Early and Noninvasive Bird Gender Identification by ATR-FTIR Spectra Coupled with a Randon Forest Algorithm

Area of Science:

• Avian biology
• Spectroscopy
• Machine learning

Background:

• Early sex identification in birds is vital for breeding programs, wildlife management, and conservation efforts.
• Current DNA-based methods are accurate but costly and time-consuming, requiring specialized facilities.
• There is a need for accessible, rapid, and precise techniques for avian gender determination.

Purpose of the Study:

• To investigate the efficacy of Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) coupled with the Random Forest algorithm for noninvasive, early gender determination in birds.
• To evaluate this novel approach as a cost-effective and precise alternative to traditional molecular methods.
• To assess the accuracy of ATR-FTIR in distinguishing sexes across four different bird species.

Main Methods:

• Feather samples (vexillum) from four bird species (Bicudo, Cockatiel, Curio, Ring-necked Parakeet) were analyzed using ATR-FTIR spectroscopy (3800-800 cm⁻¹).
• Spectra were pre-processed using Standard Normal Variate (SNV) and dimensionality was reduced via Principal Component Analysis (PCA).
• The Random Forest algorithm was employed to classify bird sexes based on the processed spectral data, with DNA analysis used for confirmation.

Main Results:

• The Random Forest classifier, utilizing ATR-FTIR spectral data, demonstrated significant potential for avian sex determination.
• High accuracy rates were achieved in external validation: 94.4% for Bicudo and Curio, 77.8% for Cockatiel, and 72.2% for the Ring-necked Parakeet.
• The study successfully identified spectral differences indicative of sex in the analyzed feather samples.

Conclusions:

• ATR-FTIR spectroscopy combined with the Random Forest algorithm presents a promising, noninvasive, and cost-effective tool for early bird gender identification.
• This technique offers a viable alternative to conventional DNA-based methods, particularly in resource-limited settings.
• Further research can optimize this spectroscopic approach for broader application in avian conservation and management.