Image Enhancement and U-Net Based Brain Tumor Segmentation Using MRI

Shahd Hashim Hassan Abdalgadir; Mahmut Öztürk

Makale detayları

Image Enhancement and U-Net Based Brain Tumor Segmentation Using MRI

Yazarlar:

Shahd Hashim Hassan Abdalgadir | Mahmut Öztürk

Yayıncı:

Tekirdag Namik Kemal University

Ana Başlıklar:

Biyomedikal Mühendisliği (Diğer)

Dergi:

European Journal of Engineering and Applied Sciences Vol.8 Issue 1, pp.31-38

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001342941

pISSN:

2651-3412

Anahtar kelimeler:

Özet:

 Brain tumor is a very fatal health problem and unfortunately it is getting more common in modern society. Developing medical methods and technologies make possible to detect the disease earlier, slow down its progress and treat it. Early detection is very crucial for the success of treatment processes. Usage of image processing and artificial intelligence methods can help medics for early detection of the disease. In this study, a deep learning based enhanced image segmentation approach has been proposed to detect brain tumors. Segmentation was performed on the brain magnetic resonance MR images which were taken from a public dataset. Classical U-Net structure were employed at segmentation process because of its compatibility and success in medical image segmentation. Performance of the proposed model was increased with the help of image processing techniques used in pre- and post-processing stages. After using some image enhancement techniques as post-processing, a 0.89 of the dice coefficient, a 0.85 of the sensitivity and a 0.89 of the F-score were obtained.

 Brain tumor is a very fatal health problem and unfortunately it is getting more common in modern society. Developing medical methods and technologies make possible to detect the disease earlier, slow down its progress and treat it. Early detection is very crucial for the success of treatment processes. Usage of image processing and artificial intelligence methods can help medics for early detection of the disease. In this study, a deep learning based enhanced image segmentation approach has been proposed to detect brain tumors. Segmentation was performed on the brain magnetic resonance MR images which were taken from a public dataset. Classical U-Net structure were employed at segmentation process because of its compatibility and success in medical image segmentation. Performance of the proposed model was increased with the help of image processing techniques used in pre- and post-processing stages. After using some image enhancement techniques as post-processing, a 0.89 of the dice coefficient, a 0.85 of the sensitivity and a 0.89 of the F-score were obtained.

 Brain tumor is a very fatal health problem and unfortunately it is getting more common in modern society. Developing medical methods and technologies make possible to detect the disease earlier, slow down its progress and treat it. Early detection is very crucial for the success of treatment processes. Usage of image processing and artificial intelligence methods can help medics for early detection of the disease. In this study, a deep learning based enhanced image segmentation approach has been proposed to detect brain tumors. Segmentation was performed on the brain magnetic resonance MR images which were taken from a public dataset. Classical U-Net structure were employed at segmentation process because of its compatibility and success in medical image segmentation. Performance of the proposed model was increased with the help of image processing techniques used in pre- and post-processing stages. After using some image enhancement techniques as post-processing, a 0.89 of the dice coefficient, a 0.85 of the sensitivity and a 0.89 of the F-score were obtained.

 Beyin tümörü oldukça ölümcül bir sağlık sorunudur ve ne yazık ki modern toplumda giderek yaygınlaşmaktadır. Gelişen tıbbi yöntem ve teknolojiler, hastalığın daha erken tespit edilmesini, ilerlemesinin yavaşlatılmasını ve tedavi edilmesini mümkün kılmaktadır. Tedavi süreçlerinin başarısı için erken teşhis çok önemlidir. Görüntü işleme ve yapay zeka yöntemlerinin kullanılması, sağlık görevlilerinin hastalığın erken tespitine yardımcı olabilir. Bu çalışmada, beyin tümörlerinin tespiti için derin öğrenme tabanlı geliştirilmiş bir görüntü bölütleme yaklaşımı önerilmiştir. Halka açık bir veri setinden alınan beyin manyetik rezonans MR görüntüleri üzerinde bölütleme yapılmıştır. Medikal görüntü bölütlemesine uygunluğu ve başarısı nedeniyle bölütleme işleminde klasik U-Net yapısı kullanılmıştır. Ön işleme ve son işleme aşamalarında kullanılan görüntü işleme teknikleri yardımıyla önerilen modelin performansı arttırılmıştır. Son işlem olarak bazı görüntü iyileştirme teknikleri uygulandıktan sonra, 0,89 zar katsayısı, 0,85 hassasiyet ve 0,89 F-puanı elde edilmiştir.

Makale detayları

Image Enhancement and U-Net Based Brain Tumor Segmentation Using MRI

Yazarlar:

Shahd Hashim Hassan Abdalgadir | Mahmut Öztürk

Yayıncı:

Tekirdag Namik Kemal University

Ana Başlıklar:

Biyomedikal Mühendisliği (Diğer)

Dergi:

European Journal of Engineering and Applied Sciences Vol.8 Issue 1, pp.31-38

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001342941

pISSN:

2651-3412

MRI

Deep Learning

U-Net

Brain Tumor SegmentationMRG

Derin Öğrenme

U-Net

Beyin Tümörü Bölütlemesi.

Sayı

Makale detayları

Yazarlar: Shahd Hashim Hassan Abdalgadir | Mahmut Öztürk

Yayıncı: Tekirdag Namik Kemal University

Ana Başlıklar: Biyomedikal Mühendisliği (Diğer)

Dergi: European Journal of Engineering and Applied Sciences Vol.8 Issue 1, pp.31-38

Belge Tipi: Makale

Yayın Tarihi: 7/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001342941

pISSN: 2651-3412

Sayı

Yazarlar:

Shahd Hashim Hassan Abdalgadir | Mahmut Öztürk

Yayıncı:

Tekirdag Namik Kemal University

Ana Başlıklar:

Biyomedikal Mühendisliği (Diğer)

Dergi:

European Journal of Engineering and Applied Sciences Vol.8 Issue 1, pp.31-38

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001342941

pISSN:

2651-3412