PAWPAL: AI-Based Pet Health Assistance System Using Image-based Disease Detection and NLP Symptom Analysis

Abstract

Early identification of health problems in pets is generally hindered by the unavailability of immediate veterinary services and a low level of awareness among pet owners. Presenting here is PAWPAL, a pet health support system powered by AI, which combines image-based disease classification and a natural language-based symptom analyzer. The system makes use of Convolutional Neural Networks (CNN) to diagnose common skin diseases in dogs and cats from the images provided by the users. Moreover, the text-based symptom checker figures out the condition\'s severity and suggests care instructions. The frontend was built with React (Next.js), and the backend was developed in Flask, the integration of TensorFlow models and an AI reasoning module were done.The test performance results showed that the model for diagnosing dog skin diseases was 93% accurate, while the model for diagnosing cat skin diseases 80% accurate. The system offers help to users in real-time, thus the delays in diagnosis are lessened and the accessibility to the first veterinary guidance is improved.

Introduction

The rapid growth in pet ownership has increased the demand for accessible and intelligent veterinary support, particularly in areas with limited access to clinics. PawPal addresses this gap through a dual AI-based system that combines image-based disease detection and symptom-based text analysis to assist in the early diagnosis of dermatological conditions in pets. By integrating computer vision (CNNs) and natural language processing (NLP), PawPal provides health insights, severity assessment, care recommendations, and guidance on whether urgent veterinary attention is required, making veterinary assistance more accessible to non-technical pet owners.

The literature highlights the effectiveness of CNNs in medical image classification and the growing role of NLP in interpreting user-described symptoms. Transfer learning using pretrained models such as MobileNetV2 helps overcome limited veterinary datasets, enabling accurate detection of common skin conditions in dogs and cats. Compared to rule-based systems, modern context-aware NLP models offer better reasoning and flexibility when interpreting informal symptom descriptions from pet owners.

Methodologically, PawPal employs separate MobileNetV2 CNN models for dogs and cats, trained on curated dermatology datasets with data augmentation to improve robustness. The image-based system predicts disease categories and generates actionable veterinary instructions, while the symptom checker uses NLP to identify key symptoms, assess severity, and recommend care or emergency action. The system is optimized for lightweight deployment, making it suitable for mobile and low-resource environments.

Results show strong performance for the dog skin disease model, achieving approximately 88–89% accuracy, while the cat model reached moderate accuracy (around 65% test accuracy) due to smaller datasets. The NLP-based symptom checker effectively assessed severity and probable conditions, though its accuracy depended on the completeness of user descriptions.

Overall, PawPal demonstrates the potential of combining vision and language AI to bridge the gap between professional veterinary care and home-based monitoring. Future enhancements include expanding and diversifying datasets, integrating real-time tele-veterinary consultations, improving NLP models with domain-specific training, enabling offline mobile use, incorporating wearable-based behavior monitoring, and extending the platform to additional animal species. These improvements aim to make PawPal a comprehensive, reliable, and widely accessible AI-assisted veterinary decision support system.

Conclusion

The PAWPAL system is a fine example of how AI integration can effectively elevate the early detection of diseases and the making of health-related decisions for pets to a large extent. By design the system merges CNNs for image-based skin disease classification with the NLP-based Symptom Checker, which, by definition, addresses two major issues that pet owners encounter - misinterpretation of the visual symptoms and lack of vet consultation. Besides classification reports and confusion matrices, the paper also presents the experimental results to demonstrate that the proposed disease prediction model has attained high diagnostic accuracy for dogs and good performance for cats. Moreover, the symptom analysis unit is instrumental in further system enhancement by determining symptoms\' severity, suggesting condition indicators, and recommending first-aid measures. The whole system is mostly aimed at pet owners, is quite affordable, and has a huge potential for further development, thus, it helps pet owners by providing them with simple and clear directions and giving them the opportunity to carry out interventions at an early stage, which, therefore, can considerably reduce the risk of disease progression. It is important to mention that PAWPAL is not a replacement for a professional veterinary consultation; nevertheless, to a large extent, it serves as a means of instant support, thus, facilitating the identification of the necessity of medical care. Later, the updates will entail broadening the datasets, enhancing diagnostic accuracy for various breeds, allowing live veterinary consultation, and adding more species of pets. As a result, PAWPAL is one of the essential digitally connected components of the veterinary health care system and a vivid illustration of how AI can bring real-life conveniences to animal welfare.

References

[1] Zhang, L., Wang, M., Liu, Y., Chen, H., and Kim, J., “Deep Learning for Automated Veterinary Diagnosis: An Exhaustive Study of CNN Implementation in Pet Health Evaluation,” Journal of Veterinary and Animal Science, vol. 12, no. 3, pp. 100–115, 2021. [2] Kim, S. H., Park, J. W., Lee, M. K., and Thompson, R., “Convolutional Neural Networks to Identify Canine Skin Disorders: A Multi-Breed Analysis,” Computers in Biology and Medicine, vol. 118, p. 103112, 2020. [3] Thompson, R. J., Martinez, A. C., Davis, K. L., and Wilson, P. M., “Automated Detection of Eye Diseases in Companion Animals Using Deep Learning: Clinical Validation Study,” Veterinary Ophthalmology, vol. 24, no. 3, pp. 201–210, 2021. [4] Laranjo, L., Dunn, A. G., Tong, H. L., Kocaballi, A. B., Chen, J., Bashir, R., and Coiera, E., “Conversational AI in Healthcare: A Comprehensive Review of Application and Challenges,” NPJ Digital Medicine, vol. 5, p. 45, 2022. [5] Rodriguez, A., Patel, S., Kumar, V., and Anderson, M., “Machine Learning in Veterinary Medicine: A Systematic Review and Meta-Analysis,” Animals (MDPI), vol. 12, no. 13, p. 1687, 2022. [6] Patel, N., Singh, R., Kumar, A., and Sharma, D., “Color-Based Feature Extraction for Skin Lesion Classification in Veterinary Dermatology,” Biomedical Signal Processing and Control, vol. 62, p. 102115, 2020. [7] Williams, K. L., Brown, S. A., Johnson, T. M., and Lee, C. H., “Digital Health Technologies in Veterinary Medicine: Present and Future Scenarios,” Frontiers in Veterinary Science, vol. 9, p. 892905, 2022. [8] Chen, X., Liu, H., Wang, S., and Zhang, Y., “Computer Vision Applications in Animal Health Monitoring: A Review of Current Technologies and Future Prospects,” Computers and Electronics in Agriculture, vol. 187, p. 106118, 2021. [9] Mandrioli, L., Siracusa, F., and Cantile, C., “Veterinary Practice Telemedicine: Current Status and Future Direction,” Animals (MDPI), vol. 11, no. 7, p. 1988, 2021.

Copyright

Copyright © 2025 Aishwarya AP, Ananya Rao R, Shubha C, Varsha V, Malathi P. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.