Validation of a semi-automatic image analysis system to age squids and its application to age Illex coindetii statoliths

Abstract

The validity of a semi-automatic image analysis system employing digital image enhancement was demonstrated by re-analysing a set of previously analysed, known-age Sepioteuthis lessoniana statoliths. Linear regression models using both the manual and semi-automatic growth increment counts successfully predicted the true age of the test statoliths to a high degree of accuracy and precision and fully supported the one increment per day hypothesis. Thus this set of known age statoliths served as a reference standard for the counting method. After "calibrating" the semi-automatic image analysis system with the reference standards the method was then used to determine the ages of a large (n=312) set of Illex coindetii statoliths of unknown age. The results of this analysis were less clear than the S. lessoniana analysis. The semi-automatic counts predicted mantle lengths (ML) more precisely than did the manual counts. Significant differences were found between the two counting methods. For the Illex data set, the linear model to estimate ML from putative age predicted negative y-intercept values for both sexes and counting methods. Thus a power function model was selected as the more biologically meaningful predictor of ML. The accuracy of the I. coindetii analysis cannot be determined as the daily increment hypothesis remains to be verified for this species.