Nondestructive 3D phenotyping method of passion fruit based on X-ray micro-computed tomography and deep learning

Area of Science:

• Agricultural Science
• Plant Biology
• Biotechnology

Background:

• Traditional passion fruit phenotyping is labor-intensive and lacks accuracy, especially for internal structures.
• Developing efficient and objective methods is crucial for improving passion fruit varieties.
• Understanding genotype-phenotype relationships requires precise trait data.

Purpose of the Study:

• To develop an automated, non-destructive Micro-CT imaging system for passion fruit phenotyping.
• To accurately quantify multiple internal and external fruit traits.
• To analyze correlations between phenotypic traits and assess comprehensive fruit quality.

Main Methods:

• Utilized a Micro-CT system for automated, non-destructive 3D imaging of passion fruit.
• Employed a U-Net based deep learning model for tissue segmentation with >0.95 accuracy.
• Calculated fourteen phenotypic traits including volume, surface area, and sarcocarp content.

Main Results:

• Micro-CT system achieved high accuracy comparable to manual measurements for external traits (e.g., length, width).
• Deep learning model demonstrated excellent segmentation accuracy for internal tissues.
• Identified significant correlations among fourteen phenotypic traits using Pearson analysis.
• Principal component analysis enabled comprehensive quality scoring for passion fruit.

Conclusions:

• The developed Micro-CT system offers a non-destructive, accurate, and efficient method for passion fruit phenotyping.
• This approach facilitates molecular breeding and comprehensive quality evaluation.
• The methodology has potential for application to other fruit crops.