Class Imbalance in Deep Learning-Based Image Classification: A Systematic Review
DOI:
https://doi.org/10.30871/jaic.v10i5.13881Keywords:
Class Imbalance, Deep Learning, Image Classification, Imbalanced Learning, Systematic Literature ReviewAbstract
Class imbalance is a major challenge in deep learning-based image classification because unequal class distributions can bias model learning toward majority classes and reduce minority-class recognition. This study conducted a Systematic Literature Review (SLR) to identify, classify, and synthesize recent approaches for handling class imbalance in deep learning-based image classification. The review followed the PRISMA 2020 framework and used Scopus as the primary bibliographic database, with publications from 2022 to 2026. A total of 1,630 records were initially retrieved, and 34 articles underwent full-text assessment. After applying the predefined eligibility criteria, 31 studies were included in the final thematic synthesis. The studies were organized into four methodological themes: Traditional Data-Level Approaches, Algorithm-Level and Loss Function-Based Approaches, Generative and Hybrid Approaches, and Emerging Architecture-Integrated Approaches. The synthesis showed that traditional methods such as SMOTE, ADASYN, and augmentation remained useful, while algorithm-level, generative, and architecture-integrated approaches provided alternative mechanisms for addressing minority-class learning. No single approach consistently demonstrated superior performance across different datasets and experimental settings. Effectiveness was influenced by dataset characteristics, class distribution, architecture, evaluation procedures, and computational requirements. The review identifies research gaps in cross-dataset validation, standardized benchmarking, minority-sensitive evaluation, computational-efficiency reporting, and validation of synthetic-data quality and its effect on classification performance.
Downloads
References
[1] M. Shafiq and Z. Gu, “applied sciences Deep Residual Learning for Image Recognition : A Survey,” Appl. Sci., pp. 1–43, 2022.
[2] J. Plested, M. Phiri, and T. & Gedeon, “Deep transfer learning for image classification: A survey,” Artif. Intell. Rev., 2026, doi: 10.1007/s10462-026-11491-z.
[3] W. Chen, K. Yang, Z. Yu, Y. Shi, and C. L. P. Chen, A survey on imbalanced learning : latest research , applications and future directions, vol. 123. 2024. doi: 10.1007/s10462-024-10759-6.
[4] C. J. Hellín, A. A. Olmedo, and A. Tayebi, “applied sciences Unraveling the Impact of Class Imbalance on Deep-Learning Models for Medical Image Classification,” 2024.
[5] F. Rodríguez-torres and J. F. Martínez-trinidad, “applied sciences An Oversampling Method for Class Imbalance Problems on Large Datasets,” 2022.
[6] M. O. Id, J. Ben Hayfron-acquah, T. Frimpong, and G. Abdul-, “Imbalanced class distribution and performance evaluation metrics : A systematic review of prediction accuracy for determining model performance in healthcare systems,” PLOS Digit. Heal., pp. 1–19, 2023, doi: 10.1371/journal.pdig.0000290.
[7] B. Więckowska and P. & Guzik, “Usmile likelihood evaluation provides robust threshold free assessment of binary classification models for balanced and imbalanced datasets,” Sci. Rep., 2026, doi: 10.1038/s41598-026-40545-z.
[8] D. Holst, K. Moenck, J. Koch, O. Schmedemann, and T. Schüppstuhl, “Transparent Reporting of AI in Systematic Literature Reviews : Development of the PRISMA-trAIce Checklist,” J. Med. Internet Res., vol. 4, pp. 1–8, 2025, doi: 10.2196/80247.
[9] V. Kumari, S. Gupta, and Y. Sharma, “A Multi-Modal Attentive Framework That Can Interpret Text ( MMAT ),” IEEE Access, vol. 13, no. April, pp. 121955–121969, 2025, doi: 10.1109/ACCESS.2025.3586900.
[10] M. Shahid, M. Ševrović, and M. Olyslagers, “Multi-Scale Spatio-Temporal Feature Aggregation for Road Safety Attributes Classification From Videos,” IEEE Access, vol. 14, no. May, 2026, doi: 10.1109/ACCESS.2026.3686110.
[11] H. Bechinia, D. Benmerzoug, and N. Khlifa, “Approach Based Lightweight Custom Convolutional Neural Network and Fine-Tuned MobileNet-V2 for ECG Arrhythmia Signals Classification,” IEEE Access, vol. 12, no. February, pp. 40827–40841, 2024, doi: 10.1109/ACCESS.2024.3378730.
[12] S. Jayakumar, V. Sounderajah, P. Normahani, L. Harling, S. R. Markar, and H. Ashra, “Quality assessment standards in arti fi cial intelligence diagnostic accuracy systematic reviews : a meta-research study,” pp. 1–13, 2022, doi: 10.1038/s41746-021-00544-y.
[13] D. Schwabe, K. Becker, and M. Seyferth, “The METRIC-framework for assessing data quality for trustworthy AI in medicine: A systematic review,” npj Digit. Med. 7, 203, 2024, doi: 10.1038/s41746-024-01196-4.
[14] M. Muhammad et al., “ADD-Net : An Effective Deep Learning Model for Early Detection of Alzheimer Disease in MRI Scans,” IEEE Access, vol. 10, no. September, pp. 96930–96951, 2022, doi: 10.1109/ACCESS.2022.3204395.
[15] A. F. A. Alshamrani, F. Saleh, and Z. Alshomrani, “Optimizing Breast Cancer Mammogram Classification Through a Dual Approach : A Deep Learning Framework Combining ResNet50 , SMOTE , and Fully Connected Layers for Balanced and Imbalanced Data,” IEEE Access, vol. 13, no. December 2024, pp. 4815–4826, 2025, doi: 10.1109/ACCESS.2024.3524633.
[16] A. S. Farhan, M. Khalid, and U. Manzoor, “Combined Oriented Data Augmentation Method for Brain MRI Images,” IEEE Access, vol. 13, no. December 2024, pp. 9981–9994, 2025, doi: 10.1109/ACCESS.2025.3526684.
[17] S. Akhtar, M. Hanif, A. Rashid, and K. Javed, “An Optimized Data and Model Centric Approach for Multi-Class Automated Urine Sediment Classification,” IEEE Access, vol. 12, no. May, 2024.
[18] M. S. Mohammed, A. Al-dhamari, W. Saeed, F. N. Al-aswadi, S. Abdulla, and M. Saleh, “Motion Pattern-Based Scene Classification Using Adaptive Synthetic Oversampling and Fully Connected Deep Neural Network,” IEEE Access, vol. 11, no. October, pp. 119659–119675, 2023, doi: 10.1109/ACCESS.2023.3327463.
[19] V. Hoang, X. Vo, and S. Member, “Categorical Weighting Domination for Imbalanced Classification With Skin Cancer in Intelligent Healthcare Systems,” vol. 11, no. August, 2023.
[20] W. Nazih and A. O. Aseeri, “Vision Transformer Model for Predicting the Severity of Diabetic Retinopathy in Fundus Photography-Based Retina Images,” IEEE Access, vol. 11, no. October, pp. 117546–117561, 2023, doi: 10.1109/ACCESS.2023.3326528.
[21] F. E. Karahan, “An End-to-End Deep Learning Framework for Crop Type and Phenological Stage Identification,” IEEE Access, vol. 14, no. June, pp. 97546–97558, 2026, doi: 10.1109/ACCESS.2026.3707086.
[22] T. Ali et al., “CAROTIDNet : A Novel Carotid Symptomatic / Asymptomatic Plaque Detection System Using CNN-Based Tangent Optimization Algorithm in B-Mode Ultrasound Images,” IEEE Access, vol. 12, no. June, pp. 73970–73979, 2024, doi: 10.1109/ACCESS.2024.3404023.
[23] V. Yamaçli, “Battery Usage-Profile Classification via Multimodal Vision Transformer With Feature-Level Fusion and Ensemble Method,” IEEE Access, vol. 13, no. September, 2025, doi: 10.1109/ACCESS.2025.3618776.
[24] X. Qin, F. M. Bui, H. A. H. Nguyen, S. Member, and Z. H. U. Han, “Learning From Limited and Imbalanced Medical Images With Finer Synthetic Images From GANs,” IEEE Access, vol. 10, no. September, pp. 91663–91677, 2022, doi: 10.1109/ACCESS.2022.3202560.
[25] Q. Su, H. Nuzly, A. Hamed, and M. A. Isa, “A GAN-Based Data Augmentation Method for Imbalanced Multi-Class Skin Lesion Classification,” IEEE Access, vol. 12, no. February, 2024.
[26] C. Venkatachalam, “A Dual-Branch Deep Learning Framework Combining Xception and ResNet for Accurate Lung and Colon Cancer Detection,” IEEE Access, vol. 13, no. July, pp. 131483–131497, 2025, doi: 10.1109/ACCESS.2025.3589390.
[27] M. Park, H. W. A. J. Song, and D. Kang, “Imbalanced Classification via Feature Dictionary-Based Minority Oversampling,” IEEE Access, vol. 10, pp. 34236–34245, 2022, doi: 10.1109/ACCESS.2022.3161510.
[28] A. Shafique, M. Siraj, S. A. Alsaif, T. Sadad, and B. Cheng, “Hyperspectral Imaging and Advanced Vision Transformers for Identifying Pure and Pesticide-Coated Apples,” IEEE Access, no. April, pp. 66405–66419, 2025.
[29] G. Mohammad, I. Alam, N. Tasnia, T. Biswas, and S. A. Tanim, “Real-Time Detection of Forest Fires Using FireNet-CNN and Explainable AI Techniques,” IEEE Access, vol. 13, no. February, 2025.
[30] M. Kovac, P. Macsik, V. Kurilova, Z. Fellner, and J. Pavlovicova, “Unconditional GAN-Based Generation of Stage-Specific Fundus Images for Augmenting Diabetic Retinopathy Screening,” IEEE Access, vol. 14, no. April, pp. 47800–47815, 2026, doi: 10.1109/ACCESS.2026.3675325.
[31] A. Asokan and B. R. Manju, “Lesion-Fidelity-Aware Synthetic Data Augmentation for Diabetic Retinopathy Grading Using Conditional GANs,” IEEE Access, vol. 14, no. April, pp. 69998–70010, 2026, doi: 10.1109/ACCESS.2026.3690867.
[32] V. Sreelekshmi, K. Pavithran, and J. J. Nair, “Unleashing the Power of Hierarchical Variational Autoencoder for Predicting Breast Cancer,” IEEE Access, vol. 12, pp. 195658–195670, 2024, doi: 10.1109/ACCESS.2024.3518612.
[33] H. I. Ahmad and M. Ali, “Classification of Psoriasis Using Vision Transformer : A Deep Learning-Based Approach,” IEEE Access, vol. 13, no. September, pp. 163885–163895, 2025, doi: 10.1109/ACCESS.2025.3607989.
[34] S. Chatterjee and J. Gil, “Early Detection of Multiclass Skin Lesions Using Transfer Learning-Based IncepX-Ensemble Model,” IEEE Access, vol. 12, no. August, pp. 113677–113693, 2024, doi: 10.1109/ACCESS.2024.3432904.
[35] R. D. W. I. Prayogo, N. U. R. Hamid, and H. Nambo, “Hybrid CNN-Based Transfer Learning Enhances Brain Tumor Classification on MRI Images,” IEEE Access, vol. 13, no. July, 2025.
[36] E. Silva et al., “Enhancing Diabetic Foot Ulcer Classification Through Fine-Tuned Multilevel CNN,” IEEE Access, vol. 13, no. July, 2025, doi: 10.1109/ACCESS.2025.3583111.
[37] L. Arora, K. K. Yadav, A. Malik, N. Kumar, U. Mamodiya, and R. Jain, “Cross-Modality Transfer Learning for Breast Cancer Detection Using Ultrasound-Pretrained Deep Network on Histopathology Images,” IEEE Access, vol. 14, no. May, pp. 76510–76525, 2026, doi: 10.1109/ACCESS.2026.3694268.
[38] A. V. Sri and M. K. Morampudi, “Privacy-Preserving Federated Learning for Retinal Disease Diagnosis Using Paillier Homomorphic Encryption With Multiple Encryption Keys,” IEEE Access, vol. 14, no. January, pp. 26391–26415, 2026, doi: 10.1109/ACCESS.2026.3664444.
[39] S. M. Vemuri, A. Gundrapally, Z. Xia, J. Kim, and K. K. Choi, “CB-DistillGrad : Class-Balanced Distillation and Gradient Conflict Resolution for Low,” IEEE Access, vol. 14, no. December 2025, pp. 3040–3061, 2026, doi: 10.1109/ACCESS.2025.3650071.
[40] R. F. Rocha, F. H. A. M. Neto, P. H. Ferreira, H. O. Paz, A. K. Suzuki, and F. Louzada, “Addressing Data Imbalance in Medical Imaging : A Simulation Study Using Ensemble Transfer Learning for Skin Cancer Detection,” IEEE Access, vol. 14, no. April, pp. 75259–75271, 2026, doi: 10.1109/ACCESS.2026.3692474.
[41] S. Ahmed, B. Hasan, T. Ahmed, R. K. Sony, and H. Kabir, “Less Is More : Lighter and Faster Deep Neural Architecture for Tomato Leaf Disease Classification,” IEEE Access, vol. 10, no. July, pp. 68868–68884, 2022, doi: 10.1109/ACCESS.2022.3187203.
[42] H. Ali, M. A. Khan, S. Member, G. Barb, and A. Silaghi, “AttFKAN-DEQ : Attention-Enhanced Fourier Kolmogorov-Arnold Networks Using Deep Equilibrium for Breast Cancer Histopathology Classification,” IEEE Access, vol. 14, no. May, pp. 91285–91302, 2026, doi: 10.1109/ACCESS.2026.3698647.
[43] J. Z. Diaz, M. A. L. I. Farooq, and P. Corcoran, “Automatic Inspection of Seal Integrity in Sterile Barrier Packaging : A Deep Learning Approach,” vol. 12, no. January, 2024.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Mochamad Arief Saputra
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License (Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) ) that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
