In today’s world, there has been a significant increase in the diversity of data sources and the volume of data. This situation especially necessitates the use of technologies such as deep learning in data processing. This study thoroughly examines the processing of computed tomography (CT) images with deep learning models and their role in the diagnosis of brain hemorrhages, proposing an innovative deep learning-based model for accurately detecting and segmenting brain hemorrhages. This model combines the architectures of Mask Scoring R-CNN and EfficientNet-B2, offering an effective approach for the detection and classification of brain hemorrhages. MS R-CNN is used to detect potential hemorrhage areas in CT images, while the EfficientNet-B2 architecture serves a classification function to determine whether these areas indeed contain hemorrhages. Thus, the model offers a two-stage verification process that enhances accuracy and precision. The performance of the model has been evaluated under patient-based and random partitioning techniques using by employing two distinct datasets: an open-access and a private. In patient-based evaluation, the proposed model has an accuracy of %91.59 on open dataset and an accuracy of %90.46 on private dataset for SDH hemorrhages. In the random partitioning method, the model’s accuracy rate has risen to %94.30 on open dataset and %97.33 on private dataset. Compared with similar studies in the literature, these results demonstrate that the model has a high level of accuracy and reliability. This study highlights the potential and importance of AI-supported methods in the detection of brain hemorrhages and provides a solid foundation for future work in this area. Additionally, the results obtained from an open dataset by the proposed model offer a realistic and comparable reference for future work in this field.The results obtained from a second data set also clearly demonstrate the validity of the model.
Teknolojinin gelişmesiyle birlikte yapay zekâ temelli uygulamalar pek çok alanda destek amaçlı kullanılmaktadır. Sağlık sektörü de bu tür uygulamaların yaygın kullanıldığı alanlardan bir tanesidir. Sağlık sektöründe teknolojik gelişime bağlı olarak meydana gelen bilgi artışı beraberinde radyolojik değerlendirmede uzmanlık gereğini doğurmuştur. Yoğun çalışma saatleri, sağlık kurumlarında her branştan uzmana ulaşılamaması ve özellikle acil patolojilerde erken teşhisin önemi göz önünde bulundurulduğunda hekimlere teşhis sürecinde destek olacak uygulamalara olan ihtiyacın önemi anlaşılmaktadır. Çalışma kapsamında Bilgisayarlı Tomografi (BT) görüntüleri kullanılarak beyin kanamalarının tespitini gerçekleştirmek amacıyla güncel derin öğrenme yöntemlerinden Görsel Geometri Grubu (VGG), Artık Sinir Ağı (ResNet) ve EfficientNet mimarileri yine güncel bir veri kümesi olan PhysioNet’e uygulanmıştır. Modeller doğruluk, kesinlik, hassasiyet ve F1 skor metrikleri kullanılarak hem kendi aralarında hem de literatürdeki çalışmalarla karşılaştırılmıştır. Gerçekleştirilen çalışma ile veri kümesine uygun model seçiminin önemi güncel modeller üzerinden ortaya konulmuştur. EfficientNet-B2 modelinin başarısı hem literatürdeki çalışmalardan hem de makale kapsamında değerlendirilen modellerden yüksek olmuştur. Elde edilen sonuçlar güncel derin öğrenme modellerinin, beyin kanaması teşhisine yardımcı olabilecek potansiyelde olduğunu göstermiştir. Çalışma acil servislerin yükünü çeken pratisyen hekimleri en azından beyin kanamasının varlığı konusunda uyarıp kanama durumunun gözden kaçmamasını sağlaması ve erken teşhisi açısından önem arz etmektedir.
1 January 2023 in Turkish Journal of Trauma and Emergency Surgery
The article was planned to make the first evaluation in terms of acute subdural hemorrhages, thinking that it can help in appropriate pathologies by tomography interpretation with the artificial intelligence (AI) method, at least in a way to quickly warn the responsible doctor. A two-level AI-based hybrid method was developed. The proposed model uses the mask-region convolutional neural network (Mask R-CNN) technique, which is a deep learning model, in the hemorrhagic region's mask generation stage, and a problem-specific, optimized support vector machines (SVM) technique which is a machine learning model in the binary classification stage. Furthermore, the bee colony algorithm was used for the optimization of SVM algorithms' parameters. In the first stage, the mean average precision (mAP) value was obtained as 0.754 when the intercept over union (IOU) value was taken as 0.5 with the Mask R-CNN architecture used. At the same time, when a 5-fold cross-validation was applied, the mAP value was obtained 0.736. With the hyperparameter optimization for both Mask R-CNN and the SVM algorithm, the accuracy of the two-level classification process was obtained as 96.36%. Furthermore, final false-negative rate and false-positive rate values were obtained as 6.20%, and 2.57%, respectively. With the proposed model, both the detection of hemorrhage and the presentation of the suspicious area to the physician were performed more successfully on two dimensional (2D) images with low cost and high accuracy compared to similar studies and today's interpretations with telemedicine techniques.