Skin Disease Detection using Machine Learning

Abstract

Cancer is a serious illness due to someone’s abnormalcellgrowththatcanspreadthroughoutthe body. Although skin cancer is uncommon, it isstill quite dangerous. Basal cell carcinoma, squamous cell carcinoma, and melanoma are the three main skin cancer types. In the U.S., melanoma is 1% of skin cancer cases, but it is the reason for the most skin cancer deaths. The American Cancer Society estimated that about 8,430 melanoma deaths will occur in 2025, 5470ofwhomaremenand2960are women.Fortunately,thedeathratesfrommelanoma are decreasing much more thanks to better treatments from 2013 to 2022. To effectively treat skin cancer, it should be detected early. When necessary, the diagnostic process usually begins with a clinical examination by a dermatologist, followed by a dermoscopy, then a biopsy, and a histopathological evaluation.Thesediagnosticsteps can take several weeks in total. Many healthcare professionals argue that early diagnosis brings greater success. Machine learning algorithms can help detect skin cancer quickly, which can ultimately improve the prognosis of the patient.

Introduction

Cancer involves uncontrolled cell growth that invades nearby tissues, with skin cancer being less common but still deadly. Melanoma, a type of skin cancer, accounts for about 4% of skin cancer cases but causes roughly 75% of skin cancer deaths. Many studies use image processing and machine learning (ML) to classify skin cancer for better detection. Previous models often relied on small or low-resolution datasets and mostly used traditional ML algorithms like K-Nearest Neighbors (KNN). This project improves upon past work by using a larger, higher-resolution dataset (ISIC_MSK-2 with 1535 images) and a Convolutional Neural Network (CNN) model for more accurate skin cancer classification.

Objectives:

• Study skin cancer and related research

• Develop a model for predicting skin cancer using high-resolution images

• Automate detection for faster and more accurate classification

• Reduce treatment time through improved prediction accuracy

Literature Review Highlights:

• Various image processing and ML techniques have been explored, such as K-means clustering, Support Vector Machines (SVM), Artificial Neural Networks (ANN), and ensemble models involving VGG16, ResNet50, and InceptionV3.

• Methods often involve lesion segmentation, feature extraction, and classification phases.

• Accuracy in skin disease classification models ranges typically from 80% to 95%, with deep learning models generally outperforming traditional ML approaches.

• Some studies focus on melanoma detection using datasets like PH2 and employ advanced techniques like fuzzy clustering, GrabCut segmentation, and deep learning.

Tools and Techniques Used:

• Python programming language with libraries such as NumPy, Keras, Pandas, and Matplotlib for data handling, machine learning, and visualization.

• Neural Networks simulate brain-like operations for pattern recognition and decision-making.

• Convolutional Neural Networks (CNNs) are specialized deep neural networks effective for image recognition and feature extraction, inspired by the human visual cortex.

• VGG16, a popular CNN architecture known for high accuracy on large-scale image recognition tasks, is highlighted as an influential model.

The project aims to leverage CNNs and high-resolution dermoscopic images to create a scalable, cost-effective, and accurate system for early detection and classification of skin cancer, improving patient outcomes.

Conclusion

Thisworktriestohelppatientsspotskincancerby using automated prediction with neural networks, so they don\'t have to go to the hospital right away. Different models, like ResNet50, CNN, and VGGNet16 were used for classification.VGGNet16 got the best accuracy of 81.91%. The waythismodelworkscanbeimprovedbytweaking hyperparameters or making the dataset larger with more samples. Also, trying out different pooling methods, architectures, and optimizers can lead to big changes in how well the model works.Dataaugmentationtechniquesmaybeapplied in future works to address class imbalance and promote generalization. Further, an exploration of more advanced deep learning architectures accompanied by ensemble learning would likely lead to the further optimization of accuracy in skin cancer detection.

References

[1] Kibria, G., Firoze, A., Amini, A., & Yan, H. (2012) “Dermatological Disease Diagnosis Using Color-Skin Images.” Xian: International Conference on Machine Learning and Cybernetics. [2] S.Mustafa,A.B.Dauda,andM.Dauda,?Image processing and SVM classification formelanomadetection,?2017InternationalConference on Computing Networking and Informatics (ICCNI), 2017. [3] M. E. Celebi, H. A. Kingravi, Y. A. Aslandogan,andW.V.Stoecker,?Detection of blue-white veil areas in dermoscopy images using machine learning techniques,? Medical Imaging 2006: Image Processing, Feb. 2006. [4] Ali, S. N. et al. Monkeypox Skin Lesion Detection Using Deep Learning Models: A Feasibility Study, Comput. Vis. Pattern Recognit., pp. 2–5, [Online]. Available: (2022). http://arxiv.org/abs/2207.03342 [5] I.A. Ozkan and M. Koklu, ?Skin Lesion Classification using Machine Learning Algorithms,? International Journal of Intelligent Systems and Applications in Engineering,vol.4,no.5,pp.285–289,2017. [6] P.R.Kshirsagar,H.Manoharan,S.Shitharth,A.M.Alshareef,N.Albishry,P.K. Balachandran,Deeplearningapproachesforprognosis of automated skin disease, Life 2022 12 (426) (2022). [7] Z.Waheed,A.Waheed,M.Zafar,andF.Riaz,?An efficient machine learning approach for the detection of melanoma using dermoscopic International images,? 2017 Conference on Communication, Computing medicine and Biology Society (EMBC), 2016. [8] R.S.S.SundarandM.Vadivel,?Performance analysis of melanoma early detection using skin lesion classification system,? 2016 International Conference on Circuit, Power and ComputingTechnologies (ICCPCT), 2016. [9] A.Masood,A.A.-Jumaily,andK.Anam,?Self-supervised learning model for skin cancer diagnosis,? 2015 7th International IEEE/EMBS Conference on Neural Engineering (NER), 2015. [10] S. Mustafa and A. Kimura, ?A SVM-based diagnosis of melanoma using only useful image features,? 2018 International Workshop on Advanced Image Technology (IWAIT), 2018. [11] Jagdisetal., J.A.D.L. Cruz-Vargas, M.E.R. Camacho, Advance study of skin diseases detection using image processing methods, Nat.VolatilesEssent.OilsJ.9(1)(2022)997–1007. [12] www.medium.com

Copyright

Copyright © 2025 Sachin Bansal, Swanjal Gupta, Talha Pathan, Ansh Gangwar, Vansh Agarwal. 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.