<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.4 20241031//EN"
        "https://jats.nlm.nih.gov/publishing/1.4/JATS-journalpublishing1-4.dtd">
<article  article-type="research-article"        dtd-version="1.4">
            <front>

                <journal-meta>
                                                                <journal-id>ij3dptdi</journal-id>
            <journal-title-group>
                                                                                    <journal-title>International Journal of 3D Printing Technologies and Digital Industry</journal-title>
            </journal-title-group>
                            <issn pub-type="ppub">2602-3350</issn>
                                        <issn pub-type="epub">2602-3350</issn>
                                                                                            <publisher>
                    <publisher-name>Kerim ÇETİNKAYA</publisher-name>
                </publisher>
                    </journal-meta>
                <article-meta>
                                        <article-id pub-id-type="doi">10.46519/ij3dptdi.1804803</article-id>
                                                                <article-categories>
                                            <subj-group  xml:lang="en">
                                                            <subject>Machine Learning (Other)</subject>
                                                    </subj-group>
                                            <subj-group  xml:lang="tr">
                                                            <subject>Makine Öğrenme (Diğer)</subject>
                                                    </subj-group>
                                    </article-categories>
                                                                                                                                                        <title-group>
                                                                                                                                                            <trans-title-group xml:lang="en">
                                    <trans-title>IMPROVING DEEP LEARNING CLASSIFICATION PERFORMANCE WITH GAN-BASED DATA AUGMENTATION IN IMBALANCED SKIN LESION DATASETS</trans-title>
                                </trans-title-group>
                                                                                                                                                                                                <article-title>DENGESİZ CİLT LEZYONU VERİ SETLERİNDE GAN TABANLI VERİ ARTIRIMI İLE DERİN ÖĞRENME SINIFLANDIRMA BAŞARIMININ İYİLEŞTİRİLMESİ</article-title>
                                                                                                    </title-group>
            
                                                    <contrib-group content-type="authors">
                                                                        <contrib contrib-type="author">
                                                                    <contrib-id contrib-id-type="orcid">
                                        https://orcid.org/0009-0005-4080-1245</contrib-id>
                                                                <name>
                                    <surname>Özdemir</surname>
                                    <given-names>Serkan</given-names>
                                </name>
                                                                    <aff>BİLECİK ŞEYH EDEBALİ ÜNİVERSİTESİ, LİSANSÜSTÜ EĞİTİM ENSTİTÜSÜ</aff>
                                                            </contrib>
                                                    <contrib contrib-type="author">
                                                                    <contrib-id contrib-id-type="orcid">
                                        https://orcid.org/0000-0003-0274-6175</contrib-id>
                                                                <name>
                                    <surname>Ceyhan</surname>
                                    <given-names>Salim</given-names>
                                </name>
                                                                    <aff>BİLECİK ŞEYH EDEBALİ ÜNİVERSİTESİ, MÜHENDİSLİK FAKÜLTESİ</aff>
                                                            </contrib>
                                                                                </contrib-group>
                        
                                        <pub-date pub-type="pub" iso-8601-date="20260430">
                    <day>04</day>
                    <month>30</month>
                    <year>2026</year>
                </pub-date>
                                        <volume>10</volume>
                                        <issue>1</issue>
                                        <fpage>84</fpage>
                                        <lpage>93</lpage>
                        
                        <history>
                                    <date date-type="received" iso-8601-date="20251016">
                        <day>10</day>
                        <month>16</month>
                        <year>2025</year>
                    </date>
                                                    <date date-type="accepted" iso-8601-date="20260113">
                        <day>01</day>
                        <month>13</month>
                        <year>2026</year>
                    </date>
                            </history>
                                        <permissions>
                    <copyright-statement>Copyright © 2017, International Journal of 3D Printing Technologies and Digital Industry</copyright-statement>
                    <copyright-year>2017</copyright-year>
                    <copyright-holder>International Journal of 3D Printing Technologies and Digital Industry</copyright-holder>
                </permissions>
            
                                                                                                                        <trans-abstract xml:lang="en">
                            <p>Imbalanced skin lesion datasets cause deep learning models to develop a bias towards the majority class, thereby compromising the detection performance of rare but vital malignant lesions. This study systematically investigates the impact of a hybrid data balancing strategy, combining Deep Convolutional Generative Adversarial Networks (DCGAN) and classical data augmentation techniques, on classification performance to address the class imbalance problem. Within the scope of the study, DenseNet121, ResNet50, VGG16, EfficientNetB0, and ConvNeXt architectures were compared on the original imbalanced dataset and a balanced dataset where synthetic images were added exclusively to the training set to prevent data leakage. Experimental results revealed that although the models exhibited high accuracy (up to 89%) on the imbalanced dataset, an &quot;Accuracy Paradox&quot; occurred, and minority classes (dermatofibroma, vascular lesions) could not be detected. In models trained with the proposed hybrid balancing method, while a marginal change was observed in overall accuracy, a significant average increase of 21% was achieved in Macro-F1 scores, which represent inter-class balance. On the balanced dataset, DenseNet121 (Macro-F1=0.803; Accuracy=0.868) and ConvNeXt models displayed the most stable performance, while the detection success of rare classes increased dramatically. These findings demonstrate that GAN-supported hybrid data augmentation is an effective strategy for enhancing the reliability and generalization capability of clinical decision support systems operating with limited data.</p></trans-abstract>
                                                                                                                                    <abstract><p>Dengesiz cilt lezyonu veri setleri, derin öğrenme modellerinin çoğunluk sınıfına karşı yanlılık geliştirmesine neden olarak, nadir fakat hayati risk taşıyan malign lezyonların tespit başarısını düşürmektedir. Bu çalışma, sınıf dengesizliği problemini ele almak amacıyla, Derin Evrişimli Üretici Karşıt Ağlar (DCGAN) ve klasik veri artırma tekniklerini birleştiren hibrit bir veri dengeleme stratejisinin sınıflandırma performansına etkisini sistematik olarak incelemektedir. Çalışma kapsamında DenseNet121, ResNet50, VGG16, EfficientNetB0 ve ConvNeXt mimarileri; orijinal dengesiz veri seti ve veri sızıntısını önlemek amacıyla yalnızca eğitim kümesine sentetik görüntülerin eklendiği dengelenmiş veri seti üzerinde karşılaştırılmıştır. Deneysel sonuçlar, dengesiz veri setinde modellerin yüksek doğruluk (%89&#039;a varan) göstermesine rağmen &quot;Doğruluk Paradoksu&quot; yaşandığını ve azınlık sınıflarının (dermatofibroma, vasküler lezyonlar) tespit edilemediğini ortaya koymuştur. Önerilen hibrit dengeleme yöntemiyle eğitilen modellerde ise, genel doğrulukta marjinal bir değişim gözlemlenirken, sınıflar arası dengeyi ifade eden Macro-F1 skorlarında ortalama %21 oranında belirgin bir artış sağlanmıştır. Dengelenmiş veri setinde en kararlı başarıyı DenseNet121 (Macro-F1=0.803; Doğruluk=0.868) ve ConvNeXt modelleri sergilerken, nadir sınıfların tespit başarısı dramatik şekilde yükselmiştir. Bu bulgular, GAN destekli hibrit veri artırımının, sınırlı veri ile çalışan klinik karar destek sistemlerinin güvenilirliğini ve genelleme yeteneğini artırmada etkili bir strateji olduğunu kanıtlamaktadır.</p></abstract>
                                                            
            
                                                                                                                    <kwd-group>
                                                    <kwd>Cilt Kanseri</kwd>
                                                    <kwd>  Dermoskopi</kwd>
                                                    <kwd>  Derin Öğrenme</kwd>
                                                    <kwd>  GAN Tabanlı Veri Artırımı</kwd>
                                                    <kwd>  Sınıf Dengesizliği</kwd>
                                                    <kwd>  HAM10000</kwd>
                                                    <kwd>  Hibrit Dengeleme.</kwd>
                                            </kwd-group>
                            
                                                                            <kwd-group xml:lang="en">
                                                    <kwd>Skin Cancer</kwd>
                                                    <kwd>  Dermoscopy</kwd>
                                                    <kwd>  Deep Learning</kwd>
                                                    <kwd>  GAN-Based Data Augmentation</kwd>
                                                    <kwd>  Class Imbalance</kwd>
                                                    <kwd>  HAM10000</kwd>
                                                    <kwd>  Hybrid Balancing.</kwd>
                                            </kwd-group>
                                                                                                                                        </article-meta>
    </front>
    <back>
                            <ref-list>
                                    <ref id="ref1">
                        <label>1</label>
                        <mixed-citation publication-type="journal">1. Kanca, E., Ayas, S., “Dengesiz veri kümelerinde topluluk yöntemlerine dayalı melanom sınıflandırılması”, Karadeniz Fen Bilimleri Dergisi, Cilt 12, Sayı 2, Sayfa 1122-1137, 2022.</mixed-citation>
                    </ref>
                                    <ref id="ref2">
                        <label>2</label>
                        <mixed-citation publication-type="journal">2. Özcan, Ö., Karaaltun, M., “Görüntü artırma tekniklerinin cilt kanseri türleri üzerinde evrişimsel sinir ağları ile sınıflandırma başarılarının karşılaştırılması”, Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi, Cilt 12, Sayı 4, Sayfa 1-1, 2025.</mixed-citation>
                    </ref>
                                    <ref id="ref3">
                        <label>3</label>
                        <mixed-citation publication-type="journal">3. Sarı, A., Nizam, A., Aydın, M., “Implementing of transfer learning method in the diagnosis of skin diseases with convolutional neural networks”, 7th International Conference on Computer Science and Engineering (UBMK), Sayfa 142-147, Diyarbakır, 2022.</mixed-citation>
                    </ref>
                                    <ref id="ref4">
                        <label>4</label>
                        <mixed-citation publication-type="journal">4. Kaplan, A., Güldogan, E., Arslan, A.K., “Prediction of melanoma from dermoscopic images using deep learning-based artificial intelligence techniques”, International Artificial Intelligence and Data Processing Symposium (IDAP), Sayfa 1-5, Malatya, 2019.</mixed-citation>
                    </ref>
                                    <ref id="ref5">
                        <label>5</label>
                        <mixed-citation publication-type="journal">5. Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q., “Densely connected convolutional networks”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , Sayfa 4700-4708, Honolulu, 2017.</mixed-citation>
                    </ref>
                                    <ref id="ref6">
                        <label>6</label>
                        <mixed-citation publication-type="journal">6. He, K., Zhang, X., Ren, S., Sun, J., “Deep residual learning for image recognition”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Sayfa 770-778, Las Vegas, 2016.</mixed-citation>
                    </ref>
                                    <ref id="ref7">
                        <label>7</label>
                        <mixed-citation publication-type="journal">7. Simonyan, K., Zisserman, A., “Very deep convolutional networks for large-scale image recognition”, ArXiv Preprint, arXiv:1409.1556, https://arxiv.org/abs/1409.1556, May 22, 2025, 2014.</mixed-citation>
                    </ref>
                                    <ref id="ref8">
                        <label>8</label>
                        <mixed-citation publication-type="journal">8. Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.A., “Convnet for the 2020s”, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Sayfa 11976-11986, New Orleans, 2022.</mixed-citation>
                    </ref>
                                    <ref id="ref9">
                        <label>9</label>
                        <mixed-citation publication-type="journal">9. Hoang, V.T., Jo, K.H., “Practical analysis on architecture of EfficientNet”, 14th International Conference on Human System Interaction (HSI), Sayfa 1-4, Gdańsk, 2021.</mixed-citation>
                    </ref>
                                    <ref id="ref10">
                        <label>10</label>
                        <mixed-citation publication-type="journal">10. Ergün, E., Kılıç, K., “Derin öğrenme ile artırılmış görüntü seti üzerinden cilt kanseri tespiti”, Black Sea Journal of Engineering and Science, Cilt. 4, Sayı 4, Sayfa 192-200, 2021.</mixed-citation>
                    </ref>
                                    <ref id="ref11">
                        <label>11</label>
                        <mixed-citation publication-type="journal">11. Aruk, İ., Toprak, A.N., “Cilt kanseri tanısı için farklı evrişimsel sinir ağı modellerinin karşılaştırılması”, Journal of the Institute of Science and Technology, Cilt 15, Sayı 1, Sayfa 25-38, 2025.</mixed-citation>
                    </ref>
                                    <ref id="ref12">
                        <label>12</label>
                        <mixed-citation publication-type="journal">12. Arora, G., Dubey, A.K., Jaffery, Z.A., Rocha, A.A., “Comparative study of fourteen deep learning networks for multi skin lesion classification (MSLC) on unbalanced data”, Neural Computing and Applications, Cilt 35, Sayı 11, Sayfa 7989-8015, 2022.</mixed-citation>
                    </ref>
                                    <ref id="ref13">
                        <label>13</label>
                        <mixed-citation publication-type="journal">13. Chatterjee, S., Gil, J.M., Byun, Y.C., “Early detection of multiclass skin lesions using transfer learning-based IncepX-ensemble model”, IEEE Access, Cilt 12, Sayfa 113677–113693, 2024.</mixed-citation>
                    </ref>
                                    <ref id="ref14">
                        <label>14</label>
                        <mixed-citation publication-type="journal">14. Selvaraj, K.M., Gnanagurusubbiah, S., Roy, R.R.R., Balu, S., “Enhancing skin lesion classification with advanced deep learning ensemble models: A path towards accurate medical diagnostics”, Current Problems in Cancer, Cilt 49, Sayfa 101077, 2024.</mixed-citation>
                    </ref>
                                    <ref id="ref15">
                        <label>15</label>
                        <mixed-citation publication-type="journal">15. Gayatri, E., Aarthy, S.L., “Reduction of overfitting on the highly imbalanced ISIC-2019 skin dataset using deep learning frameworks”, Journal of X-Ray Science and Technology, Cilt 32, Sayı 1, Sayfa 53-68, 2024.</mixed-citation>
                    </ref>
                                    <ref id="ref16">
                        <label>16</label>
                        <mixed-citation publication-type="journal">16. Nguyen, V.D., Bui, N.D., Do, H.K., “Skin lesion classification on imbalanced data using deep learning with soft attention”, Sensors, Cilt 22, Sayı 19, Sayfa 7530, 2022.</mixed-citation>
                    </ref>
                                    <ref id="ref17">
                        <label>17</label>
                        <mixed-citation publication-type="journal">17. Yao, P., Shen, S., Xu, M., Liu, P., Zhang, F., Xing, J., Shao, P., Kaffenberger, B., Xu, R.X., “Single model deep learning on imbalanced small datasets for skin lesion classification”, IEEE Transactions on Medical Imaging, Cilt 41, Sayı 5, Sayfa 1242-1254, 2021.</mixed-citation>
                    </ref>
                                    <ref id="ref18">
                        <label>18</label>
                        <mixed-citation publication-type="journal">18. Su, Q., Hamed, H.N.A., Isa, M.A., Hao, X., Dai, X.A., “GAN-based data augmentation method for imbalanced multi-class skin lesion classification”, IEEE Access, Cilt 12, Sayfa 16498-16513, 2024.</mixed-citation>
                    </ref>
                                    <ref id="ref19">
                        <label>19</label>
                        <mixed-citation publication-type="journal">19. P. Tschandl, C. Rosendahl, and H. Kittler, “The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions”, Scientific Data, Cilt 5, Sayfa 180161, 2018.</mixed-citation>
                    </ref>
                            </ref-list>
                    </back>
    </article>
