Inferring tephritid fly pupal development stage using non-invasive near infrared imaging and machine learning...

Area of Science:

• Entomology
• Agricultural Science
• Computer Science

Background:

• Accurate determination of insect pupal age is vital for managing agriculturally significant pests like Tephritid flies.
• Current methods for age determination are invasive, requiring specimen dissection and hindering longitudinal studies.
• Non-invasive techniques are needed to assess pupal physiological age ethically and efficiently.

Purpose of the Study:

• To develop and validate a non-invasive method for determining the physiological age of Mexican fruit fly (Anastrepha ludens) pupae.
• To apply machine learning algorithms to near-infrared (NIR) imaging data for pupal age estimation.
• To assess the potential of this method for estimating pupal viability without awaiting adult emergence.

Main Methods:

• Utilized non-invasive near-infrared (NIR) imaging (850-1100 nm) to capture images of Anastrepha ludens pupae throughout their development.
• Photographed pupae at regular intervals under controlled conditions (26°C, 75-80% RH).
• Analyzed NIR images using a convolutional neural network (CNN) to classify pupal developmental stages and estimate physiological age.

Main Results:

• The intrapuparial period for Anastrepha ludens ranged from 17 to 19 days.
• Key morphological changes, including eye darkening (day 12) and wing/leg/setae melanization (days 13-15), were observed.
• The developed CNN model achieved an average accuracy of 71.77% in identifying physiological age ranges.

Conclusions:

• Near-infrared imaging combined with machine learning provides a non-invasive method for determining the physiological age of Anastrepha ludens pupae.
• This approach allows for developmental stage assessment without harming the pupae.
• The method shows promise for estimating pupal viability, improving pest management strategies.