Design of a tool for the classification of skin cancer images using Deep Neural Networks (DNN)
Abstract
Skin cancer is one of the most common diseases in the world population. Usually, the diagnosis requires the acquisition of dermatoscopic images. Both biopsy and histopathology have been used in advanced stages. Its early detection is very important to increase patient life quality and life expectancy. In Colombia, the lack of qualified professionals and medical instruments difficulties this task. The automatic classification is a huge challenge, due to ample variety and morphology in skin lesions. Nowadays, Deep Learning reaches elevated accuracy levels in image classification tasks and is set to become a reliable solution for medical image classification. In this research, used these DNN advantages to build a convolutional neural network (CNN) trained with open source databases to the classification of skin lesions benign and malignant. After the training process, we develop an embedded system with raspberry Pi 3 B+ with a generic camera and implemented the CNN described in Python coded-based. For the benign and malignant classification, the prototype reached an accuracy level of 91.06% in the F1 score and a recall of 91.98%.
Downloads
References
Castillo, J. A., Granados, Y. C., & Fajardo, C. A. (2020). Patient-specific detection of atrial fibrillation in segments of ecg signals using deep neural networks. Ciencia E Ingenieria Neogranadina, vol. 30, no. 1, 45–58.
Chanampe, H., Aciar, S., Vega, M. d., Molinari Sotomayor, J. L., Carrascosa, G., & Lorefice, A. (2019). Modelo de redes neuronales convolucionales profundas para la clasificacion de lesiones en ecograf´ ´ıas mamarias. in XXI Workshop de Investigadores en Ciencias de la Computacion (WICC´ 2019, Universidad Nacional de San Juan).
Chicco, D. J. (2020). The advantages of the matthews correlation coefficient (mcc) over f1 score and accuracy in binary classification evaluation. BMC genomics, vol. 21, no. 1, 6.
Diana, H. (2021). Tesisredes-dianahellis-2021/skincancerclassifier. Obtenido de https://github.com/TesisRedes-DianaHellis-2021?tab=repositorie
Gamerosa, P. C., & Tellez, J. E. (2016). El cancer de piel, un problema actual. Revista de la Facultad de Medicina UNAM, vol. 59, no. 2, 6–14.
Glorot, X., & Bengio, Y. (2010). Understanding the difficulty of training deep feedforward neural networks. Proceedings of the thirteenth international conference on artificial intelligence and statistics, 249–256.
Instituto de Hidrología, Meteorología y Estudios Ambientales. (2014). “Indice ultravioleta (iuv) - ideam. Obtenido de http://www.ideam.gov.co/web/tiempo-y-clima/indice-ultravioleta-iuv-
Kaggle. (s.f.). Skin Cancer ISIC. Obtenido de https://www.kaggle.com/nodoubttome/skin-cancer9-classesisic
Mao, H. H. (2020). A survey on self-supervised pre-training for sequential transfer learning in neural networks. arXiv preprint arXiv:2007.00800.
Mikołajczyk, A., & Grochowski, M. (2018). Data augmentation for improving deep learning in image classification problem. 2018 international interdisciplinary PhD workshop (IIPhDW). IEEE, 117–122.
Morocho Jiménez, J. I. (2019). Detección de tumores cutáneos malignos y benignos usando una red neuronal convolucional. Quito: B.S. thesis.
Ng, A., Besounda Mourris, Y., & Karantoforoosh, K. (2021). Deep learning. Obtenido de https://www.coursera.org/specializations/deep-learning
Pérez Lorenzo, C., & et al. (2019). Detección precoz de cáncer de piel en imágenes basado en redes convolucionales,. B.S. thesis.
Powers, D. M. (2010). Evaluation: from precesion, recall and f-measure to roc, informedness, makedness and correlation. arXiv preprint a arxiv, 16061.
Qin, Z., Yu, F., Liu, C., & Chen, X. (2018). How convolutional neural network see the world-a survey of convolutional neural network visualization methods. arXiv preprint arXiv:1804.11191.
Sanjay, M. (2018). Why and how to Cross Validate a Model? Obtenido de https://towardsdatascience.com/why-and-how-to-cross-validate-a-model-d6424b45261f
Shin, H.-C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., . . . Summers, R. M. (2016). Deep convolutional neural networks for computer-aided detection: Cnn architectures, dataset characteristics and transfer learning. IEEE transactions on medical imaging, vol. 35, no. 5, 1285–1298.
Shorten, C., & Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, vol. 6, no. 1,, 60.
Sokolova, M., Japkowicz, N., & Szpakowicz, S. (2006). Beyond accuracy, f-score and roc: a family of discriminant measures for performance evaluation. Australasian joint conference on artificial intelligence. Springer, 1015–1021.
Static Raspberrypi. (s.f.). Raspberry pi modelbplus product brief. Obtenido de https://static.raspberrypi.org/files/product-briefs/raspberry-pi-modelbplus-product-brief.pdf
Team, K. (s.f.). Keras documentation: Keras applications. Obtenido de https://keras.io/api/applications
Tschandl, P. (2018). The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Obtenido de https://doi.org/10.7910/DVN/DBW86T
Xia, B., Zhang, H., & Li, Q. L. (2015). Pets: A stable and accurate redictor of protein-protein interacting sites based on xtremely-randomized ree. IEEE Transactions on NanoBioscience, vol. 14, 1–1, 11 .
The articles published in the journal Ciencia y Tecnología are the exclusive property of their authors. Their opinions and content belong to their authors, and the Universidad de Palermo declines all responsibility for the rights that may arise from reading and/or interpreting the content of the published articles.
The reproduction, use or exploitation by any third party of the published articles is not authorized. Its use is only authorized for exclusively academic and/or research purposes.
