Develop an AI-based decision support system to improve viability status monitoring of in vitro collections.

Challenge Summary

The CIP’s genebank is one of the world’s largest in vitro genebanks, conserving the global collection of potato, sweetpotato, and Andean Roots and Tuber crops. In vitro conservation applies plant tissue culture techniques to culture clonal crop propagules (e.g., stem cuttings, shoots, meristems) in a sterile medium. These cultures are kept in a controlled environment to ensure their healthy and safe growth. 

However, in vitro conservation, while the most effective way to preserve and distribute the genetic resources of clonal crops, has several disadvantages. Maintenance costs are high due to the need for specialized equipment and materials.  The process is labor-intensive, requiring skilled personnel for precise handling and maintenance. Additionally, constant monitoring and periodic regeneration of the cultures are necessary to maintain the viability of the plants, further increasing the resources and effort needed. 

Currently, the collections are monitored every 1-3 months through a process in which the tubes containing the plantlets are individually inspected and then classified into four viability categories: good, medium, regular, and critical. Plantlets falling into the regular and critical categories are subjected to immediate regeneration to preserve their viability. This evaluation relies on visual assessments of the extent of necrosis on shoots, roots, and leaves, as well as changes in leaf coloration, such as chlorosis and dark pigmentation. Highly skilled and experienced technicians are required for this critical task, as incorrect categorization can lead to reduced viability, necessitating additional labor, financial resources, and time to restore plant vigor and viability. In the worst cases, it may even result in the loss of an accession. The distinction between the medium and regular categories is particularly susceptible to errors.  

The challenge focuses on creating an AI-driven image scanning and classification process with the aim to serve as a valuable decision support tool, enhancing the accuracy of classifications and improving the efficiency of in vitro maintenance. 

Potential AI Use

AI models trained on plant images can automatically classify them in real time into the viability categories (good, medium, regular, critical) based on visual cues such as necrosis, leaf coloration, and other health indicators. This reduces reliance on manual, subjective assessments and speeds up the categorization process.  

Machine learning algorithms can be designed to identify any anomalies or unusual patterns in plant growth that might not fit neatly into the predefined categories but indicate a need for attention. This can help in early detection of potential issues before they become critical. 

AI can analyze historical data on plantlet viability and outcomes to predict future regeneration needs. By understanding patterns in how certain conditions affect plant health, AI can forecast which plantlets are likely to require regeneration soon, optimizing the timing and resources used for maintenance.

Deep learning techniques can be applied to extract detailed phenotypic data from images, such as precise measurements of leaf size, shape, and color variations. This detailed analysis can enhance the understanding of plant health and growth patterns, leading to more informed decisions about regeneration and maintenance. 

Implementing blockchain technology can ensure the traceability of each decision made by the AI system, providing a transparent and immutable record of when and why each plantlet was classified into its viability category. This can help in refining the AI model over time and ensuring accountability in the maintenance process.

Social Impact – Context

The integration of AI-driven image scanning and classification processes improves the efficiency and reliability of maintaining plant collections, allowing genebanks to provide broader access to these resources. This effort aligns with broader goals of enhancing food security, promoting sustainable agriculture, and preserving genetic diversity. By ensuring the accurate categorization and health of plant collections, genebanks can more effectively contribute to preserving agricultural biodiversity. This is crucial for mitigating challenges posed by climate change, pests, and diseases. Moreover, the detailed and reliable data generated through AI technologies offer invaluable insights for policy and decision-making processes. These insights are essential for formulating strategies related to genetic conservation, agricultural biodiversity, and food security, thereby reinforcing the role of genebanks in shaping national and international approaches to sustainable agriculture. Through these contributions, genebanks, powered by AI can maintain their critical role in conserving genetic resources in securing the future of global food systems and preserving the genetic heritage of essential crops. 

Solution Impact

The development and implementation of an AI-driven image scanning and classification procedure for in vitro viability monitoring is expected to yield significant benefits:  

  • Increased Accuracy in Viability Assessments: Enhanced precision in categorizing plant health and viability reduces errors, leading to better maintenance and conservation outcomes. 
  • Efficiency Improvements: Streamlines the monitoring process, saving time and resources by automating the categorization of plantlets. 
  • Reduced Dependency on Skilled Labor: While expertise remains valuable, the system’s reliance on highly skilled technicians for routine assessments is diminished, easing labor constraints. 
  • Enhanced Genetic Resource Preservation: Improved detection and categorization capabilities aid in the early identification of issues, enhancing the overall conservation of genetic diversity. 
  • Cost Reduction: Automating the monitoring process can lead to significant savings in labor and operational costs over time. 
  • Data-Driven Insights for Research: Generates valuable data on plant viability trends and responses to in vitro conditions, supporting scientific research and eventually breeding programs. 
  • Global Accessibility and Sharing: Facilitates the sharing of genetic resources by providing reliable data on plantlet viability, benefiting global agricultural development and food security initiatives. 

CIP Mentoring Team

2   0   2   4