Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water

arXiv:2611.16859 · eess.IV, eess.SP · Submitted 2026-11-12 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.

Rosa: We're looking at the paper: "Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water".

Dev: Here's the gist of the paper. Transmission-mode multispectral imaging (MSI) is proposed as a framework to estimate both Microcystis spp.

Rosa: cell concentration and ELISA-equivalent microcystin concentration in water through culture derived dilution samples, demonstrating the feasibility of this dual output optical calibration framework The study evaluates a custom-built MSI system acquiring twelve spectral bands from near-ultraviolet to near-infrared wavelengths to achieve these estimations Dual Target Estimation Feasibility The research demonstrates the feasibility of using transmission mode MSI to estimate both Microcystis spp.

Dev: cell concentration and ELISA-equivalent MC concentration within a controlled culture derived dilution series This dual target formulation avoids explicitly using one target as a proxy for the other and provides two complementary outputs from a single optical acquisition Both calibration relationships showed comparable performance on held-out intermediate concentration levels, indicating stability when applied to intermediate concentration levels not used during model fitting Spectral Response and Data Preparation The relationship between the aggregated Wasserstein distance descriptor and Microcystis spp.

Rosa: cell concentration was modeled using the linear regression equation Cˆcell = −0.1322 + 110.8299W The decrease in normalized transmitted intensity with increasing cell concentration is consistent with increasing optical attenuation within the cyanobacterial suspension, arising from the combined effects of pigment related absorption and light scattering Spectral signatures were plotted to provide an initial visualization of the spectral behavior associated with the prepared concentration levels Model Construction Pipeline The framework combines twelve band MSI acquisition, spectral preprocessing, principal component analysis (PCA) based dimensionality reduction, and an aggregated Wasserstein distance descriptor The analysis first involved image preprocessing, including dark current subtraction, region of interest (ROI) selection, and bilateral filtering to reduce acquisition related noise and standardize the spectral images prior to feature extraction Subsequently, square root and Yeo-Johnson transformations were applied to the normalized calibration and held-out validation matrices for feature transformation for modeling PCA was fitted using only the transformed calibration matrix to obtain a compact representation of the multispectral information Finally, a single scalar descriptor, Wi, was constructed by combining four component wise Wasserstein distances using an unweighted sum Performance Metrics and Interpretation The calibration relationship for cell concentration achieved a coefficient of determination, R2 = 0.9819, with a root mean square error (RMSE) of 8.3034×105 cells mL−1 The ELISA-equivalent MC estimation showed an R2 of 0.9788 for calibration and R2 = 0.9786 on held-out intermediate dilution levels The results demonstrate the feasibility of transmission-mode MSI as a dual output optical calibration framework for estimation of Microcystis spp.

Title and authors: Rosa: So we're looking at this paper titled "Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water" and the authors are Jayakody, Thilakarathne, Ranasinghe, De Silva, Perera, Godaliyadda, Herath, Ekanayake Athukorala and Sirimanna.

Dev: That title sounds like a lot to digest for a regular water quality person. It’s about using a specific imaging technique to guess two different things about Microcystis in the water: how many cells there are and how much microcystin toxin is present, even though they aren't directly the same thing.

Rosa: Exactly. The core idea here is this dual output optical calibration framework, which tries to estimate both cell concentration and ELISA-equivalent microcystin concentration from culture derived dilution samples using transmission mode multispectral imaging.

Dev: So it’s not just trying to measure one thing, like just the cells or just the toxin; it’s aiming for two complementary outputs from a single optical acquisition process.

Taro: From my angle, I'm thinking about why that complementarity is important—because cell abundance and toxin concentration don't always correlate well on their own, which means relying only on one indicator can be incomplete information for risk assessment.

Rosa: Right. And what the paper shows is that they built a custom system acquiring twelve spectral bands from near-ultraviolet to near-infrared wavelengths measured transmitted light spectral responses. This is how they get the data points for their estimation process, and it’s pretty detailed about that setup in page two of this paper <ref:2611.16859#pg1>.

Dev: Twelve bands spanning from ultraviolet to infrared sounds like a lot of data coming in real time or near real time, so I'm wondering about the practical limits there. How stable is this system when the water conditions actually change outside of a perfect lab setting?

Rosa: That’s what I want to know. The paper addresses that by showing how they used principal component analysis based dimensionality reduction followed by an aggregated Wasserstein distance descriptor for regression based prediction, which helps quantify the spectral variation.

Taro: So, it’s taking all that raw spectral variation and squeezing it down into one single scalar descriptor, which is that Wasserstein distance descriptor, to feed into their final regression model? That sounds like a way to make sense of a huge amount of spectral information efficiently.

Dev: It’s an attempt at dimensionality reduction before the final modeling step. I'm looking at the results now—they report R squared values of zero point nine eight one nine for cell concentration calibration and zero point nine seven eight eight for ELISA-equivalent microcystin concentration calibration, and also on held-out intermediate dilution levels <ref:2611.16859#pg1>.

Rosa: Those numbers are pretty high, especially the R squared of zero point nine eight one nine for the cell concentration estimation when calibrated with their data <ref:2611.16859#pg1>. The paper also shows that the relationship between aggregated Wasserstein distance descriptor and Microcystis spp. cell concentration is modeled by a linear regression equation where C hat cell equals negative zero point one three two two plus one hundred ten point eight two nine nine W.

Title and authors: Taro: That specific linear relationship, using that one scalar descriptor Wi, means they’ve established a very direct mathematical link between their spectral features and the actual cell count. But what about the error in that link?

Dev: They give you the root mean square error for that cell concentration estimation as eight point three zero three four times ten to five hundred cells per milliliter when calibrated, which is a concrete number we can look at. And for the toxin estimation, they report an R squared of zero point nine seven eight eight for calibration and zero point nine seven eight six on held-out intermediate dilution levels for ELISA-equivalent MC concentration, which is also quite strong performance <ref:2611.16859#pg1>.

Rosa: They show that the calibration relationship remained stable when applied to intermediate concentration levels that weren't used during the initial model fitting, which suggests the model has some good generalization capability across different concentrations.

Taro: That stability is key for real-world deployment because you don't want a system calibrated on low concentrations suddenly failing when it sees a high bloom. But what about what they didn't test? I’m looking at page three and they mention that cell counts or pigment based indicators alone may provide an incomplete basis for toxin risk assessment <ref:2611.16859#pg1>.

Dev: They acknowledge that the cell count doesn't necessarily correlate with microcystin concentration, which is a crucial caveat. The paper points out that MC cell quota and congener composition can change based on things like nutrient availability, temperature, light exposure, and competition with other phytoplankton.

Rosa: So the framework they propose isn't a magic solution that instantly tells you everything about the toxin; it’s presented as one part of a larger approach to rapid, sensitive, and practical assessment of both bloom abundance and associated cyanotoxin concentrations.

Taro: It sounds like this MSI framework is valuable because it provides complementary information on the chemical hazard alongside the bloom magnitude. It gives you two pieces of data at once for water quality management.

Dev: From an engineering standpoint, the entire pipeline—from image preprocessing, including dark current subtraction and region of interest selection, through to feature transformation and PCA based dimensionality reduction—is what makes this work happen in a controlled environment before we worry about the field.

Rosa: And they have made all that code publicly available on GitHub so others can see how the entire analysis pipeline works, which is good for transparency when you're trying to build something useful outside of just a single study.

Taro: If you think about this for someone monitoring a real bloom, does this system offer them anything practical beyond just having two R squared numbers? Does it help them make a faster decision?

Dev: It supports rapid indirect estimation of water quality parameters by combining compact LED based spectrophotometric measurements with data driven calibration, which is the broader implication they are suggesting. This lets you get an estimate without needing a huge, complex laboratory assay immediately.

Title and authors: Rosa: So to wrap up this paper on Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water, they demonstrate the feasibility of using transmission mode MSI as a dual output optical calibration framework for estimation of Microcystis spp. cell concentration and ELISA-equivalent MC concentration in water.

Dev: Yeah, it shows that this combined approach is feasible when you use twelve spectral bands from near-ultraviolet to near-infrared wavelengths measured transmitted light spectral responses.

Taro: It’s about proving that having a method to jointly estimate both the cell abundance and the associated toxin concentration using a single optical measurement technique is achievable with good accuracy.

Rosa: The main implication for me is how this moves us toward more rapid, sensitive, and practical assessment methods for water quality management by providing this end-to-end framework.

Dev: For real-time loop engineers, the implication is that you can potentially use these compact LED based spectrophotometric measurements combined with data driven calibration to support quick indirect estimation of water quality parameters.

Taro: I think the big picture here is establishing a robust method for joint optical assessment of Microcystis spp. abundance and ELISA-equivalent MC concentration.

Rosa: We’ve looked at the title, we’ve discussed the dual outputs, and we’ve seen how they achieved those high R squared values for both targets using that specific spectral descriptor technique.

Dev: The caveat remains that this is currently based on culture derived dilution samples, so the next step in testing needs to be moving this out of the lab setting and into actual bloom samples in the field.

Taro: That’s exactly where I’d push for more work—establishing the robustness and transferability of this approach for field oriented monitoring outside of controlled conditions.

Rosa: So, we've covered how they built a system with twelve bands, how they used PCA and Wasserstein distance to create that scalar descriptor Wi, and what the resulting performance metrics look like for both cell concentration and microcystin estimation.

Dev: The next thing is seeing if this method holds up when you apply it to samples collected from natural blooms in the field, which is the logical next step beyond culture derived dilution series.

Taro: And we should also check how gracefully this system recovers or responds when the world misbehaves and things get noisy out there.

Rosa: Overall, this paper shows that transmission mode MSI can be used as a dual output optical calibration framework for estimation of Microcystis spp. cell concentration and ELISA-equivalent MC concentration in water, providing complementary information on the associated chemical hazard alongside the bloom magnitude.

Dev: It’s a solid demonstration of how data driven calibration can support rapid indirect estimation of water quality parameters by combining compact LED based spectrophotometric measurements with that MSI data.

Taro: We've seen how they constructed this entire pipeline, and it seems like the most important part is proving that you can get both targets reliably from one measurement process.

Rosa: And we’re wrapping up on this paper about Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water.

The paper's summary: Rosa: So we're looking at this paper titled "Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water." It proposes using transmission mode multispectral imaging to estimate two things: how many Microcystis cells there are, and the concentration of microcystin toxin, using samples from controlled culture dilutions.

Dev: Right, so it's not just trying to measure one thing; they're aiming for a dual output optical calibration framework that gives you both cell concentration and ELISA-equivalent microcystin concentration from a single measurement setup.

Taro: That’s interesting because cell abundance and toxin concentration don't always move together, so having two complementary outputs avoids you just using one as a proxy for the other.

Rosa: Exactly. They built this custom system acquiring twelve spectral bands, from near-ultraviolet to near-infrared wavelengths, measuring the light that gets transmitted through the water samples.

Dev: Twelve bands spanning that range sounds like a lot of spectral data coming in at once, so I'm wondering about how stable this system is when you try to use it outside of a perfectly controlled lab setting.

Taro: That stability is key for real-world monitoring, because we need to know what happens when the actual water conditions change and things get noisy.

Rosa: The paper addresses that by showing they used principal component analysis based dimensionality reduction, followed by an aggregated Wasserstein distance descriptor to model how the spectral variation relates to cell concentration.

Dev: So they take all that raw spectral variation and squash it down into one single scalar descriptor, which is this Wasserstein distance value, and then use that for their final regression model.

Taro: That's a way of making sense of a huge amount of spectral information efficiently before feeding it into the final prediction step.

Rosa: They report some performance numbers: an R squared value of zero point nine eight one nine for the cell concentration calibration, and zero point nine seven eight eight for the ELISA-equivalent microcystin concentration calibration.

Dev: Those are pretty high numbers, especially that zero point nine eight one nine for the cell concentration estimation when they calibrated it with their data. And they also showed good performance on those intermediate dilution levels that weren't used in the initial model fitting.

Taro: The stability of those results across different concentration levels suggests the fitted relationships have some good generalization capability when applied to different concentrations.

Rosa: They show that the calibration relationship for cell concentration is modeled by a linear regression equation where cell concentration hat equals negative zero point one three two two plus one hundred ten point eight two nine nine times that Wasserstein distance value.

Dev: That specific linear relationship using that single scalar descriptor, which we called Wi, means they’ve established a very direct mathematical link between their spectral features and the actual cell count. But what about the error in that link?

Taro: They give you the root mean square error for that cell concentration estimation as eight point three zero three four times ten to five hundred cells per milliliter when calibrated. That's a concrete number we can look at.

Rosa: And for the toxin estimation, they report an R squared of zero point nine seven eight eight for calibration and zero point nine seven eight six on held-out intermediate dilution levels for ELISA-equivalent microcystin concentration.

Dev: So it shows the feasibility of using transmission mode multispectral imaging as a dual output optical calibration framework to estimate both Microcystis spp. cell concentration and ELISA-equivalent MC concentration in water.

Taro: The main implication for me is how this moves us toward more rapid, sensitive, and practical assessment methods for water quality management by providing this end-to-end framework.

Rosa: Yeah, it shows that combining compact LED based spectrophotometric measurements with data driven calibration can support rapid indirect estimation of water quality parameters.

Dev: So the next thing is seeing if this method holds up when you apply it to samples collected from natural blooms in the field, which is the logical next step beyond culture derived dilution series.

The paper's improvements: Rosa: So we're talking about how they improve this transmission mode multispectral imaging method for estimating microcystis cells and toxin concentration using their results from culture derived dilution samples.

Dev: Right, what are these improvements? What's the next step in making this framework better than what they just showed us?

Taro: They suggest moving beyond just the initial calibration to test how gracefully the system recovers when things get noisy or when you move it out of that perfect lab setting.

Rosa: They’re suggesting that further work should evaluate this entire framework using independently collected cultures and natural bloom samples, which is crucial for establishing its robustness and transferability for field oriented monitoring.

Dev: So they aren't just stopping at the controlled dilution series; they want to see if this method actually works when you apply it to real water quality data from the field.

Taro: That makes sense because we need to know if this approach can handle the unpredictable variability of a natural bloom, where things like nutrient availability and light exposure keep changing.

Rosa: They also highlighted that the complete analysis pipeline is important, meaning they emphasize that everything from controlled illumination and wavelength selection to the preprocessing and PCA based dimensionality reduction matters for quantitative estimation.

Dev: So it’s not just about getting a good final R squared number; it’s about validating every single step of this whole process so you can trust the estimate in a real-time loop scenario.

Taro: I'm interested in how they handle the uncertainty, because when you get estimates for both cells and toxin, there has to be some way to quantify how much confidence you should have in each one.

Rosa: They pointed out that since cell counts and microcystin concentration don't perfectly match up on their own, the dual output is still valuable because it provides complementary information about the chemical hazard alongside the bloom magnitude.

Dev: So, while they can't perfectly predict both at once from this single measurement, having both pieces of data helps you build a more complete picture for water quality management decisions.

Taro: It’s less about a perfect prediction and more about providing two distinct indicators that work together to give you better risk assessment information.

Rosa: And they also mentioned the potential for using these compact LED based spectrophotometric measurements combined with data driven calibration to support rapid indirect estimation of water quality parameters, which is a pretty practical application.

Dev: That's a good point because it means you don't need a massive, complex laboratory assay right away; you can get an estimate quickly.

Taro: So the big picture here is that this framework gives us a way to jointly assess both the abundance of microcystis and its associated toxic risk using just optical data.

Rosa: Yeah, it's about establishing a robust method for joint optical assessment of microcystis spp. abundance and ELISA-equivalent MC concentration.

Dev: And they’ve made the code for all that preprocessing, PCA, and modeling publicly available on GitHub so others can see exactly how the entire analysis pipeline works step by step.

Conclusion: Rosa: So to wrap up this paper, "Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water," they show that using transmission mode multispectral imaging can create a dual output optical calibration framework for estimating both microcystis cell concentration and microcystin concentration in water.

Dev: Right, so the main point is that you can get two different types of information—cell count and toxin level—from one single optical measurement process.

Taro: It proves that joint optical assessment of bloom abundance and associated cyanotoxin concentrations is feasible using this kind of setup.

Rosa: They built a system with twelve spectral bands from near-ultraviolet to near-infrared wavelengths measured transmitted light spectral responses, which they used for these estimations.

Dev: That full analysis pipeline—from image preprocessing to PCA based dimensionality reduction and the Wasserstein distance descriptor—is what makes this whole process work in a controlled environment before we worry about the field.

Taro: It’s important because it shows how to get both targets reliably from one measurement process, which is a big step for monitoring.

Rosa: They also made all that code available on GitHub so others can see exactly how the analysis pipeline works step by step, which is good for transparency.

Dev: And they showed that the calibration relationship remained stable when applied to intermediate concentration levels not used during model fitting, which means the model has some good generalization capability.

Taro: That stability is important because you don't want a system calibrated on low concentrations suddenly failing when it sees a high bloom.

Rosa: So overall, this paper shows that this method is feasible for quantitative estimation from transmission mode MSI data.

Dev: It demonstrates how data driven calibration can support rapid indirect estimation of water quality parameters by combining compact LED based spectrophotometric measurements with that MSI data.

Taro: The next logical step for me is seeing if this method holds up when you apply it to samples collected from natural blooms in the field, testing its robustness and transferability outside of those controlled conditions.

Rosa: Yeah, we need to see how gracefully it recovers when the world misbehaves out there.

Dev: That's a necessary test because real-world conditions are never as perfect as a lab setting.

Taro: It really highlights how this framework provides complementary information on the chemical hazard alongside the bloom magnitude.

Rosa: So, to wrap up this paper, "Transmission-Mode Multispectral Imaging for Estimation of Microcystis spp. Cell Concentration and ELISA-Equivalent Microcystin Concentration in Water," they demonstrate that transmission mode MSI can be used as a dual output optical calibration framework for estimation of microcystis spp. cell concentration and ELISA-equivalent MC concentration in water.

Dev: It’s a solid demonstration of how to use twelve spectral bands from near-ultraviolet to near-infrared wavelengths measured transmitted light spectral responses for these estimations.

Taro: It's about proving that having a method to jointly estimate both the cell abundance and the associated toxin concentration using a single measurement technique is achievable with good accuracy.

J.A.S.T. Jayakody, G. Thilakarathne, G.A.S.L. Ranasinghe, M.C.-L.-De Silva, K.-A.-D.-S.-A. Perera, G.M.-R.ÍI.­dyadda, H.M.-V.R.Ч Herath, M.P.-B.▱ Ekanayake, A.D.-S.-N.P.▱ Athukorala, H.A.-D.-R.P.▱ Sirimanna

Department of Electrical and Electronic Engineering, University of Peradeniya · Farbe Technologies (Pvt) Ltd · Department of Electrical and Information Engineering, University of Ruhuna · Multidisciplinary AI Research Centre, University of Peradeniya · Department of Botany, Faculty of Science, University of Peradeniya

eess.IV, eess.SP

Submitted: 2026-11-12

Updated: 2026-11-12

Comments: 30 pages, 9 figures, 2 tables. Includes Supplementary Information

Code: https://github.com/SenithJayakody/microcystis-microcystin-msi

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 86/100

The gist: Transmission-mode multispectral imaging (MSI) is proposed as a framework to estimate both Microcystis spp.

Key concepts

Transmission-Mode MSI
This involves taking images of water where light passes through the sample. The system captures twelve different colors (spectral bands) from near-ultraviolet to near-infrared wavelengths. By measuring how much light is transmitted, the system can infer properties of the organisms inside, like cell concentration and toxin levels.
Dual Target Estimation
The framework estimates two things at once: Microcystis cell count and microcystin concentration. This avoids needing one measurement to approximate the other. Both estimations are derived from the same spectral data, offering complementary information about the water quality simultaneously.
Wasserstein Distance Descriptor
This is a mathematical tool used to measure how different two sets of data distributions are. In this study, it quantifies the dissimilarity between spectral features derived from different concentration levels of Microcystis cells and microcystin concentrations, helping build the predictive model.
Calibration Framework
This is the process of creating a mathematical relationship between the raw MSI data and the actual concentrations. The researchers used this to create models that accurately predict cell counts (R2=0.9819) and toxin levels, showing how reliable this optical method is for quantitative estimation.

Terminology

Summary

Transmission-mode multispectral imaging (MSI) is proposed as a framework to estimate both Microcystis spp. p. 1 cell concentration and ELISA-equivalent microcystin concentration in water through culture derived dilution samples, demonstrating the feasibility of this dual output optical calibration framework The study evaluates a custom-built MSI system acquiring twelve spectral bands from near-ultraviolet to near-infrared wavelengths to achieve these estimations

Dual Target Estimation Feasibility

The research demonstrates the feasibility of using transmission mode MSI to estimate both Microcystis spp. cell concentration and ELISA-equivalent MC concentration within a controlled culture derived dilution series This dual target formulation avoids explicitly using one target as a proxy for the other and provides two complementary outputs from a single optical acquisition Both calibration relationships showed comparable performance on held-out intermediate concentration levels, indicating stability when applied to intermediate concentration levels not used during model fitting

Spectral Response and Data Preparation

The relationship between the aggregated Wasserstein distance descriptor and Microcystis spp. cell concentration was modeled using the linear regression equation Cˆcell = −0. p. 41322 + 110. p. 48299W The decrease in normalized transmitted intensity with increasing cell concentration is consistent with increasing optical attenuation within the cyanobacterial suspension, arising from the combined effects of pigment related absorption and light scattering Spectral signatures were plotted to provide an initial visualization of the spectral behavior associated with the prepared concentration levels

Model Construction Pipeline

The framework combines twelve band MSI acquisition, spectral preprocessing, principal component analysis (PCA) based dimensionality reduction, and an aggregated Wasserstein distance descriptor The analysis first involved image preprocessing, including dark current subtraction, region of interest (ROI) selection, and bilateral filtering to reduce acquisition related noise and standardize the spectral images prior to feature extraction Subsequently, square root and Yeo-Johnson transformations were applied to the normalized calibration and held-out validation matrices for feature transformation for modeling PCA was fitted using only the transformed calibration matrix to obtain a compact representation of the multispectral information Finally, a single scalar descriptor, Wi, was constructed by combining four component wise Wasserstein distances using an unweighted sum

Performance Metrics and Interpretation

The calibration relationship for cell concentration achieved a coefficient of determination, R2 = 0.9819, with a root mean square error (RMSE) of 8. p. 43034×105 cells mL−1 The ELISA-equivalent MC estimation showed an R2 of 0. p. 49788 for calibration and R2 = 0. p. 19786 on held-out intermediate dilution levels The results demonstrate the feasibility of transmission-mode MSI as a dual output optical calibration framework for estimation of Microcystis spp. p. 1 cell concentration and ELISA-equivalent MC concentration in water The close agreement between the calibration and held-out results for both targets indicates that the fitted relationships remained stable when applied to intermediate concentration levels not used during model fitting

Broader Implications

The proposed methodology provides an end-to-end framework for joint optical assessment of Microcystis spp. abundance and ELISA-equivalent MC concentration The results highlight the contribution of the complete analysis pipeline, including controlled illumination, selected wavelength acquisition, consistent imaging conditions, preprocessing, feature transformation, PCA based dimensionality reduction, aggregated Wasserstein distance descriptor construction, and regression modeling to quantitative estimation from transmission mode MSI data Further work should evaluate the framework using independently collected cultures and natural bloom samples to establish the robustness and transferability of this approach for field oriented monitoring The results therefore demonstrate the feasibility of using the MSI derived descriptor for ELISA-equivalent MC estimation The compact LED based spectrophotometric measurements combined with data driven calibration can support rapid indirect estimation of water quality parameters The proposed methodology therefore provides an end-to-end framework for joint optical assessment of Microcystis spp. p. 8 abundance and ELISA-equivalent MC concentration The results therefore demonstrate the feasibility of using the MSI derived descriptor for ELISA-equivalent MC estimation The results therefore demonstrate the feasibility of using transmission mode MSI as a dual output optical calibration framework for estimation of Microcystis spp. p. 1 cell concentration and ELISA-equivalent MC concentration in water The results therefore demonstrate the feasibility of using transmission mode MSI as a dual output optical calibration framework for estimation of Microcystis spp. p. 1 cell concentration and ELISA-equivalent MC concentration in water

Reference Data Summary

The reference cell concentrations and ELISA-equivalent MC concentrations for the calibration and held-out validation datasets are summarized in Supplementary Table S1 The level-specific reference values were assigned to the eight corresponding MSI imaging replicates at each dilution level The 88 calibration descriptors were associated with 11 distinct level wise reference concentrations, while the 48 held-out descriptors were associated with six distinct intermediate concentration levels The code used for multispectral image preprocessing, feature transformation, principal component analysis, Wasserstein distance descriptor construction, regression modeling, and performance evaluation is publicly available at https://github.com/SenithJayakody/microcystis-microcystin-msi The code used for multispectral image preprocessing, feature transformation, principal component analysis, Wasserstein distance descriptor construction, regression modeling, and performance evaluation is publicly available at https://github. p. 8com/SenithJayakody/microcystis-microcystin-msi

Mathematical Formulation of Distance

The 1-Wasserstein distance, W1, is used to quantify the dissimilarity between two probability distributions in terms of the minimum cost required to transport probability mass from one distribution to another The general mathematical formulation of the p-Wasserstein distance is defined as Wp(µ, ν) = inf γ∈Π(µ,ν) Z Rd×Rd∥x − y∥ p 1/p In this study, p = 1 was used, yielding the 1-Wasserstein distance, W1, which is commonly referred to as the Earth Mover’s Distance The component wise Wasserstein distance was calculated as w(k)i = W1 Z(k)ref, Z(k)i, k = 1...

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Precision estimation of Microcystis spp. cell concentration can be achieved by using a custom built MSI system acquiring twelve spectral bands from near-ultraviolet to near-infrared wavelengths measured transmitted light spectral responses. This allows for an estimation of Microcystis spp. cell concentration and ELISAequivalent MC concentration from culture derived dilution samples, with calibration yielding R2 values as high as 0.9819 for cell concentration.

  2. Dual output optical calibration frameworks can be established to simultaneously estimate both Microcystis spp. cell concentration and ELISAequivalent MC concentration in water, providing complementary information that avoids explicitly using one target as a proxy for the other.

  3. The system can perform rapid indirect estimation of water quality parameters by combining compact LED based spectrophotometric measurements with data driven calibration, illustrating the broader applicability of reference calibrated optical measurements for water quality characterization.

Related papers