瞳孔対光応答の波形特徴による眼疾患の推定に関する検討

摘要

本報告では，加齢などによる眼疾患である加齢黄斑変性（AMD）を，4種類の光刺激による瞳孔対光応答の波形特徴間の非類似度による判別可能性を検討した．瞳孔応答の波形形状をフーリエ記述子で特徴量を表現し，重み係数をつけた瞳孔応答の振幅成分を特徴に加えた．健康，高齢，AMDの瞳孔応答を，RandomFbrest法を用いて判別分析を行った．誤判別率は振幅成分の重み係数が1．5の時に，最小値となった．誤判別率は健康，AMD，高齢の順に高くなった．また，判別に寄与する特徴量は，赤色の低輝度光と青色光との瞳孔応答における非類似度が大きいことを示した．%This report determine a possibility of classification for Pupil Light Response (PLR) with Age-Related Macular Degeneration (AMD) using the features of waveform shapes to four chromatic stimuli. The features are extracted with Fourier descriptors, and the weighted amplitudes of waveform are also considered as a component. Classification analysis was conducted with Random Forests technique using Euclid distances amongst features of waveforms. The weight value is optimized as 1.5, the error rates were evaluated from low to high as Healthy, Aged, and AMD. The distances between waveforms for red light-low luminance and blue lights contribute on the classification.