DAC: Differentiable Auto-Cropping in Deep Learning

Ahmet Esad Top; Mustafa Yeniad; Mahmut Sertaç Özdoğan; Fatih Nar

Makale detayları

DAC: Differentiable Auto-Cropping in Deep Learning

Yazarlar:

Ahmet Esad Top | Mustafa Yeniad | Mahmut Sertaç Özdoğan | Fatih Nar

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Bilgisayar Yazılımı

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.24 Issue 6, pp.1382-1394

Belge Tipi:

Makale

Yayın Tarihi:

2/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001330234

pISSN:

2149-3367

Anahtar kelimeler:

Özet:

 Auto-cropping, the process of automatically adjusting the boundaries of an image to focus on the region of interest, is crucial to improving the diagnostic quality of dental panoramic radiographs. Its importance lies in its ability to standardize the size of different input images with minimal loss of information, thus ensuring consistency and improving the performance of subsequent image-processing tasks. Despite the widespread use of CNNs in many studies, research on auto-cropping for different-sized images remains limited. This study aims to explore the potential of differentiable auto-cropping in dental panoramic radiographs. A unique dataset of 20,973 dental panoramic radiographs, mostly with a resolution of 28361536 or close, divided into five classes by 3 dentists, was used, which is the same dataset from the previous study Top et al. 2023. ResNet-101 model, which was the most successful network for the dataset Top et al. 2023, was used for the evaluation. To reduce variance, the model was evaluated using 10-fold cross-validation for both non-auto-cropped and auto-cropped trainings. Data augmentation was also used to produce more accurate and robust results. For auto-cropped training, it was adjusted to be much less effective than the non-auto-cropped one. Accuracy was improved by 1.8, from 92.7 to 94.5, thanks to the proposed auto-crop optimization developed to reduce dataset-related issues. Its macro-average AUC was also raised from 0.989 to 0.993. The proposed auto-crop optimization can be implemented as a trainable network layer in an end-to-end CNN and can be used for other problems as well. Increasing the accuracy from 92.7 to 94.5 is a very challenging task due to diminishing returns, as there is little room for improvement. The results show the potential of the proposed differentiable auto-crop algorithm and encourages its use in different fields.

 Auto-cropping, the process of automatically adjusting the boundaries of an image to focus on the region of interest, is crucial to improving the diagnostic quality of dental panoramic radiographs. Its importance lies in its ability to standardize the size of different input images with minimal loss of information, thus ensuring consistency and improving the performance of subsequent image-processing tasks. Despite the widespread use of CNNs in many studies, research on auto-cropping for different-sized images remains limited. This study aims to explore the potential of differentiable auto-cropping in dental panoramic radiographs. A unique dataset of 20,973 dental panoramic radiographs, mostly with a resolution of 28361536 or close, divided into five classes by 3 dentists, was used, which is the same dataset from the previous study Top et al. 2023. ResNet-101 model, which was the most successful network for the dataset Top et al. 2023, was used for the evaluation. To reduce variance, the model was evaluated using 10-fold cross-validation for both non-auto-cropped and auto-cropped trainings. Data augmentation was also used to produce more accurate and robust results. For auto-cropped training, it was adjusted to be much less effective than the non-auto-cropped one. Accuracy was improved by 1.8, from 92.7 to 94.5, thanks to the proposed auto-crop optimization developed to reduce dataset-related issues. Its macro-average AUC was also raised from 0.989 to 0.993. The proposed auto-crop optimization can be implemented as a trainable network layer in an end-to-end CNN and can be used for other problems as well. Increasing the accuracy from 92.7 to 94.5 is a very challenging task due to diminishing returns, as there is little room for improvement. The results show the potential of the proposed differentiable auto-crop algorithm and encourages its use in different fields.

 Auto-cropping, the process of automatically adjusting the boundaries of an image to focus on the region of interest, is crucial to improving the diagnostic quality of dental panoramic radiographs. Its importance lies in its ability to standardize the size of different input images with minimal loss of information, thus ensuring consistency and improving the performance of subsequent image-processing tasks. Despite the widespread use of CNNs in many studies, research on auto-cropping for different-sized images remains limited. This study aims to explore the potential of differentiable auto-cropping in dental panoramic radiographs. A unique dataset of 20,973 dental panoramic radiographs, mostly with a resolution of 28361536 or close, divided into five classes by 3 dentists, was used, which is the same dataset from the previous study Top et al. 2023. ResNet-101 model, which was the most successful network for the dataset Top et al. 2023, was used for the evaluation. To reduce variance, the model was evaluated using 10-fold cross-validation for both non-auto-cropped and auto-cropped trainings. Data augmentation was also used to produce more accurate and robust results. For auto-cropped training, it was adjusted to be much less effective than the non-auto-cropped one. Accuracy was improved by 1.8, from 92.7 to 94.5, thanks to the proposed auto-crop optimization developed to reduce dataset-related issues. Its macro-average AUC was also raised from 0.989 to 0.993. The proposed auto-crop optimization can be implemented as a trainable network layer in an end-to-end CNN and can be used for other problems as well. Increasing the accuracy from 92.7 to 94.5 is a very challenging task due to diminishing returns, as there is little room for improvement. The results show the potential of the proposed differentiable auto-crop algorithm and encourages its use in different fields.

 Bir görüntünün sınırlarını ilgi alanına odaklanacak şekilde otomatik olarak ayarlama işlemi olan oto-kırpma, panoramik diş radyografilerinin teşhis kalitesinin iyileştirilmesi açısından çok önemlidir. Önemi, minimum bilgi kaybıyla farklı girdi görüntülerinin boyutunu standartlaştırma yeteneğinde yatmaktadır, böylece tutarlılık sağlanmakta ve sonraki görüntü işleme görevlerinin performansı iyileştirilmektedir. Çalışmaların birçoğunda CNNler yaygın olarak kullanılmasına rağmen, farklı boyutlardaki görüntüler için oto-kırpma kullanan araştırmalar sınırlı kalmaktadır. Bu çalışma, panoramik diş radyografilerinde türevlenebilir oto-kırpma kullanmanın potansiyelini araştırmayı amaçlamaktadır. Çalışmada, çoğunlukla 28361536 veya buna yakın çözünürlüklü, 3 diş hekimi tarafından beş farklı sınıfa bölünmüş 20.973 panoramik diş radyografisinden oluşan benzersiz bir veri kümesi kullanıldı; bu, önceki çalışmadaki aynı veri kümesidir Top et al. 2023. Değerlendirme için bu veri kümesine en başarılı sonucu veren ResNet-101 modeli kullanıldı Top et al. 2023. Varyansı azaltmak için, hem oto-kırpma olan hem de oto-kırpma olmayan eğitimlere 10 kat çapraz doğrulama kullanılarak model değerlendirildi. Daha doğru ve sağlam sonuçlara ulaşmak için veri artırma yöntemi de kullanıldı. Veri artırma, oto-kırpma olan eğitim için, oto-kırpma olmayan eğitime göre çok daha az etkili olacak şekilde ayarlandı. Veri kümesiyle ilgili sorunları azaltmak için geliştirilen önerilen oto-kırpma optimizasyonu sayesinde doğruluk 1,8 artarak 92,7den 94,5e çıktı. Makro ortalama AUCsi de 0,989dan 0,993e yükseldi. Önerilen oto-kırpma optimizasyonu, uçtan uca bir CNNde eğitilebilir bir ağ katmanı olarak uygulanabilir ve diğer problemler için de kullanılabilir. Doğruluğu 92,7den 94,5e çıkarmak, iyileştirme için çok az alan kaldığından, azalan faydalar kanununa da bağlı olarak çok zorlu bir iştir. Sonuçlar, önerilen türevlenebilir oto-kırpma algoritmasının potansiyelini göstermekte ve farklı alanlarda kullanımını teşvik etmektedir.

Makale detayları

DAC: Differentiable Auto-Cropping in Deep Learning

Yazarlar:

Ahmet Esad Top | Mustafa Yeniad | Mahmut Sertaç Özdoğan | Fatih Nar

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Bilgisayar Yazılımı

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.24 Issue 6, pp.1382-1394

Belge Tipi:

Makale

Yayın Tarihi:

2/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001330234

pISSN:

2149-3367

Computer aided diagnosis

Convolutional Neural Networks

Differentiable cropping

Gradient ascent

Panoramic radiographBilgisayar destekli teşhis

CNN

Türevlenebilir kırpma

Gradyan yükselme

Panoramik radyografi

Sayı

Makale detayları

Yazarlar: Ahmet Esad Top | Mustafa Yeniad | Mahmut Sertaç Özdoğan | Fatih Nar

Yayıncı: Afyon Kocatepe University

Ana Başlıklar: Bilgisayar Yazılımı

Dergi: Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.24 Issue 6, pp.1382-1394

Belge Tipi: Makale

Yayın Tarihi: 2/2024

Belge Dili: İngilizce

Document ID (Org No.): 2001330234

pISSN: 2149-3367

Sayı

Yazarlar:

Ahmet Esad Top | Mustafa Yeniad | Mahmut Sertaç Özdoğan | Fatih Nar

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Bilgisayar Yazılımı

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.24 Issue 6, pp.1382-1394

Belge Tipi:

Makale

Yayın Tarihi:

2/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001330234

pISSN:

2149-3367