Healthc Inform Res.  2023 Apr;29(2):112-119. 10.4258/hir.2023.29.2.112.

Design of a Machine Learning System to Predict the Thickness of a Melanoma Lesion in a Non-Invasive Way from Dermoscopic Images

• Szijártó Á
• Somfai E
• Lőrincz A

• KMID: 2542137
• DOI: http://doi.org/10.4258/hir.2023.29.2.112

Abstract

Objectives
Melanoma is the deadliest form of skin cancer, but it can be fully cured through early detection and treatment in 99% of cases. Our aim was to develop a non-invasive machine learning system that can predict the thickness of a melanoma lesion, which is a proxy for tumor progression, through dermoscopic images. This method can serve as a valuable tool in identifying urgent cases for treatment.
Methods
A modern convolutional neural network architecture (EfficientNet) was used to construct a model capable of classifying dermoscopic images of melanoma lesions into three distinct categories based on thickness. We incorporated techniques to reduce the impact of an imbalanced training dataset, enhanced the generalization capacity of the model through image augmentation, and utilized five-fold cross-validation to produce more reliable metrics.
Results
Our method achieved 71% balanced accuracy for three-way classification when trained on a small public dataset of 247 melanoma images. We also presented performance projections for larger training datasets.
Conclusions
Our model represents a new state-of-the-art method for classifying melanoma thicknesses. Performance can be further optimized by expanding training datasets and utilizing model ensembles. We have shown that earlier claims of higher performance were mistaken due to data leakage during the evaluation process.

Keyword

Melanoma, Supervised Machine Learning, Dermoscopy, Medical Image Processing, Classification

Figure

• 

  Figure 1 Balanced accuracy of each subset as a function of the dataset size ratio and the reverse exponential curve fit.

• 

  Figure 2 Comparison between an original dermoscopic image (A) and a SMOTE-generated image (B). The two images are very similar, but not identical; for example, the curved line (piece of hair) on the right-hand side of the generated image is copied from a different sample. The original image is from the ISIC2018 Task-1 Challenge dataset (https://challenge.isic-archive.com/data/) provided with a CC0 license. SMOTE: synthetic minority oversampling technique.

Reference

References

1. National Cancer Institute. SEER Cancer Statistics Review (CSR) 1975–2018[Internet]. Bethesda (MD): National Cancer Institute;2021. [cited at 2023 Apr 15]. Available: https://seer.cancer.gov/archive/csr/1975_2018/.

2. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015; 521(7553):436–44. https://doi.org/10.1038/nature14539.

3. Tschandl P, Rinner C, Apalla Z, Argenziano G, Codella N, Halpern A, et al. Human-computer collaboration for skin cancer recognition. Nat Med. 2020; 26(8):1229–34. https://doi.org/10.1038/s41591-020-0942-0.

4. Tschandl P, Codella N, Akay BN, Argenziano G, Braun RP, Cabo H, et al. Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study. Lancet Oncol. 2019; 20(7):938–47. https://doi.org/10.1016/S1470-2045(19)30333-X.

5. Rotemberg V, Halpern A, Dusza S, Codella NC. The role of public challenges and data sets towards algorithm development, trust, and use in clinical practice. Semin Cutan Med Surg. 2019; 38(1):E38–E42. https://doi.org/10.12788/j.sder.2019.013.

6. Rotemberg V, Kurtansky N, Betz-Stablein B, Caffery L, Chousakos E, Codella N, et al. A patient-centric dataset of images and metadata for identifying melanomas using clinical context. Sci Data. 2021; 8(1):34. https://doi.org/10.1038/s41597-021-00815-z.

7. Shaikh WR, Dusza SW, Weinstock MA, Oliveria SA, Geller AC, Halpern AC. melanoma thickness and survival trends in the United States, 1989 to 2009. J Natl Cancer Inst. 2015. 108(1):djv294. https://doi.org/10.1093/jnci/djv294.

8. Rubegni P, Cevenini G, Sbano P, Burroni M, Zalaudek I, Risulo M, et al. Evaluation of cutaneous melanoma thickness by digital dermoscopy analysis: a retrospective study. Melanoma Res. 2010; 20(3):212–7. https://doi.org/10.1097/CMR.0b013e328335a8ff.

9. Saez A, Sanchez-Monedero J, Gutierrez PA, Hervas-Martinez C. Machine learning methods for binary and multiclass classification of melanoma thickness from dermoscopic images. IEEE Trans Med Imaging. 2016; 35(4):1036–45. https://doi.org/10.1109/TMI.2015.2506270.

10. Sanchez-Monedero J, Saez A, Perez-Ortiz M, Gutierrez PA, Hervas-Martinez C. Classification of melanoma presence and thickness based on computational image analysis. In : Proceedings of the 11th International Conference on Hybrid Artificial Intelligent Systems (HAIS); 2016 Apr 18–20; Seville, Spain. p. 427–38. https://doi.org/10.1007/978-3-319-32034-2_36.

11. Jaworek-Korjakowska J, Kleczek P, Gorgon M. Melanoma thickness prediction based on convolutional neural network with VGG-19 model transfer learning. In : Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops; 2019 Jun 16–20; Long Beach, CA. p. 2748–56. https://doi.org/10.1109/CVPRW.2019.00333.

12. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. SMOTE: synthetic minority over-sampling technique. J Artif Intell Res. 2002; 16:321–57. https://doi.org/10.1613/jair.953.

13. Kawahara J, Daneshvar S, Argenziano G, Hamarneh G. 7-Point checklist and skin lesion classification using multi-task multi-modal neural nets. IEEE J Biomed Health Inform. 2018; 23(2):538–46. https://doi.org/10.1109/JBHI.2018.2824327.

14. Tan M, Le Q. EfficientNet: rethinking model scaling for convolutional neural networks. Proc Mach Learn Res. 2019; 97:6105–14.

15. Kuhn M, Johnson K. Feature engineering and selection: a practical approach for predictive models. Boca Raton (FL): CRC Press;2019.

16. Somfai E, Baffy B, Fenech K, Guo C, Hosszu R, Korozs D, et al. Minimizing false negative rate in melanoma detection and providing insight into the causes of classification. Ithaca (NY): arXiv.org;2021. [cited at 2023 Apr 15]. Available from: https://arxiv.org/abs/2102.09199.

17. Brownlee J. Data preparation for machine learning: data cleaning, feature selection, and data transforms in Python [Internet]. [place unknown]: Machine Learning Mastery;2020. [cited at 2023 Apr 15]. Available from: https://machinelearningmastery.com/data-preparation-for-machine-learning/.

18. Frid-Adar M, Klang E, Amitai M, Goldberger J, Greenspan H. Synthetic data augmentation using GAN for improved liver lesion classification. In : Proceedings of 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI); 2018 Apr 4–7; Washington, DC. 289–93. https://doi.org/10.1109/ISBI.2018.8363576.