variations and acquired neuro-ophthalmic diseases.
• Current CVD tests detect presence but lack type/severity quantification.

Purpose of the Study:

• Introduce Foraging Interactive D-prime (FInD), a novel computer-based vision assessment tool for colour vision testing.
• Evaluate FInD's sensitivity, repeatability, and ability to classify CVD type and severity.

Main Methods:

• FInD utilizes a signal detection theory-based adaptive paradigm with chromatic blobs in dynamic luminance noise.
• Participants detected blobs or discriminated differing colours; stimulus intensity was derived from d-prime analysis.
• Compared FInD against Hardy-Rand-Rittler and Farnsworth-Munsell 100 hue tests in colour-normal and atypical observers.

Main Results:

• Atypical observers had higher detection and discrimination thresholds than typical observers.
• Threshold elevations corresponded to specific CVD types, enabling classification.
• Unsupervised machine learning confirmed functional CVD subtypes based on FInD results.

Conclusions:

• FInD reliably detects inherited CVDs and quantifies type and severity.
• This novel tool shows potential for basic and clinical colour vision research.
• FInD offers a generalisable, rapid, and self-administered approach to vision assessment.