Deep Learning Pipeline for Accurate Wheat Spike Phenotypes Reveals Correlation with Yield

Researchers from the Chinese Academy of Sciences have developed a deep learning pipeline, called Speakerphone, to acquire precise wheat spike phenotypes. According to the study, traditional measurement methods are limited in capturing complex spike phenotypes with high precision, thus limiting progress in yield-related trait analysis. The pipeline achieved a mean intersection over union (mIoU) of 0.948 in spike segmentation and measured spike traits strongly agreed with manually measured values, with Pearson correlation coefficients of 0.9865 for spike length, 0.9753 for the number of spikelets per spike, and 0.9635 for fertile spikelets.

Key Takeaways:

• The research highlights the importance of accurate measurement of wheat spike phenotypes in determining grain yield.
• The deep learning pipeline, Speakerphone, achieved high precision in spike segmentation with a mean intersection over union (mIoU) of 0.948.
• The pipeline measured spike traits strongly agreed with manually measured values, with high Pearson correlation coefficients for spike length, number of spikelets per spike, and fertile spikelets.
• The study extracted 45 phenotypes from 221 wheat cultivars from various regions of Zhao County, Hebei Province, China, and analyzed their correlations with thousand-grain weight (TGW) and spike yield.
• The research identified phenotypic differences among wheat spikes, categorized them into six classes, and revealed effects on TGW and yield.
• The study also revealed phenotypic differences among wheat cultivars from different geographical regions and over decades, with an increase in the number of large-spike cultivars over time, especially in southern China.

Statistics:

• The deep learning pipeline, Speakerphone, achieved a mean intersection over union (mIoU) of 0.948 in spike segmentation.
• The pipeline measured spike traits with high Pearson correlation coefficients: 0.9865 for spike length, 0.9753 for the number of spikelets per spike, and 0.9635 for fertile spikelets.
• The study analyzed 45 phenotypes extracted from 221 wheat cultivars from various regions of Zhao County, Hebei Province, China.
• The research revealed phenotypic differences among wheat spikes, categorized them into six classes, and identified effects on TGW and yield.

Sources:

• NewsRx. New Plant Phenomics Findings Reported from Chinese Academy of Sciences (Analysis of Wheat Spike Morphological Traits By 2d Imaging). Life Science Weekly. October 21, 2025; p 3502.
• Analysis of Wheat Spike Morphological Traits By 2d Imaging. Plant Phenomics, 2025;7(3).
• Biological Breeding-National Sci-ence and Technology Major Project, National Key Research & Development Program of China.
• Chinese Academy of Sciences, Institute of Genetics and Developmental Biology, Beijing 100101, People's Republic of China.
• Amer Assoc Advancement Science, 1200 New York Ave, NW, Washington, DC 20005, USA.