Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches

Mahbub Hasan; Md Shohel Babu

doi:doi:10.11648/j.mcs.20251006.11

Research Article |

| Peer-Reviewed

Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches

Mahbub Hasan

, Md Shohel Babu^*

Published in Mathematics and Computer Science (Volume 10, Issue 6)

Received: 15 September 2025 Accepted: 25 September 2025 Published: 10 November 2025

Views: Downloads:

Download PDF

Share This Article

Twitter
Linked In
Facebook

Abstract

Melanoma is one of the most aggressive forms of skin cancer, accounting for an excessive number of skin cancer-related deaths despite its relatively low incidence. Early and accurate detection is crucial to improving survival rates, but manual diagnosis through dermoscopy remains challenging due to the subtle lesion characteristics. In this study, we propose a lightweight and resource-efficient framework for melanoma detection that integrates image segmentation with multiple instance learning (MIL)-based classification. Dermoscopic images from the ISIC 2018 dataset were preprocessed using resizing, noise removal, and hair artifact reduction to improve visual consistency. Lesion segmentation was performed using a statistical measure of pixel intensity standard deviation (SD), which adaptively determined the most suitable enhancement and thresholding techniques across three criteria. Segmentation outputs were then used to extract region-specific features, which were classified using three MIL-based approaches: mi-KSPSVM, mi-SPSVM, and MIL with Lagrangian relaxation, along with standard SVM baselines. Model performance was evaluated under 5-fold, 10-fold, and Leave-One-Out (LOO) cross-validation schemes using accuracy, sensitivity, specificity, F1-score, and computational efficiency as metrics. Experimental results demonstrate that segmentation significantly enhances classification performance, particularly in terms of sensitivity and F1-score. Notably, mi-KSPSVM achieved the highest sensitivity of 94.74% under 10-fold CV, while mi-SPSVM exhibited superior generalization under LOO validation. Additionally, preprocessing and segmentation consistently reduced misclassification of melanoma cases while preserving computational efficiency, with the SVM (RBF) baseline remaining the fastest model. Future work will explore the integration of deep learning-based segmentation to improve accuracy further and extend applicability to other skin cancer subtypes.

Published in	Mathematics and Computer Science (Volume 10, Issue 6)
DOI	10.11648/j.mcs.20251006.11
Page(s)	92-101
Creative Commons	This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.
Copyright	Copyright © The Author(s), 2025. Published by Science Publishing Group

Keywords

Melanoma, Segmentation, MIL, SVM, Machine Learning, Classification, Multiple Instance Learning

References

[1]	D. Whiteman, A. Green, and J. Olsen, “Melanoma incidence and mortality trends in 21 countries,” Int. J. Cancer, vol. 138, no. 2, pp. 406-418, Jan. 2016, https://doi.org/10.1002/ijc.29663
[2]	J. E. Gershenwald et al., “Melanoma staging: Evidence based changes in the American Joint Committee on Cancer eighth edition cancer staging manual,” CA Cancer J. Clin., vol. 67, no. 6, pp. 472-492, Nov. 2017, https://doi.org/10.3322/caac.21409
[3]	G. Argenziano et al., “Dermoscopy of pigmented skin lesions: A valuable tool for early diagnosis,” Lancet Oncol., vol. 4, no. 7, pp. 429-439, Jul. 2003, https://doi.org/10.1016/S1470-2045(03)01183-6
[4]	A. Esteva et al., “Dermatologist-level classification of skin cancer with deep neural networks,” Nature, vol. 542, pp. 115-118, Jan. 2017, https://doi.org/10.1038/nature21056
[5]	G. Campanella et al., “Clinical-grade computational pathology using weakly supervised deep learning on whole-slide images,” Nat. Med., vol. 25, pp. 1301-1309, Aug. 2019, https://doi.org/10.1038/s41591-019-0508-1
[6]	Codella, N., Rotemberg, V., Tschandl, P., Celebi, M. E., Dusza, S., Gutman, D., ... & Halpern, A. (2019). Skin lesion analysis toward melanoma detection 2018: A challenge hosted by the International Skin Imaging Collaboration (isic). arXiv preprint arXiv:1902.03368.
[7]	P. Tschandl, C. Rosendahl, and H. Kittler, “The HAM10000 dataset: A large collection of multi source dermatoscopic images of common pigmented skin lesions,” Sci. Data, vol. 5, no. 180161, Aug. 2018, https://doi.org/10.1038/sdata.2018.161
[8]	Mendonca, T., Ferreira, P. M., Marques, J. S., Marcal, A. R., & Rozeira, J. (2013). PH2- a dermoscopic image database for research and benchmarking. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference, 2013, 5437-5440. https://doi.org/10.1109/EMBC.2013.6610779
[9]	J. Winkler et al., “Association between training data and performance of deep learning algorithms in dermatology,” JAMA Dermatol., vol. 155, no. 11, pp. 1282-1288, Nov. 2019, https://doi.org/10.1001/jamadermatol.2019.1735
[10]	M. E. Celebi, H. A. Kingravi, B. Uddin, H. Iyatomi, Y. A. Aslandogan, W. V. Stoecker, and R. H. Moss, “A methodological approach to the classification of dermoscopy images,” Comput. Med. Imaging Graph., vol. 31, no. 6, pp. 362-373, Sep. 2007, https://doi.org/10.1016/j.compmedimag.2007.01.003
[11]	A. S. Ashour, Y. Guo, E. Kucukkulahli, P. Erdogmus, and K.Polat, “A hybrid dermoscopy images segmentation approach based on neutrosophic clustering and histogram estimation,” Appl. Soft Comput., vol. 69, pp. 426-434, Aug. 2018, https://doi.org/10.1016/j.asoc.2018.05.003
[12]	X. Bian, H. Pan, K. Zhang, C. Chen, P. Liu, and K. Shi, “NeDSeM: Neutrosophy domain-based segmentation method for malignant melanoma images,” Entropy, vol. 24, no. 6, p. 783, Jun. 2022, https://doi.org/10.3390/e24060783
[13]	S. Oukil, R. Kasmi, K. Mokrani, and B. Garc˜ Aa-Zapirain, “Automatic segmentation and melanoma detection based on color and texture features in dermoscopic images,” Skin Res. Technol., vol. 28, no. 2, pp. 203-211, Mar. 2022, https://doi.org/10.1111/srt.13111
[14]	M. Mabrouk, H. Benali, and N. Almotiri, “Automated detection of pigmented skin lesions,” J. King Saud Univ. Comput. Inf. Sci., vol. 32, no. 1, pp. 90-96, Jan. 2020, https://doi.org/10.1016/j.jksuci.2017.10.002
[15]	Y. Zhang, X. Liu, and H. Chen, “Unsupervised deep learning for dermoscopic image segmentation,” Pattern Recognit. Lett., vol. 146, pp. 1-7, Jan. 2021, https://doi.org/10.1016/j.patrec.2021.03.012
[16]	H. Li, L. Xu, and P. Liu, “Unsupervised histopathology image segmentation for melanoma,” Biomed. Signal Process. Control, vol. 70, p. 103012, Sep. 2021, https://doi.org/10.1016/j.bspc.2021.103012
[17]	N. Codella, D. Gutman, M. E. Celebi, B. Helba, M. A. Marchetti, S. W. Dusza, A. K. Kalloo, K. Liopyris, N. K. Mishra, H. Kittler, and A. Halpern, “Skin lesion analysis toward melanoma detection: A challenge at the 2018 ISIC workshop,” arXiv:1902.03368, Feb. 2019, https://doi.org/10.48550/arXiv.1902.03368
[18]	K. Liu, Y. Song, L. Zhang, et al., “Learning melanocytic proliferation segmentation in histopathology images from imperfect annotations,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. Workshops (CVPRW), Nashville, TN, USA, 2021, pp. 3761-3770, https://doi.org/10.1109/CVPRW53098.2021.00417
[19]	Y. Wang, J. Zhao, and H. Liu, “Boundary-aware transformer for skin lesion segmentation,” IEEE Trans. Med. Imaging, vol. 40, no. 10, pp. 2627-2637, Oct. 2021, https://doi.org/10.1109/TMI.2021.3073616
[20]	L. Cai, J. Wang, S. Wang, and M. Ma, “Intelligent skin lesion segmentation using deformable attention transformer U-Net (BiADATU-Net),” Front. Neurorobot., vol. 18, p. 1368990, May 2024, https://doi.org/10.3389/fnbot.2024.1368990
[21]	J. Xu, Y. Zhang, and M. Liu, “Lightweight boundary assisted UNet for skin lesion segmentation,” in Proc. MICCAI, 2024, pp. 1-12, https://doi.org/10.1007/978-3 031-70353-1 35
[22]	A. Author, B. Author, and C. Author, “Transformer-based networks for skin lesion segmentation,” in Proc. Mach. Learn. Med. Imaging (MLMI), MICCAI, 2023, pp. 351 360, https://doi.org/10.1007/978-3-031-43990-4 35
[23]	M. Kreouzi, E. Iliopoulou, and K. Marias, “Deep learning for melanoma detection: A review,” Cancers, vol. 17, no. 1, p. 28, Jan. 2024, https://doi.org/10.3390/cancers17010028
[24]	Y. Wu et al., “Skin cancer classification with deep learning: A systematic review,” Frontiers in Oncology, vol. 12, Art. no. 893972, Jul. 2022.[Online]. Available: https://doi.org/10.3389/fonc.2022.893972
[25]	Z. Yu, M. Gutman, A. Codella, et al., “Deep learning for melanoma diagnosis: Results from a large-scale, multicenter study,” JAMA Dermatol., vol. 156, no. 10, pp. 1136-1142, Oct. 2020, https://doi.org/10.1001/jamadermatol.2020.1708
[26]	Y. Zhang, H. Zhang, and L. Li, “Deep learning-based melanoma detection from histopathological images,” J. Pathol. Inform., vol. 12, p. 53, 2021, https://doi.org/10.4103/jpi.jpi 53 21
[27]	M. Alshawi, A. Alshamrani, and A. Alyami, “Deep ensemble learning for multiclass skin lesion classification,” Bioengineering, vol. 12, no. 9, p. 934, Sep. 2023, https://doi.org/10.3390/bioengineering12090934
[28]	G. Yang, S. Wang, and J. Li, “Boosting skin cancer classification: A multi-scale attention and ensemble learning approach,” Sensors, vol. 25, no. 8, p. 2479, Apr. 2025, https://doi.org/10.3390/s25082479
[29]	L. Shao, B. Zhu, and W. Xie, “TransMIL: Transformer based correlated multiple instance learning for whole slide image classification,” in Proc. 35th AAAI Conf. Artif. Intell. (AAAI), vol. 35, no. 3, pp. 6141-6149, May 2021, https://doi.org/10.1609/aaai.v35i7.16726
[30]	Y. Li, Z. Liu, and T. Jiang, “Dual-stream multiple instance learning with contrastive learning for whole-slide classification,” Med. Image Anal., vol. 73, p. 102157, Nov. 2021, https://doi.org/10.1016/j.media.2021.102157
[31]	L. Godson, B. E. Bejnordi, M. Veta, et al., “Immune subtyping of melanoma whole slide images using multiple instance learning,” Med. Image Anal., vol. 93, p. 103097, Apr.2024, https://doi.org/10.1016/j.media.2024.103097
[32]	P. Meseguer-Esbri, R. del Amor, and V. Naranjo, “MiCIL: Multiple-instance class-incremental learning for skin cancer whole slide images,” Artif. Intell. Med., vol. 152, p. 102870, Feb. 2024, https://doi.org/10.1016/j.artmed.2024.102870
[33]	Z. Gao, A. Mao, Y. Dong, et al., “Accurate spatial quantification in computational pathology with multiple instance learning,” medRxiv, preprint, Mar. 2024, https://doi.org/10.1101/2024.03.20.24305321
[34]	H. Xu, J. Zhang, and X. Zhou, “When multiple instance learning meets foundation models in computational pathology,” Patterns, in press, 2025, https://doi.org/10.1016/j.patter.2025.100987
[35]	Astorino, A., Fuduli, A. and Gaudioso, M., 2019. A Lagrangian relaxation approach for binary multiple instance classification. IEEE transactions on neural networks and learning systems, 30(9), pp. 2662-2671. https://doi.org/10.1109/TNNLS.2018.2885852
[36]	Avolio, M. and Fuduli, A., 2020. A semiproximal support vector machine approach for binary multiple instance learning. IEEE transactions on neural networks and learning systems, 32(8), pp. 3566-3577. https://doi.org/10.1109/TNNLS.2020.3015442
[37]	Avolio, M. and Fuduli, A., 2024. The semiproximal SVM approach for multiple instance learning: a kernel-based computational study. Optimization Letters, 18(2), pp. 635-649. https://doi.org/10.1007/s11590-023-02022-8

Cite This Article

Plain Text BibTeX RIS

APA Style

Hasan, M., Babu, M. S. (2025). Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Mathematics and Computer Science, 10(6), 92-101. https://doi.org/10.11648/j.mcs.20251006.11

Copy | Download

ACS Style

Hasan, M.; Babu, M. S. Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Math. Comput. Sci. 2025, 10(6), 92-101. doi: 10.11648/j.mcs.20251006.11

Copy | Download

AMA Style

Hasan M, Babu MS. Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Math Comput Sci. 2025;10(6):92-101. doi: 10.11648/j.mcs.20251006.11

Copy | Download

@article{10.11648/j.mcs.20251006.11,
  author = {Mahbub Hasan and Md Shohel Babu},
  title = {Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches},
  journal = {Mathematics and Computer Science},
  volume = {10},
  number = {6},
  pages = {92-101},
  doi = {10.11648/j.mcs.20251006.11},
  url = {https://doi.org/10.11648/j.mcs.20251006.11},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.mcs.20251006.11},
  abstract = {Melanoma is one of the most aggressive forms of skin cancer, accounting for an excessive number of skin cancer-related deaths despite its relatively low incidence. Early and accurate detection is crucial to improving survival rates, but manual diagnosis through dermoscopy remains challenging due to the subtle lesion characteristics. In this study, we propose a lightweight and resource-efficient framework for melanoma detection that integrates image segmentation with multiple instance learning (MIL)-based classification. Dermoscopic images from the ISIC 2018 dataset were preprocessed using resizing, noise removal, and hair artifact reduction to improve visual consistency. Lesion segmentation was performed using a statistical measure of pixel intensity standard deviation (SD), which adaptively determined the most suitable enhancement and thresholding techniques across three criteria. Segmentation outputs were then used to extract region-specific features, which were classified using three MIL-based approaches: mi-KSPSVM, mi-SPSVM, and MIL with Lagrangian relaxation, along with standard SVM baselines. Model performance was evaluated under 5-fold, 10-fold, and Leave-One-Out (LOO) cross-validation schemes using accuracy, sensitivity, specificity, F1-score, and computational efficiency as metrics. Experimental results demonstrate that segmentation significantly enhances classification performance, particularly in terms of sensitivity and F1-score. Notably, mi-KSPSVM achieved the highest sensitivity of 94.74% under 10-fold CV, while mi-SPSVM exhibited superior generalization under LOO validation. Additionally, preprocessing and segmentation consistently reduced misclassification of melanoma cases while preserving computational efficiency, with the SVM (RBF) baseline remaining the fastest model. Future work will explore the integration of deep learning-based segmentation to improve accuracy further and extend applicability to other skin cancer subtypes.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches
AU  - Mahbub Hasan
AU  - Md Shohel Babu
Y1  - 2025/11/10
PY  - 2025
N1  - https://doi.org/10.11648/j.mcs.20251006.11
DO  - 10.11648/j.mcs.20251006.11
T2  - Mathematics and Computer Science
JF  - Mathematics and Computer Science
JO  - Mathematics and Computer Science
SP  - 92
EP  - 101
PB  - Science Publishing Group
SN  - 2575-6028
UR  - https://doi.org/10.11648/j.mcs.20251006.11
AB  - Melanoma is one of the most aggressive forms of skin cancer, accounting for an excessive number of skin cancer-related deaths despite its relatively low incidence. Early and accurate detection is crucial to improving survival rates, but manual diagnosis through dermoscopy remains challenging due to the subtle lesion characteristics. In this study, we propose a lightweight and resource-efficient framework for melanoma detection that integrates image segmentation with multiple instance learning (MIL)-based classification. Dermoscopic images from the ISIC 2018 dataset were preprocessed using resizing, noise removal, and hair artifact reduction to improve visual consistency. Lesion segmentation was performed using a statistical measure of pixel intensity standard deviation (SD), which adaptively determined the most suitable enhancement and thresholding techniques across three criteria. Segmentation outputs were then used to extract region-specific features, which were classified using three MIL-based approaches: mi-KSPSVM, mi-SPSVM, and MIL with Lagrangian relaxation, along with standard SVM baselines. Model performance was evaluated under 5-fold, 10-fold, and Leave-One-Out (LOO) cross-validation schemes using accuracy, sensitivity, specificity, F1-score, and computational efficiency as metrics. Experimental results demonstrate that segmentation significantly enhances classification performance, particularly in terms of sensitivity and F1-score. Notably, mi-KSPSVM achieved the highest sensitivity of 94.74% under 10-fold CV, while mi-SPSVM exhibited superior generalization under LOO validation. Additionally, preprocessing and segmentation consistently reduced misclassification of melanoma cases while preserving computational efficiency, with the SVM (RBF) baseline remaining the fastest model. Future work will explore the integration of deep learning-based segmentation to improve accuracy further and extend applicability to other skin cancer subtypes.

VL  - 10
IS  - 6
ER  -

Copy | Download

Author Information

Mahbub Hasan

Department of Computer Science and Engineering, Southeast University, Dhaka, Bangladesh

Contact Email

http://orcid.org/0009-0000-7906-1823
Md Shohel Babu

Department of Computer Science and Engineering, Southeast University, Dhaka, Bangladesh

Contact Email

http://orcid.org/0000-0002-7319-9028

Download PDF

Submit an Article

Sections

Plain Text BibTeX RIS

APA Style

Hasan, M., Babu, M. S. (2025). Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Mathematics and Computer Science, 10(6), 92-101. https://doi.org/10.11648/j.mcs.20251006.11

Copy | Download

ACS Style

Hasan, M.; Babu, M. S. Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Math. Comput. Sci. 2025, 10(6), 92-101. doi: 10.11648/j.mcs.20251006.11

Copy | Download

AMA Style

Hasan M, Babu MS. Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches. Math Comput Sci. 2025;10(6):92-101. doi: 10.11648/j.mcs.20251006.11

Copy | Download

@article{10.11648/j.mcs.20251006.11,
  author = {Mahbub Hasan and Md Shohel Babu},
  title = {Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches},
  journal = {Mathematics and Computer Science},
  volume = {10},
  number = {6},
  pages = {92-101},
  doi = {10.11648/j.mcs.20251006.11},
  url = {https://doi.org/10.11648/j.mcs.20251006.11},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.mcs.20251006.11},
  abstract = {Melanoma is one of the most aggressive forms of skin cancer, accounting for an excessive number of skin cancer-related deaths despite its relatively low incidence. Early and accurate detection is crucial to improving survival rates, but manual diagnosis through dermoscopy remains challenging due to the subtle lesion characteristics. In this study, we propose a lightweight and resource-efficient framework for melanoma detection that integrates image segmentation with multiple instance learning (MIL)-based classification. Dermoscopic images from the ISIC 2018 dataset were preprocessed using resizing, noise removal, and hair artifact reduction to improve visual consistency. Lesion segmentation was performed using a statistical measure of pixel intensity standard deviation (SD), which adaptively determined the most suitable enhancement and thresholding techniques across three criteria. Segmentation outputs were then used to extract region-specific features, which were classified using three MIL-based approaches: mi-KSPSVM, mi-SPSVM, and MIL with Lagrangian relaxation, along with standard SVM baselines. Model performance was evaluated under 5-fold, 10-fold, and Leave-One-Out (LOO) cross-validation schemes using accuracy, sensitivity, specificity, F1-score, and computational efficiency as metrics. Experimental results demonstrate that segmentation significantly enhances classification performance, particularly in terms of sensitivity and F1-score. Notably, mi-KSPSVM achieved the highest sensitivity of 94.74% under 10-fold CV, while mi-SPSVM exhibited superior generalization under LOO validation. Additionally, preprocessing and segmentation consistently reduced misclassification of melanoma cases while preserving computational efficiency, with the SVM (RBF) baseline remaining the fastest model. Future work will explore the integration of deep learning-based segmentation to improve accuracy further and extend applicability to other skin cancer subtypes.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - Clinically Reliable Melanoma Classification from Darmoscopy Image Using Linear and Nonlinear MIL Approaches
AU  - Mahbub Hasan
AU  - Md Shohel Babu
Y1  - 2025/11/10
PY  - 2025
N1  - https://doi.org/10.11648/j.mcs.20251006.11
DO  - 10.11648/j.mcs.20251006.11
T2  - Mathematics and Computer Science
JF  - Mathematics and Computer Science
JO  - Mathematics and Computer Science
SP  - 92
EP  - 101
PB  - Science Publishing Group
SN  - 2575-6028
UR  - https://doi.org/10.11648/j.mcs.20251006.11
AB  - Melanoma is one of the most aggressive forms of skin cancer, accounting for an excessive number of skin cancer-related deaths despite its relatively low incidence. Early and accurate detection is crucial to improving survival rates, but manual diagnosis through dermoscopy remains challenging due to the subtle lesion characteristics. In this study, we propose a lightweight and resource-efficient framework for melanoma detection that integrates image segmentation with multiple instance learning (MIL)-based classification. Dermoscopic images from the ISIC 2018 dataset were preprocessed using resizing, noise removal, and hair artifact reduction to improve visual consistency. Lesion segmentation was performed using a statistical measure of pixel intensity standard deviation (SD), which adaptively determined the most suitable enhancement and thresholding techniques across three criteria. Segmentation outputs were then used to extract region-specific features, which were classified using three MIL-based approaches: mi-KSPSVM, mi-SPSVM, and MIL with Lagrangian relaxation, along with standard SVM baselines. Model performance was evaluated under 5-fold, 10-fold, and Leave-One-Out (LOO) cross-validation schemes using accuracy, sensitivity, specificity, F1-score, and computational efficiency as metrics. Experimental results demonstrate that segmentation significantly enhances classification performance, particularly in terms of sensitivity and F1-score. Notably, mi-KSPSVM achieved the highest sensitivity of 94.74% under 10-fold CV, while mi-SPSVM exhibited superior generalization under LOO validation. Additionally, preprocessing and segmentation consistently reduced misclassification of melanoma cases while preserving computational efficiency, with the SVM (RBF) baseline remaining the fastest model. Future work will explore the integration of deep learning-based segmentation to improve accuracy further and extend applicability to other skin cancer subtypes.

VL  - 10
IS  - 6
ER  -

Copy | Download