Purpose Hyperspectral imaging combined with machine learning offers a promising, cost-effective alternative to invasive chemical analysis for early plant disease detection. In this study, the use of 3D Convolutional Neural Networks (3D-CNNs) was explored to detect presymptomatic viral infections in the model plant Nicotiana benthamiana L. and assess the generalization of these models across different plant genotypes. Methods Four genotypes of Nicotiana benthamiana L. (wild-type, DCL2/4, AGO2, and NahG) were inoculated with different potexviruses (PepMV mild or severe strain, PVX, BaMV). Viral infection was verified via northern blot analysis at 5 and 10 days post inoculation (DPI). Hyperspectral images were captured over 10 days following inoculation, focusing on the top 3 leaves where symptoms typically appear. The dataset was carefully processed to remove errors, and raster masks were generated to isolate only the leaf pixels. The Extremely Randomized Trees algorithm was used for Effective Wavelength selection, and a novel 3D-CNN architecture was developed to classify <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq1.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="91" /> </InlineMediaObject> <EquationSource Format="TEX">$16 \times 16 \times 16$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>16</mn> <mo>×</mo> <mn>16</mn> <mo>×</mo> <mn>16</mn> </mrow> </math></EquationSource> </InlineEquation> nonoverlapping cubes extracted from the unmasked leaf surfaces. The aim was to classify each cube into healthy or diseased for each of the four viruses at different time points. Results Accuracies of <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq2.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.78$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.78</mn> </mrow> </math></EquationSource> </InlineEquation>–<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq3.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.87$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.87</mn> </mrow> </math></EquationSource> </InlineEquation> were achieved for AGO2 mutants at the cube level, and overall plant-level accuracies of <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq4.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.68$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.68</mn> </mrow> </math></EquationSource> </InlineEquation>–<InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq5.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.89$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.89</mn> </mrow> </math></EquationSource> </InlineEquation>. The model’s generalization capabilities were tested across other genotypes, yielding accuracies of up to <InlineEquation ID="IEq6"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq6.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.75$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.75</mn> </mrow> </math></EquationSource> </InlineEquation> for DCL2/4, <InlineEquation ID="IEq7"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq7.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.83$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.83</mn> </mrow> </math></EquationSource> </InlineEquation> for NahG, and <InlineEquation ID="IEq8"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq8.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="30" /> </InlineMediaObject> <EquationSource Format="TEX">$0.78$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0.78</mn> </mrow> </math></EquationSource> </InlineEquation> for the wild-type. The timing of disease detection was also assessed, finding that accuracies approached 0.8 as early as <InlineEquation ID="IEq9"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq9.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="10" /> </InlineMediaObject> <EquationSource Format="TEX">$6$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>6</mn> </mrow> </math></EquationSource> </InlineEquation>–<InlineEquation ID="IEq10"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13007_2025_1337_Article_IEq10.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="10" /> </InlineMediaObject> <EquationSource Format="TEX">$8$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>8</mn> </mrow> </math></EquationSource> </InlineEquation> DPI depending on the virus. The results were validated against northern blot analyses and benchmarked against another state-of-the-art methodology for Nicotiana benthamiana viral infections, achieving superior overall classification accuracies. Conclusion The proposed patch-based method demonstrated key advantages: (a) exploiting both spectral and textural information, (b) deriving a large training dataset from few hyperspectral images, (c) providing localized classification explainability within leaf regions, and (d) achieving high accuracy for early detection of viral infections.

3D-CNN detection of systemic symptoms induced by different Potexvirus infections in four Nicotiana benthamiana genotypes using leaf hyperspectral imaging

期刊论文
出版模式：

开放获取
发表日期：

2025年2月10日

查看全文

Rizos-Theodoros Chadoulis,
Ioannis Livieratos,
Ioannis Manakos,
Theodore Spanos,
Zeinab Marouni,
Christos Kalogeropoulos,
Constantine Kotropoulos

摘要

Purpose

Hyperspectral imaging combined with machine learning offers a promising, cost-effective alternative to invasive chemical analysis for early plant disease detection. In this study, the use of 3D Convolutional Neural Networks (3D-CNNs) was explored to detect presymptomatic viral infections in the model plant Nicotiana benthamiana L. and assess the generalization of these models across different plant genotypes.

Methods

Four genotypes of Nicotiana benthamiana L. (wild-type, DCL2/4, AGO2, and NahG) were inoculated with different potexviruses (PepMV mild or severe strain, PVX, BaMV). Viral infection was verified via northern blot analysis at 5 and 10 days post inoculation (DPI). Hyperspectral images were captured over 10 days following inoculation, focusing on the top 3 leaves where symptoms typically appear. The dataset was carefully processed to remove errors, and raster masks were generated to isolate only the leaf pixels. The Extremely Randomized Trees algorithm was used for Effective Wavelength selection, and a novel 3D-CNN architecture was developed to classify $16 \times 16 \times 16$ $16 \times 16 \times 16$ nonoverlapping cubes extracted from the unmasked leaf surfaces. The aim was to classify each cube into healthy or diseased for each of the four viruses at different time points.

Results

Accuracies of $0.78$ $0.78$ – $0.87$ $0.87$ were achieved for AGO2 mutants at the cube level, and overall plant-level accuracies of $0.68$ $0.68$ – $0.89$ $0.89$ . The model’s generalization capabilities were tested across other genotypes, yielding accuracies of up to $0.75$ $0.75$ for DCL2/4, $0.83$ $0.83$ for NahG, and $0.78$ $0.78$ for the wild-type. The timing of disease detection was also assessed, finding that accuracies approached 0.8 as early as $6$ $6$ – $8$ $8$ DPI depending on the virus. The results were validated against northern blot analyses and benchmarked against another state-of-the-art methodology for Nicotiana benthamiana viral infections, achieving superior overall classification accuracies.

Conclusion

The proposed patch-based method demonstrated key advantages: (a) exploiting both spectral and textural information, (b) deriving a large training dataset from few hyperspectral images, (c) providing localized classification explainability within leaf regions, and (d) achieving high accuracy for early detection of viral infections.