<?xml version="1.0" encoding="UTF-8"?>



<records>

  <record>
    <language>eng</language>
          <publisher>Oriental Scientific Publishing Company</publisher>
        <journalTitle>Biomedical and Pharmacology Journal</journalTitle>
          <issn>0974-6242</issn>
            <publicationDate>2026-11-06</publicationDate>
    
        <volume>19</volume>
        <issue>4</issue>

 
    <startPage></startPage>
    <endPage></endPage>

	    <publisherRecordId>74059</publisherRecordId>
    <documentType>article</documentType>
    <title language="eng">Edge-Optimized Explainable Graph Neural Learning for Lumpy Skin Disease Classification</title>

    <authors>
	 


      <author>
       <name>Sandeep Sharma</name>

 
		
	<affiliationId>1</affiliationId>
      </author>
    

	 


      <author>
       <name>Amandeep Kaur</name>


		
	<affiliationId>2</affiliationId>

      </author>
    

	 


      <author>
       <name>Suman Pant</name>

		
	<affiliationId>3</affiliationId>
      </author>
    

	 


      <author>
       <name>Ruhi Saxena</name>

		
	<affiliationId>4</affiliationId>
      </author>
    


	 


      <author>
       <name>Gopal Ghosh</name>

		
	<affiliationId>5</affiliationId>
      </author>
    


	 


      <author>
       <name>Ashish Rayal</name>

		
	<affiliationId>6</affiliationId>
      </author>
    
    </authors>
    
	    <affiliationsList>
	    
		
		<affiliationName affiliationId="1">School of AI and Emerging Technologies, Lovely Professional University, Phagwara Punjab, India</affiliationName>
    

		
		<affiliationName affiliationId="2">Computer Science and Engineering, M.M. Engineering College Maharishi Markandeshwar (Deemed to be University), Mullana, Ambala, Haryana, India.</affiliationName>
    
		
		<affiliationName affiliationId="3">School of Science and Technology, Swami Rama Himalayan University Jolly Grant, Doiwala, Dehradun Uttarakhand, India.</affiliationName>
    
		
		<affiliationName affiliationId="4">School of Engineering and Technology CGC University, Mohali Punjab India.</affiliationName>
    
		
		<affiliationName affiliationId="5">School of AI and Emerging Technologies, Lovely Professional University, Phagwara Punjab, India.</affiliationName>
    
		
		<affiliationName affiliationId="6">Department of Mathematics, School of Applied and Life Sciences, Uttaranchal University Dehradun, Uttarakhand India.</affiliationName>
    
	  </affiliationsList>






    <abstract language="eng">Lumpy Skin Disease (LSD) is a highly contagious viral disease that has a major impact on cow health and the livestock industry worldwide. The accurate, interpretable and real-time prediction of LSD is a challenging problem since traditional machine learning methods are often not easily explainable, are limited in cross domain adaptability and are hard to deploy at the edge. To address these issues, we present a novel framework in this work that combines Graph Convolutional Networks (GCN), Layer-wise Relevance Propagation (LRP), Domain Adversarial Neural Networks (DANN), and MobileNetV3-GCN with TinyML into a single system. The proposed framework includes an embedded model with a set of heterogeneous data sources, such as data obtained from Internet of Things (IoT) sensors and from satellite images, to model complex epidemiological relationships, and to achieve better transparency of the model and its generalization across different geographical areas. The results were reproducible by performing experimental evaluations with standard preprocessing, fixed random seeds, documented hyperparameters and consistent partitioning of the datasets. The proposed framework outperformed the baseline Convolutional Neural Network (CNN), Graph Attention Network (GAT) and standard GCN models, with a cross domain prediction accuracy of 95%. The use of TinyML-based quantization and pruning in the model reduced inference latency by 70-80%, making them suitable to be deployed to the resource-constrained edge devices, whereas LRP increased the relevance precision by 85% to make the outcomes of the latter more interpretable. The proposed framework offers a robust solution for early LSD prediction that is accurate, scalable, and explainable, paving the way for timely disease monitoring, informed decision-making, and practical application in smart livestock management systems.</abstract>

    <fullTextUrl format="html">https://biomedpharmajournal.org/vol19no4/edge-optimized-explainable-graph-neural-learning-for-lumpy-skin-disease-classification/</fullTextUrl>

<keywords language="eng">

      
        <keyword>Convolutional Neural Network</keyword>
      

      
        <keyword> Domain Adversarial Neural Network</keyword>
      

      
        <keyword> Gradient Reversal Layer</keyword>
      

      
        <keyword> Graph Attention Network</keyword>
      

      
        <keyword> Graph Convolutional Network</keyword>
      

      
        <keyword> Layer-wise Relevance Propagation</keyword>
      

      
        <keyword> Long Short-Term Memory</keyword>
      

      
        <keyword> Lumpy Skin Disease.</keyword>
      
</keywords>
  </record>
</records>