Equipment/Technology, Europe, UK, Ireland, Pests and Diseases, Research, Studies/Reports

Innovative potato disease detection: The potential of hyperspectral imaging

This article was prepared and written by Jorge Luis Alonso G.

Seed potatoes with latent infections contribute significantly to the spread of Phytophthora infestans. Identifying these infections before symptoms appear is critical. Relying on manual inspection alone isn’t practical, and models based on spectral data alone can be inconsistent due to varying local conditions.

Researchers at Wageningen University & Research has introduced a technique using hyperspectral imaging, which contrasts the spectra of infected plants with those of healthy plants under analogous conditions, highlighting the differences. The study was published in Volume 6 of the journal Smart Agricultural Technology.

The study and conclusions are summarized below.

Introduction

Phytophthora infestans is a formidable adversary to potato crops and is responsible for the annual late blight epidemic. Without appropriate countermeasures, the disease wreaks havoc on both the foliage and the tubers, resulting in significant losses. In the Netherlands in particular, growers resort to 10 to 15 rounds of fungicide spraying each year just to keep the disease at bay. However, there is growing concern. P. infestans has a worrying ability to rapidly develop resistance to the fungicides currently in use.

A major source of introduction of this pathogen is seed potatoes carrying latent infections. It’s worth noting that the rate at which these infected seed tubers transmit the disease to plants ranges from 1.5% to 5%. This rate of transmission suggests that up to two infected plants may occur per hectare. Therefore, early detection and removal of these plants could serve as a strategic intervention, giving potato plants more time to grow their tubers.

Turning to technology, hyperspectral imaging (HSI) is emerging as a beacon of hope. By fusing imaging techniques with point spectrometry, HSI provides a window into plant structures while shedding light on the interaction between plants and specific wavelengths. In the laboratory, this technology has proven itself by detecting early infections of P. infestans before they are visible to the naked eye.

Using HSI, pioneering researchers have calibrated a classifier to distinguish between healthy and infected plant tissue. Encouragingly, preliminary tests have demonstrated its ability to detect infection much earlier than traditional human assessment. However, it’s important to understand that relying solely on spectral data for early detection is not without its challenges.

To illustrate, in an experiment where potato leaves were treated with P. infestans, the system misclassified the vast majority of plants as healthy — even though some showed obvious symptoms. In another twist, two identical experiments conducted weeks apart still allowed the model to discriminate between untreated plants from both experiments. This underscores the inadequacy of spectral information alone.

Therefore, relying on spectrometry alone to effectively detect P. infestans in plants is proving to be insufficient. Instead, a comprehensive approach that integrates imaging is critical. This involves comparing the imaging with the reflectance spectra of healthy plants grown under comparable conditions. This study presents a method that captures image data based on spectral measurements. Importantly, this approach involves comparing these measurements to images of healthy plants. It’s critical to note that these healthy plants are grown under conditions that mirror those of the infected specimens.

Deciphering Potato Resistance: Four experiments on treatment effects

In a series of studies, the effects of different treatments on potato plants were examined in four different experiments.

Experiment A, conducted in 2020, involved 80 potato plants receiving one of four treatments. These treatments mainly revolved around different inoculation methods. A detailed breakdown of these methods can be found in the corresponding research references.

The following year, 2021, Experiment B was conducted. 45 potato plants were divided into three groups of 15 plants each. Each group received a different treatment: no treatment, spray inoculation, or stem inoculation. The method used was a specific P. infestans sporangial suspension, which was later adjusted for density. After inoculation, the plants were maintained under specific conditions to observe any changes.

Experiment C, also in 2021, dealt with another set of 45 plants. These were divided into three treatment types: no inoculation, tuber inoculation or a combination of tuber and soil inoculation. The procedure began with the application of P. infestans to potato slices, which were then applied to the tubers. The tubers were then planted and monitored for additional treatments during the growing season.

The series was completed in 2022 with Experiment D. This focused on potatoes grown in the field. In this arrangement, three rows of 80 plants each were observed. They underwent three unique treatments in which the P. infestans genotype EU-36-A2 was used for inoculation. After a specific incubation period under defined conditions, these tubers were planted and standard agricultural practices followed.

Conclusion

This research focused on using hyperspectral imaging to detect potato late blight in its early stages, before the onset of sporulation. It’s worth noting that the reflectance of plants can vary significantly depending on their growing conditions. In addition, a variety of factors, not least other diseases, can alter this spectrum. Therefore, it is important to investigate the mechanisms by which these spectral shifts occur.

Experiments A and B were specifically designed to shed light on this issue, and they successfully elucidated the patterns of these spectral changes. However, an interesting challenge arose: when the convolutional neural network (CNN) was trained on the data from Experiment A, it proved ineffective on the data from Experiment B. This inconsistency led the researchers to believe that there may be distinct differences in the way the disease manifests when tubers are inoculated as opposed to leaves. Experiments C and D were therefore commissioned.

In the subsequent field trials, the changes in reflectance were quite subtle and barely noticeable. In particular, in Experiment D, although the inoculated tubers were stunted, many didn’t show any obvious symptoms. This ambiguity posed a significant challenge in separating healthy plants from their diseased counterparts.

Source: Kool, J., & Evenhuis, A. (2023). Early detection of Phytophthora infestans in potato plants using hyperspectral imaging, local comparison and a convolutional neural network. Smart Agricultural Technology, 6, 100333. https://doi.org/10.1016/j.atech.2023.100333
Author: This article was prepared and written by Jorge Luis Alonso G.

Editor & Publisher: Lukie Pieterse


Feel free to get in touch with Lukie!
He’ll be happy to share your company’s news stories on Potato News Today:
lukie@potatonewstoday.com
Connect on LinkedIn
Follow on Twitter
About us

Stay updated with breaking news

Enter your email below for news story notifications.

Translate this site

Blog Stats

  • 1,840,526 reads of our news stories

LOCKWOOD Mfg

PULSEMASTER

FPS Food Process Solutions

FAM STUMABO

AVR

Scotts Precision Manufacturing

TOMRA FOOD

GRIMME