Abstract
Our methodology offers a straightforward solution for setting up automated, scalable and low-cost phenotyping facilities in a wide range of lab environments that could greatly increase the processing power and scalability of Arabidopsis soil growth experiments.
Introduction
In modern plant biology, the ability to sequence genomes has far outpaced the ability to measure physical traits—a phenomenon known as the "phenotyping bottleneck." While DNA mapping is now relatively inexpensive, measuring how a plant grows in response to its environment remains labor-intensive and costly. Commercial high-throughput systems are often out of reach for smaller laboratories, and many existing solutions struggle to monitor plants during the dark period without disrupting their circadian rhythms. This paper presents an affordable, automated system designed to capture high-resolution growth data across the full 24-hour diel cycle.
Hardware and Software Design
The researchers developed a hardware-software co-design approach to create a system costing less than £100.
The Image Capture System (ICS)
The hardware core consists of a Raspberry Pi equipped with a NoIR (No Infrared filter) camera. To enable nighttime imaging without affecting plant growth or circadian signaling, the team designed a custom 940 nm near-infrared (NIR) LED panel. Since plants do not utilize this wavelength for photosynthesis, the system can monitor growth 24/7 without "waking" the plants.
iDIEL Plant Software
The accompanying software module, iDIEL Plant, was developed in MATLAB to handle the complex task of image segmentation—separating the plant from the soil background.
- Daytime Imaging: The software utilizes thresholding in the HSV (Hue, Saturation, Value) color space, providing robustness against varying light conditions.
- Nighttime Imaging: To overcome the low contrast of NIR images, the system employs a Chan-Vese active contour method. It cleverly uses the final visible-light image of the day as a "starting mask" to guide the algorithm in identifying the plant in subsequent dark frames.
Experimental Validation
The system was validated using Arabidopsis thaliana wild-type (Col-0) and two Rubisco-deficient mutants (1a2b and 1a3b). Images were captured every 20 minutes over a 16-day period. The software's accuracy was benchmarked against both external datasets and "ground truth" destructive measurements, where leaves were physically scanned to compare 2D image estimates against actual leaf area.
Key Results and Insights
- High Accuracy and Low Cost: The entire setup was constructed for under £100. Despite the low cost, the software's estimates for Projected Rosette Area (PRA) achieved a Mean Absolute Percentage Error (MAPE) of approximately 5% compared to ground-truth measurements.
- Capturing the "Night Shift": High-frequency imaging revealed that growth is not constant. The system captured fluctuations in Relative Expansion Rates (RER) between day and night.
- Circadian Rhythm Analysis: The system successfully tracked hyponasty (rhythmic leaf movement). While wild-type plants showed a characteristic "dip" in projected area at dawn due to leaf movement, certain mutants lacked this rhythm, indicating compromised carbon management.
- Mutant Phenotyping: The system quantified that the 1a3b mutant was eightfold smaller than the wild type by day 24. Interestingly, younger mutants exhibited higher growth peaks during the dark period, likely a compensatory response to reduced photosynthetic capacity.
Discussion and Limitations
While highly effective, the authors note that 2D imaging serves as a proxy for growth. Top-down cameras can be confounded by leaf overlap and vertical movements, which may underrepresent the actual leaf area. Future developments could integrate 3D imaging to better account for complex plant architecture. Additionally, the use of standardized dark soil or felt covers is recommended to avoid interference from reflective soil particles like perlite or vermiculite.
Overall, this "Do-It-Yourself" system provides a scalable and accessible solution for labs to overcome the phenotyping bottleneck and gain deeper insights into plant growth dynamics.
Summary updated