Terahertz time-of-flight contrast for non-destructive visualization of bruchid damage in lentils: a feasibility study

Lentils are a key source of plant-based proteins, but their quality can be compromised by bruchids, insects that burrow inside the seeds, causing internal damage. Current detection methods rely on destructive testing, which is time-consuming, costly, and results in the loss of part of the batch. In this study, we demonstrate the potential of terahertz time-domain imaging in transmission as a non-destructive and effective method for detecting bruchids in lentils. The approach provides rapid and precise assessment, preserving the integrity of the samples and enabling quality control for the agro-industrial sector. Beyond accurate quality evaluation, this method holds promise for future applications in automated sorting systems, improving product valuation and market competitiveness. By leveraging terahertz technology, we present an innovative solution that aligns with the needs of producers, distributors, and consumers. This work is a qualitative proof-of-concept aimed at establishing physical feasibility. It does not estimate diagnostic performance (sensitivity/specificity); a powered validation on concealed larvae is planned.

Lentils occupy a important place in human nutrition thanks to their high protein (Aghababaei et al., 2024; Shrestha et al., 2023), fiber, and mineral content (Aryee and Boye, 2017). Lentil proteins (LP) are gaining attention as a sustainable and nutritional alternative to animal proteins due to their high digestibility, rich amino acid profile, and antioxidant properties (Jarpa-Parra, 2018). They demonstrate versatile functionalities, including solubility, emulsification, foaming, and gelling capacities, making them suitable for diverse food applications such as meat substitutes, bakery products, and dairy alternatives (Keskin and Sumnu, 2024; Miranda et al., 2022).

Protein production is essential because proteins are vital for human health, animal feed, and various industrial applications. In France, the protein deficit, particularly in plant-based protein sources such as soy and legumes, highlights the nation’s dependance on imports, which poses risks to food security and the country’s dependence (Gohin and Issanchou, 2022). Addressing this deficit by boosting local protein production would enhance self-sufficiency and reduce dependency on imported genetically modified organisms (GMOs), in line with environmental and consumer demands (Duflot, 2023).

However, the quality of these legumes can be considerably affected by the presence of bruchids (Baxevanos et al., 2024; Vlachostergios et al., 2018), which are insects whose larvae burrow inside the seeds. These pests cause internal damage that reduces the commercial value of the batches of lentils and raises concerns about food safety (Jaswal and Menon, 2020; Pérez de la Vega et al., 2022). Bruchid beetles significantly lower the value of lentils by damaging their appearance, making them unsuitable for premium markets like human consumption. Infested grains are often rejected, affecting both food and feed markets and reducing international competitiveness as described in market reports in European Union (Sepngang et al., 2020).

Lentil crops face significant threats from insect pests, which can impact yield and quality. To address these challenges, various pest management strategies are employed, often as part of an Integrated Pest Management (IPM) system as described in Stevenson et al. (2007). These strategies include chemical insecticides, such as methidathion, deltamethrin, and endosulfan, are commonly used to manage pests such as aphids, thrips, and pod borers, especially during flowering and pod formation. Although effective, the use of chemicals must be carefully managed to avoid resistance development and minimize environmental harm. Biological control utilizes natural enemies of pests, such as predators and parasitoids. For example, Braconidae and Trichogrammatidae species are effective against bruchids and other field and storage pests. This environmentally friendly approach reduces the reliance on synthetic chemicals. Other strategies include Insect Growth Regulators which disrupt the growth and development of insects by targeting juvenile stages, effectively reducing pest populations without harming non-target organisms or humans. However, it remains difficult to assess during or after harvest what is the rate of lentils having an insect.

Currently, the most widely used detection method is based on destructive testing of individual seeds. Each lentil is opened and examined one by one to determine whether a bruchid larva is present. This process is not only time-consuming and costly, but also results in the destruction of the tested seed. Consequently, part of the batch is sacrified, which is an issue that becomes more pressing when the available volume is limited or when the product is expensive.

In this context, developing a non-destructive, automated approach for inspecting lentils is of great interest and such a method would allow infested seeds to be identified within an entire batch without damaging or destroying the samples. The advantages are numerous: enhanced product preservation, improved cost-effectiveness, and a higher valuation for batches confirmed to be insect-free. Moreover, from a commercial point of view, having a reliable and rapid diagnostic tool simplifies traceability and certification, making the product more competitive on the market.

In this article, we will discuss how terahertz (THz) technology can be harnessed to provide this non-destructive solution to detect bruchids in lentils. This approach paves the way for automated, precise, and fast quality control that meets the needs of producers, distributors, and consumers alike.

Current lentils bruchids measurement

The presence of bruches (internal defects) in lentils is currently assessed during production and sales using a destructive testing method. This technique involves selecting a sample of a few dozen lentils from a batch and applying vertical pressure to each lentil, typically using an object like a ballpoint pen. Lentils without bruches resist the applied force, whereas those with defects yield break. By counting the number of broken lentils, the percentage of defective lentils can be estimated, providing an indication of the overall quality of the batch.

However, this method has several drawbacks. It lacks complete objectivity as the pressure applied by the operator can vary. In addition, it is a slow process and is limited to a small sample size, which may not fully reflect the quality of the entire batch.

Terahertz time domain system

Terahertz time domain systems are used in several applications (Valušis et al., 2021; Gao et al., 2024; Shen et al., 2022), including non-destructive testing (Pan et al., 2020), multilayer measurements (Cassar et al., 2019), biomedical (Smolyanskaya et al., 2017) art and heritage (Cassar et al., 2020). Due to its high penetration capability and non-ionising property, terahertz imaging is used for food control with some published examples, like non contact quality parameters evaluation, (Fu et al., 2024) chocolate inspection (Carré et al., 2022; Ok et al., 2019), fungal infections in chestnuts (Di Girolamo et al., 2021) Terahertz imaging non destructive quality control has been demonstrated in several food applications during last decade (Gowen et al., 2012; Wang et al., 2017; Han et al., 2024).

The Teraview TPS 3000 (Cambridge, U.K.) is a terahertz time-domain spectroscopy (THz-TDS), designed primarily for material characterization, spectroscopy and imaging. As illustrated in Figure 1, the system generates ultrafast THz pulses, exploiting their ability to penetrate non-conductive materials and reveal their internal structures without causing damage. The hardware of the TPS 3000 is based on an 80 fs Ti:Sa ultrashort pulsed laser operating at 800 nm with a repetition rate of 76 MHz used as a probe/pump beam, which serves as the source for the THz pulses, that are directed onto a GaAs photoconductive antenna. The THz pulse is generated when the femtosecond optical pump excites carriers in the photoconductive antenna gap. A bias field accelerates these carriers, producing a sub-picosecond transient current that radiates a broadband THz pulse. The TPS 3000 system offers a frequency range of 0.06–3.0 THz (2.0–100.0 cm −1) with a maximum dynamic range of around 75 dB. The lateral resolution of this system, depending on frequency, is approximately 300 μ m (1 THz).

These THz pulses are focused towards the sample using specially designed optics, including parabolic mirrors. The system operates in two primary configurations: reflection mode, where the terahertz radiation reflects off the surface of the sample, and transmission mode, where the radiation passes through the material. This dual configuration enables the TPS 3000 to accommodate a wide variety of materials, from thin coatings to multi-layered composites. Here, in this study, we use it in transmission mode. The detection process uses a complementary photoconductive antenna, which converts the incoming terahertz signal into an electrical response. This detection is time-resolved using a mechanical delay line, meaning that the arrival of the terahertz pulses is measured allowing the system to construct a detailed temporal profile of the interaction between the terahertz wave and the sample. By analyzing the time delay and amplitude of the reflected or transmitted signal, data processing approaches can be used to extracts information about the material, including its thickness, dielectric properties, and internal structure. That measurements were done in a nitrogen (N 2) purged chamber. The N 2 purge reduces water vapor absorption in air, which otherwise introduces strong spectral lines and temporal ringing in THz time-domain measurements.

Sample preparation

In this experiment, we employ a transmission setup. Since the motorized stage movement is vertical, we need a method to secure the lentils in a vertical orientation. As illustrated in Figure 2, the proposed solution involves the use of adhesive tape to hold the lentils, which can then be easily attached to the support structure. This setup allows for quick placement and removal of the lentils while providing a thin and uniform support. The uniformity of the support is particularly beneficial for subsequent data analysis. The lentils: they were collected by our agronomic collaborator (Association Climatologique de la Moyenne-Garonne, France).

Bruchids detection

This measurement was completed in 10 min. A temporal signal is obtained for each pixel of the image, which is first analyzed individually before the global image is constructed. When measuring through a lentil, the signals shows a delayed peak compared to the reference, as expected, due to the additional optical path introduced by the lentil. As shown in the Figure 3, the peak delay varies: it is more pronounced for some lentils and less so for others. We can clearly distinguish the good lentils (Figure 3b) based on the time of flight compared to lentil with a brushid (Figure 3a).

Healthy lentils are denser and contain intact biological material. Infested lentils have an internal cavity left by the bruchid larva. The effective refractive index along the THz path is therefore higher in a healthy lentil than in a hollowed lentil. Higher effective refractive index increases optical path length for the same geometric thickness. This produces a larger delay (longer time of flight) of the transmitted THz pulse relative to air.

Based on this observation, we propose a simple signal processing approach: extracting the peak position from each signal and plotting this position on a 2D graph. This approach generates an image that correlates with the mat. In this time-of-flight approach, each pixel is assigned a value in picoseconds corresponding to the position of the maximum peak of the signal. The result of this analysis is presented in the Figure 4. We can clearly observe the presence of the nine lentils, four of which are intact, as expected from the visible holes identified during the preliminary manual sorting. This method thus enables automatic detection of bruchids, and a second measurement was performed, as presented in 4, where healthy lentils, in red, has a higher time of flight than lentils with bruchids, and repeated as described in Appendix.

For each pixel (x, y), the time of flight contrastis defined as in the Equation 1 as the delay of the main transmitted pulse relative to the reference in air: here t peak is obtained via ∗ a r g m a x within a window centered on the first transmitted pulse, t ref is measured without sample, c is the speed of light in vacuum, and d (x, y) is the local thickness (or a constant d 0 within a single lentil). In practice, n eff provides robust contrast: healthy lentils typically show larger Δ t than infested ones.

Segmentation uses a threshold on ToF with the maximum value, as in the Equation 2 in red:

After measurement and processing, we later confirmed that red lentils are healthy using the traditional method described in introduction.

Study limitations

Several limitations must be applied to this work. First, the lentils used for this initial demonstration had bruchids that had already exited, leaving a hole, which allowed us to manually select the corresponding lentils. However, during the early stages of larval development, when the bruchid is still inside, the signal variation is expected to be smaller. Data processing with such an early stage bruchid would require signal processing other than the direct time-of-flight method.

Furthermore, these tests were conducted on lentils that had been harvested 3 months earlier. Measurements on freshly harvested lentils could alter the response due to differences in moisture content This feasibility does not generalize to early, concealed larvae. A prospective, blinded study with pre-specified thresholds, sample-size calculation, and external validation is required before any quantitative claims (Gong et al., 2022; Kashyap et al., 2022).

This study has demonstrated the feasibility of using terahertz time-domain imaging as a non-destructive method for detecting bruchids in lentils, which is the first demonstration to our knowledge. We successfully identified internal defects caused by bruchid infestation, offering a significant improvement over traditional destructive testing methods. The experimental results confirmed the ability of the terahertz imaging to distinguish between infested and intact lentils, while preserving the structural integrity of the samples. The proposed approach not only addresses the limitations of current practices, but also has the potential to be integrated into automated quality control systems. An application of such integration could be the evaluation of batches of lentils to assess the rate of infested lentils. Furthermore, more prospectively, and subject to having low-cost and high-throughput systems such as radars in electronic technology (Carré et al., 2022), it could be possible to perform classification, in order to increase the grade of batches of lentils that are partly infested.

However, some limitations of the current study must be pointed out. First, the analyzed lentils were 3 months old, and the impact of moisture content on freshly harvested lentils remains unexplored. Furthermore, the method was tested on lentils with advanced bruchid damage (visible holes), and the sensitivity of the system to detect larvae in earlier stages of development requires further investigation. Addressing these challenges will be critical to expand the applicability and robustness of this technology.

Future research should aim to build on the first results of this initial study. A more systematic investigation could include testing lentils of various ages and moisture levels to assess the consistency of THz signal detection under various conditions. It would also be valuable to examine the system’s performance on freshly harvested lentils with concealed bruchids to better understand its sensitivity at different infestation stages, and characterize the deficiencies of bad lentils due to effects other than bruchids. Incorporating advanced data processing techniques, such as machine learning, could enhance the accuracy and speed of bruchid detection, particularly in complex scenarios (Xu et al., 2022). Beyond lentil quality control, this method has broader implications for the agro-food industry. For instance, terahertz imaging could be extended to assess other crops or food products that face similar quality issues due to internal defects or infestations. By further refining this technology, it may contribute to improve food safety standards, optimizing resource use, and reducing waste across the supply chain. In addition, an automated sampling system integrated into grain elevators at cooperatives could rapidly screen incoming organic lentil batches, facilitating quick decision-making about storage suitability or segregation. Such a setup is especially relevant in organic farming contexts, where insecticidal treatments are not permitted, making non-destructive detection methods even more critical for managing infestation risks.

In conclusion, THz-TDI represents a non-destructive approach to quality control in the agricultural sector. With continued research and development, this technology has the potential to revolutionize food quality assessment and support new applications for the growing need of industry for efficient, sustainable, and precise diagnostic tools.