Skin Disease Detection Using Image Processing and CNN

Abstract

The rising prevalence of skin problems, combined with a dearth of dermatologists—particularly in rural and underserved areas—has generated a pressing demand for novel technologies that aid in early detection and diagnosis. To address this, we introduce an AI-powered solution that helps patients, general practitioners, and dermatologists by providing an efficient and user-friendly preliminary screening tool. Our method uses Convolutional Neural Networks (CNNs), specifically the Mobile Net architecture, to evaluate high-resolution dermoscopic and clinical skin images and accurately classify various dermatological diseases. The algorithm, which was trained on a large, labelled dataset, can recognize a variety of skin illnesses, including melanoma, eczema, psoriasis, and acne.Mobile Net’s lightweight architecture makes it ideal for deployment on mobile and edge devices, allowing for use in distant environments and medical applications. This system is designed to serve as a decision-support tool, enhancing diagnosis accuracy, minimizing delays, and streamlining healthcare delivery without displacing medical experts. With additional improvement and clinical validation, the system has the potential to have a global impact by improving dermatological treatment accessibility and efficiency.

Introduction

1. Introduction & Motivation

• Global Concern: Skin diseases rank fourth among the leading causes of illness worldwide.

• Challenges: There is a global shortage of dermatologists, especially in rural and underserved regions, leading to delayed diagnoses and poor health outcomes.

• Solution: Artificial Intelligence (AI)—especially Convolutional Neural Networks (CNN) and MobileNet architectures—offers a promising solution for automating dermatological diagnoses via image analysis.

• Impact: AI tools can provide early, fast, and accurate preliminary diagnoses, support telemedicine, help triage cases, and reduce the burden on healthcare systems.

2. Literature Review: Technological Evolution in Dermatology

• From Traditional to Modern Treatments:

  • Ancient remedies relied on herbal treatments and empirical methods.

  • Modern practices now include antibiotics, steroids, biologics, and laser and surgical therapies.

• AI Integration: AI is being used for diagnosis and personalized treatment planning. Emerging tools like wearable sensors and telemedicine platforms are transforming dermatological care.

• Advancements:

  • Laser therapy and Mohs surgery offer precise, targeted treatment.

  • Biologics and genetically tailored therapies are addressing complex conditions like eczema and psoriasis.

  • AI algorithms are becoming more accurate than human specialists in skin disease detection.

3. Methodology

• Model Selection: CNN and MobileNet were chosen for their efficiency in medical image recognition.

• Development Stages:

  1. Data Collection: Curated from clinics and medical databases; diverse in age, gender, skin tone, and condition.

  2. Preprocessing: Image resizing, normalization, privacy protection (anonymization).

  3. Data Annotation: Expert dermatologists label each image with diagnosis and severity.

  4. Data Augmentation: Applied techniques like rotation, flipping, and scaling to improve robustness.

  5. Training & Validation: Trained in Python using TensorFlow and Keras, with Adam optimizer and dropout to improve generalization.

  6. Testing: Final evaluation with an independent dataset for accuracy and reliability.

• Model Architecture:

  • MobileNet is used for its lightweight design and ability to run efficiently on devices with limited computational power.

  • Designed to support real-time diagnosis on mobile or edge devices.

4. Implementation & Results

• CNN Performance: Demonstrated strong capability in analyzing and diagnosing skin conditions from images.

• MobileNet Benefits: Optimized for mobile environments, balancing high accuracy with low computational cost—ideal for use in resource-constrained settings like smartphones or remote clinics.

• Outcome: The system enables quick, reliable, and scalable dermatological screening, especially helpful for early-stage diagnosis and in teledermatology applications.

Conclusion

The project\'s conclusion, which focussed on the construction of a hybrid Mobile Net and LSTM model for dermatological picture categorization, captures both the successes and problems faced in this novel attempt. The study attempted to combine the strengths of convolutional and recurrent neural networks to produce a strong tool for identifying skin disorders, and while it showed promise, the results also revealed crucial areas for improvement. The model\'s capacity to gradually learn and improve its accuracy throughout training epochs demonstrates the feasibility of merging Mobile Net and LSTM for picture classification tasks. This hybrid technique takes advantage of Mobile Net’s efficiency in processing spatial information and LSTM\'s ability to handle sequential data, making it ideal for the complex task of dermatological diagnostics.

References

[1] Al-Masni, M. A., Al-Antari, M. A., Choi, M.-T., Han, S.-M., & Kim, T.-S. (2018). Skinlesion segmentation in dermoscopy images via deep full resolution convolutionalnetworks. Computer Methods and Programs in Biomedicine, 162, 221–231. [2] Alfed, N., Khelifi, F., Bouridane, A., &Seker, H. (2015). Pigment network-based skindisease detection. In 2015 37th annual international conference of the IEEE engineeringin medicine and biology society (EMBC) (pp. 7214–7217). IEEE. [3] Aljanabi, M., Özok, Y. E., Rahebi, J., & Abdullah, A. S. (2018). Skin lesion segmentationmethod for dermoscopy images using artificial bee colony algorithm. Symmetry,10(8), 347. [4] Biessmann, F., Salinas, D., Schelter, S., Schmidt, P., & Lange, D. (2018). ‘‘ Deep\" learningfor missing value imputationin tables with non-numerical data. In Proceedings ofthe 27th ACM international conference on information and knowledge management (pp.2017–2025). [5] Canadian Disease Society’s Advisory Committee on Disease Statistics (2014). Canadiandisease statistics 2014 - special topic: Skin diseases. Retrieved from https://www.disease.ca/statistics. (Accessed 30 December 2020). [6] Carcagnì, P., Leo, M., Cuna, A., Mazzeo, P. L., Spagnolo, P., Celeste, G., et al.(2019). Classification of skin lesions by combining multilevel learnings in adensenet architecture. In International conference on image analysis and processing (pp. 335–344). Springer. [7] Dorj, U.-O., Lee, K.-K., Choi, J.-Y., & Lee, M. (2018). The skin disease classificationusing deep convolutional neural network. Multimedia Tools and Applications, 77(8),9909–9924. [8] Ech-Cherif, A., Misbhauddin, M., &Ech-Cherif, M. (2019). Deep neural network basedmobiledermoscopy application for triaging skin disease detection. In 2019 2ndinternational conference on computer applications & information security (ICCAIS) (pp.1–6). IEEE. [9] Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., et al. (2017).Dermatologist-level classification of skin disease with deep neural networks. Nature,542(7639), 115–118. [10] Garg, R., Maheshwari, S., & Shukla, A. (2019). Decision support system for detectionand classification of skin disease using CNN. arXiv preprint arXiv:1912.03798. [11] Guan, Q., Wang, Y., Ping, B., Li, D., Du, J., Qin, Y., et al. (2019). Deep convolutionalneural network VGG-16 model for differential diagnosing of papillary thyroid carcinomas in cytological images: a pilot study. Journal of Disease, 10(20), 4876. [12] Han, X., Zhong, Y., Cao, L., & Zhang, L. (2017). Pre-trained alexnet architecture with pyramid pooling and supervision for high spatial resolution remote sensing imagescene classification. Remote Sensing, 9(8), 848. [13] Hasan, M., Barman, S. D., Islam, S., & Reza, A. W. (2019). Skin disease detection usingconvolutional neural network. In Proceedings of the 2019 5th international conferenceon computing and artificial intelligence (pp. 254–258). [14] Hobbs, L., Hillson, S., Lawande, S., & Smith, P. (2011). Oracle 10g data warehousing.Elsevier. [15] Hoshyar, A. N., Al-Jumaily, A., &Hoshyar, A. N. (2014). The beneficial techniques inpreprocessing step of skin disease detection system comparing. Procedia ComputerScience, 42, 25–31.

Copyright

Copyright © 2025 B D Gnana Deepak Reddy, A Naveen Kumar, B Yashwanth, C Santosh, G Pavan Kalyan, Mr. O. G. Suresh Kumar, Dr. R. Karunia Krishnapriya, Mr. Pandreti Praveen. 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.