Comparative Analysis of GAN-based Segmentation Models on Dental Panoramic Radiography Dataset

Hakan Öcal; Gürdal Altundağ

Makale detayları

Comparative Analysis of GAN-based Segmentation Models on Dental Panoramic Radiography Dataset

Yazarlar:

Hakan Öcal | Gürdal Altundağ

Yayıncı:

Fırat University

Ana Başlıklar:

Derin Öğrenme | Şaşırtmalı Makine Öğrenimi | Biyomedikal Görüntüleme

Dergi:

Fırat Üniversitesi Mühendislik Bilimleri Dergisi Vol.37 Issue 1, pp.523-532

Belge Tipi:

Makale

Yayın Tarihi:

3/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001334890

pISSN:

1308-9072

Anahtar kelimeler:

Özet:

 Especially in criminal investigations, the identification of the victim is essential. The branch of forensic medicine that uses the method of identification from the teeth of the victims is called forensic odontology. In forensic odontology, physical information about the individual can be obtained from the bone and enamel structure of the teeth. Panoramic, periapical, and cephalometric imaging techniques are the most commonly used in the odontological identification of the individual. Forensic odontology is increasingly recognized for its essential role in personal identification during mass disasters, sexual assault cases, and child abuse investigations. Deep learning algorithms have recently successfully detected dental disorders such as caries, periodontal bone loss, and apical lesions. Generative adversarial networks GAN models have mainly achieved high segmentation performance in medical images. In this study, GAN models were designed and comparatively analyzed using U-Net, Volumetric convolutional neural network V-Net, spatial and channel Squeeze-Excitation-based U-NetscSEU-Net, Transformer-based U-Net TransU-Net, and U-Net like pure Transformer SwinU-Net segmentation architectures which are widely used in the literature as generators. As a result of the comparative analyses, scSEU-Net-based GAN achieved the highest performance values with 0.8826 Thresholded DiceDSC, 0.7901 Thresholded Intersection over Union Thresh-IoU, 0.9805 Accuracy ACC, 0.9268 Precision PREC, and 0.9001 Recall REC.

 Especially in criminal investigations, the identification of the victim is essential. The branch of forensic medicine that uses the method of identification from the teeth of the victims is called forensic odontology. In forensic odontology, physical information about the individual can be obtained from the bone and enamel structure of the teeth. Panoramic, periapical, and cephalometric imaging techniques are the most commonly used in the odontological identification of the individual. Forensic odontology is increasingly recognized for its essential role in personal identification during mass disasters, sexual assault cases, and child abuse investigations. Deep learning algorithms have recently successfully detected dental disorders such as caries, periodontal bone loss, and apical lesions. Generative adversarial networks GAN models have mainly achieved high segmentation performance in medical images. In this study, GAN models were designed and comparatively analyzed using U-Net, Volumetric convolutional neural network V-Net, spatial and channel Squeeze-Excitation-based U-NetscSEU-Net, Transformer-based U-Net TransU-Net, and U-Net like pure Transformer SwinU-Net segmentation architectures which are widely used in the literature as generators. As a result of the comparative analyses, scSEU-Net-based GAN achieved the highest performance values with 0.8826 Thresholded DiceDSC, 0.7901 Thresholded Intersection over Union Thresh-IoU, 0.9805 Accuracy ACC, 0.9268 Precision PREC, and 0.9001 Recall REC.

 Especially in criminal investigations, the identification of the victim is essential. The branch of forensic medicine that uses the method of identification from the teeth of the victims is called forensic odontology. In forensic odontology, physical information about the individual can be obtained from the bone and enamel structure of the teeth. Panoramic, periapical, and cephalometric imaging techniques are the most commonly used in the odontological identification of the individual. Forensic odontology is increasingly recognized for its essential role in personal identification during mass disasters, sexual assault cases, and child abuse investigations. Deep learning algorithms have recently successfully detected dental disorders such as caries, periodontal bone loss, and apical lesions. Generative adversarial networks GAN models have mainly achieved high segmentation performance in medical images. In this study, GAN models were designed and comparatively analyzed using U-Net, Volumetric convolutional neural network V-Net, spatial and channel Squeeze-Excitation-based U-NetscSEU-Net, Transformer-based U-Net TransU-Net, and U-Net like pure Transformer SwinU-Net segmentation architectures which are widely used in the literature as generators. As a result of the comparative analyses, scSEU-Net-based GAN achieved the highest performance values with 0.8826 Thresholded DiceDSC, 0.7901 Thresholded Intersection over Union Thresh-IoU, 0.9805 Accuracy ACC, 0.9268 Precision PREC, and 0.9001 Recall REC.

 Özellikle ceza soruşturmalarında mağdurun kimliği esastır. Mağdurların dişlerinden kimlik tespiti yöntemini kullanan adli tıp dalına adli odontoloji denir. Adli odontolojide, bireye ait fiziksel bilgiler dişlerin kemik ve mine yapısından elde edilebilir. Bireyin odontolojik kimlik tespitinde en sık kullanılanlar panoramik, periapikal ve sefalometrik görüntüleme teknikleridir. Adli odontoloji, kitlesel felaketler, cinsel saldırı vakaları ve çocuk istismarı soruşturmaları sırasında kişisel kimlik tespitindeki temel rolüyle giderek daha fazla tanınmaktadır. Derin öğrenme algoritmaları son zamanlarda çürük, periodontal kemik kaybı ve apikal lezyonlar gibi diş bozukluklarını başarıyla tespit etmiştir. Üretken Çekişmeli Ağlar GAN modelleri çoğunlukla medikal görüntülerde yüksek segmentasyon performansı elde etmiştir. Bu çalışmada, literatürde üreteç olarak yaygın olarak kullanılan U-Net, Volumetrik Evrişimli Sinir Ağı V-Net, Uzaysal ve Kanal Sıkıştırma-Uyartım Tabanlı U-Net scSEU-Net, Transformatör Tabanlı U-Net TransU-Net ve U-Net benzeri Saf Transformatör SwinU-Net segmentasyon mimarileri kullanılarak GAN modelleri tasarlanmış ve karşılaştırmalı olarak analiz edilmiştir. Karşılaştırmalı analizler sonucunda scSEU-Net tabanlı GAN, 0,8826 Eşikli Zar DSC, 0,7901 Eşikli Birleşim Kesişimi Thresh-IoU, 0,9805 Doğruluk ACC, 0,9268 Hassasiyet PREC ve 0,9001 Geri Çağırma REC değerleriyle en yüksek performans değerlerine ulaşmıştır.

Makale detayları

Comparative Analysis of GAN-based Segmentation Models on Dental Panoramic Radiography Dataset

Yazarlar:

Hakan Öcal | Gürdal Altundağ

Yayıncı:

Fırat University

Ana Başlıklar:

Derin Öğrenme | Şaşırtmalı Makine Öğrenimi | Biyomedikal Görüntüleme

Dergi:

Fırat Üniversitesi Mühendislik Bilimleri Dergisi Vol.37 Issue 1, pp.523-532

Belge Tipi:

Makale

Yayın Tarihi:

3/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001334890

pISSN:

1308-9072

Deep Learning

Dental panoramic radiography segmentation

Generative adversarial networks

U-shaped Segmentation models

V-Net.Anahtar kelimeler: Derin Öğrenme

Dental panoramik radyografi segmentasyonu

Üretken düşmanca ağlar

U şeklinde segmentasyon modelleri

V-Net.

Sayı

Makale detayları

Yazarlar: Hakan Öcal | Gürdal Altundağ

Yayıncı: Fırat University

Ana Başlıklar: Derin Öğrenme | Şaşırtmalı Makine Öğrenimi | Biyomedikal Görüntüleme

Dergi: Fırat Üniversitesi Mühendislik Bilimleri Dergisi Vol.37 Issue 1, pp.523-532

Belge Tipi: Makale

Yayın Tarihi: 3/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001334890

pISSN: 1308-9072

Sayı

Yazarlar:

Hakan Öcal | Gürdal Altundağ

Yayıncı:

Fırat University

Ana Başlıklar:

Derin Öğrenme | Şaşırtmalı Makine Öğrenimi | Biyomedikal Görüntüleme

Dergi:

Fırat Üniversitesi Mühendislik Bilimleri Dergisi Vol.37 Issue 1, pp.523-532

Belge Tipi:

Makale

Yayın Tarihi:

3/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001334890

pISSN:

1308-9072