Semantic Segmentation of Intracranial Hemorrhages in Head CT Scans

Qiu, Yuhang; Chang, Chia Shuo; Yan, Jiun Lin; Ko, Li-Wei; Chang, Tian-Sheuan

doi:10.1109/icsess47205.2019.9040733

Cited by 64 publications

(37 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…InceptionV3 (Szegedy 2015) and ResNet50 (He 2016) are two typical deep CNNs for image classification, each of which has its unique advantages and are proved in many studies [10] [18]. In this paper, by merging InceptionV3 and ResNet50, an improved deep CNN model of feature fusion was proposed to classify images of skin cancer, whose performance on the augmented HAM10000 dataset is better than each of the two models (i.e.…”

Section: Introductionmentioning

confidence: 99%

Skin Cancer Detection Based on Hybrid Model by Means of Inception V3 and ResNet 50

Zeng¹,

Zhu²

2023

Atlantis Highlights in Computer Sciences

View full text Add to dashboard Cite

Due to the rapid growth of skin cancer patients, being diagnosed and treated at an early stage has become more and more necessary. However, only mature, and experienced doctors are capable for the precise detection of skin cancer by carrying out some expensive examination. To solve this issue, several computer-based, deep-learning detection methods have been proposed over recent years. However, most of current methods only focus on algorithms and do not provide with industrial applications. In addition, the current accuracy based on these algorithms can be further improved. This paper presents an improved deep Convolutional Neural Network (CNN) model of feature fusion by merging InceptionV3 and ResNet50, to classify images of skin cancer into eight types. By concatenating, the proposed model has not only the Inception Modules, but also the Residual Blocks, which obtains the advantages of reducing the number of parameters and mitigating the gradient problem. The experiment has been carried out on the augmented HAM10000 dataset, which contains quite several images of different types of skin cancer. The experimental results show that the proposed model has the best training performance with a validation accuracy at 87.11% among InceptionV3, ResNet50, VGG19 and itself. Besides, during prediction, the proposed model also holds the best achievement with accuracy at 0.822, precision at 0.824, recall at 0.822, and F1-score at 0.817 among the four models, which has a rivalrous performance than VGG19. Furthermore, an iOS App was also developed to provide a better interactive experience for users to acquire a diagnosis through one single image of skin lesion and communicate about their results conveniently.

show abstract

Section: Introductionmentioning

confidence: 99%

Skin Cancer Detection Based on Hybrid Model by Means of Inception V3 and ResNet 50

Zeng¹,

Zhu²

2023

Atlantis Highlights in Computer Sciences

View full text Add to dashboard Cite

show abstract

“…Due to its huge advantages in image processing, GAN appeared many variants in a very short period of time, such as pix2pix, StarGAN and CycleGAN [2][3][4]. Usually, Convolutional Neural Network (CNN) are chosen as the main structure of these algorithms due to their excellent performance in different areas [5][6][7][8][9]. Among them, pix2pix can achieve paired imageto-image translation or achieve the transformation between images in different forms; CycleGAN can achieve unpaired image-to-image translation While both pix2pix and CycleGAN can only convert between two domains, StarGAN can convert multiple domains by learning just one model.…”

Section: Introductionmentioning

confidence: 99%

Transforming images into paintings in the style of Van Gogh based on CycleGAN

Wu¹

2023

Third International Conference on Intelligent Computing and Human-Computer Interaction (ICHCI 2022)

View full text Add to dashboard Cite

Nowadays, with the rapid development of entertainment industry, filter has almost become an indispensable function in photography. However, most of the filters today are not directly change the style of an image, and even those techniques that can achieve image style transfer are difficult to be applied by users of other disciplines because of the complexity of their algorithms. In this study, a Generative Adversarial Network named CycleGAN, is used to transfer normal landscape images to images with the style of Van Gogh’s paintings. By using this model, two generators and discriminators will be formed. For the generator, they both generate a domain of images to the other domain, transfer them back to the original domain, and to make sure the generated images will not be differed so much with the original images. In this way, the generator can learn the important features of a specific style of images and apply them to another set of images with a different style. Experimental results show that the model does well in transferring normal landscape images to images with the style of Van Gogh’s paintings, but for the transformation from the style of Van Gogh’s paintings to landscape images, the effect of the model is not very ideal.

show abstract

“…According to Zhao Liya's research in 2016, the test accuracy rate was 94.39%, the average accuracy rate was 93.33%, and the average recall rate was 93%. Qiu et al [5] proposed pretrained neural network based on ResNet [3] for detecting different types of brain tumor and also achieved satisfactory result. However, all mentioned studies mentioned above did not include the impact of different convolutional neural network's depth.…”

Section: Introductionmentioning

confidence: 99%

Study on the Effect of Depth on Convolutional Neural Networks for Brain Tumor Detection

Li¹

2023

Atlantis Highlights in Computer Sciences

View full text Add to dashboard Cite

Brain tumors are one of the invasive diseases in children and adults. Due to the complexity involved in brain tumors and their characteristics, manual examination may be error prone. The use of automatic classification technology using machine learning (ML) has always shown higher accuracy than manual classification [1]. Convolutional Neural Network (CNN) is widely used as a deep learning algorithm in detecting brain tumor. However, many people have not paid their attention to how the number of layers affect the accuracy of CNN when applying it in detecting tumor. To find the relationship between the number of layers of CNN model detecting the brain tumor and the accuracy of the model, this study implements three different models with different number of layers. By using a certain data set and preprocessing, the three models' codes are successfully run, and figures and output of the accuracy and loss are showed up. After comparing the graphs and output, a certain relationship is concluded. In this article, it is indicated that the number of layers of model has positive relationship with the accuracy of this model: the more layers the model has, the higher accuracy this model will provide.

show abstract

Semantic Segmentation of Intracranial Hemorrhages in Head CT Scans

Cited by 64 publications

References 9 publications

Skin Cancer Detection Based on Hybrid Model by Means of Inception V3 and ResNet 50

Skin Cancer Detection Based on Hybrid Model by Means of Inception V3 and ResNet 50

Transforming images into paintings in the style of Van Gogh based on CycleGAN

Study on the Effect of Depth on Convolutional Neural Networks for Brain Tumor Detection

Contact Info

Product

Resources

About