This paper presents a novel computerized technique for the segmentation of nuclei in hematoxylin and eosin(H&E)stained histopathology images.The purpose of this study is to overcome the challenges faced in automat...This paper presents a novel computerized technique for the segmentation of nuclei in hematoxylin and eosin(H&E)stained histopathology images.The purpose of this study is to overcome the challenges faced in automated nuclei segmentation due to the diversity of nuclei structures that arise from differences in tissue types and staining protocols,as well as the segmentation of variable-sized and overlapping nuclei.To this extent,the approach proposed in this study uses an ensemble of the UNet architecture with various Convolutional Neural Networks(CNN)architectures as encoder backbones,along with stain normalization and test time augmentation,to improve segmentation accuracy.Additionally,this paper employs a Structure-Preserving Color Normalization(SPCN)technique as a preprocessing step for stain normalization.The proposed model was trained and tested on both single-organ and multi-organ datasets,yielding an F1 score of 84.11%,mean Intersection over Union(IoU)of 81.67%,dice score of 84.11%,accuracy of 92.58%and precision of 83.78%on the multi-organ dataset,and an F1 score of 87.04%,mean IoU of 86.66%,dice score of 87.04%,accuracy of 96.69%and precision of 87.57%on the single-organ dataset.These findings demonstrate that the proposed model ensemble coupled with the right pre-processing and post-processing techniques enhances nuclei segmentation capabilities.展开更多
A brain tumor is the uncharacteristic progression of tissues in the brain.These are very deadly,and if it is not diagnosed at an early stage,it might shorten the affected patient’s life span.Hence,their classificatio...A brain tumor is the uncharacteristic progression of tissues in the brain.These are very deadly,and if it is not diagnosed at an early stage,it might shorten the affected patient’s life span.Hence,their classification and detection play a critical role in treatment.Traditional Brain tumor detection is done by biopsy which is quite challenging.It is usually not preferred at an early stage of the disease.The detection involvesMagneticResonance Imaging(MRI),which is essential for evaluating the tumor.This paper aims to identify and detect brain tumors based on their location in the brain.In order to achieve this,the paper proposes a model that uses an extended deep Convolutional Neural Network(CNN)named Contour Extraction based Extended EfficientNet-B0(CE-EEN-B0)which is a feed-forward neural network with the efficient net layers;three convolutional layers and max-pooling layers;and finally,the global average pooling layer.The site of tumors in the brain is one feature that determines its effect on the functioning of an individual.Thus,this CNN architecture classifies brain tumors into four categories:No tumor,Pituitary tumor,Meningioma tumor,andGlioma tumor.This network provides an accuracy of 97.24%,a precision of 96.65%,and an F1 score of 96.86%which is better than already existing pre-trained networks and aims to help health professionals to cross-diagnose an MRI image.This model will undoubtedly reduce the complications in detection and aid radiologists without taking invasive steps.展开更多
Diabetic Retinopathy (DR) is a serious hazard that can result inirreversible blindness if not addressed in a timely manner. Hence, numeroustechniques have been proposed for the accurate and timely detection ofthis dis...Diabetic Retinopathy (DR) is a serious hazard that can result inirreversible blindness if not addressed in a timely manner. Hence, numeroustechniques have been proposed for the accurate and timely detection ofthis disease. Out of these, Deep Learning (DL) and Computer Vision (CV)methods for multiclass categorization of color fundus images diagnosed withDiabetic Retinopathy have sparked considerable attention. In this paper,we attempt to develop an extended ResNet152V2 architecture-based DeepLearning model, named ResNet2.0 to aid the timely detection of DR. TheAPTOS-2019 datasetwas used to train the model. This consists of 3662 fundusimages belonging to five different stages of DR: no DR (Class 0), mild DR(Class 1), moderate DR (Class 2), severe DR (Class 3), and proliferativeDR (Class 4). The model was gauged based on ability to detect stage-wiseDR. The images were pre-processed using negative and positive weightedGaussian-based masks as feature engineering to further enhance the qualityof the fundus images by removing the noise and normalizing the images. Upsamplingand data augmentation methods were used to address the skewnessof the original dataset. The proposed model achieved an overall accuracyof 91% and an area under the receiver-operating characteristic curve (AUC)score of 95.1%, outperforming existing Deep Learning models by around 10%.Furthermore, the class-wise F1 score for No DR was 92%, Mild DR was 82%,Moderate DR was 66%, Severe was DR 89% and Proliferative DR was 80%.展开更多
文摘This paper presents a novel computerized technique for the segmentation of nuclei in hematoxylin and eosin(H&E)stained histopathology images.The purpose of this study is to overcome the challenges faced in automated nuclei segmentation due to the diversity of nuclei structures that arise from differences in tissue types and staining protocols,as well as the segmentation of variable-sized and overlapping nuclei.To this extent,the approach proposed in this study uses an ensemble of the UNet architecture with various Convolutional Neural Networks(CNN)architectures as encoder backbones,along with stain normalization and test time augmentation,to improve segmentation accuracy.Additionally,this paper employs a Structure-Preserving Color Normalization(SPCN)technique as a preprocessing step for stain normalization.The proposed model was trained and tested on both single-organ and multi-organ datasets,yielding an F1 score of 84.11%,mean Intersection over Union(IoU)of 81.67%,dice score of 84.11%,accuracy of 92.58%and precision of 83.78%on the multi-organ dataset,and an F1 score of 87.04%,mean IoU of 86.66%,dice score of 87.04%,accuracy of 96.69%and precision of 87.57%on the single-organ dataset.These findings demonstrate that the proposed model ensemble coupled with the right pre-processing and post-processing techniques enhances nuclei segmentation capabilities.
文摘A brain tumor is the uncharacteristic progression of tissues in the brain.These are very deadly,and if it is not diagnosed at an early stage,it might shorten the affected patient’s life span.Hence,their classification and detection play a critical role in treatment.Traditional Brain tumor detection is done by biopsy which is quite challenging.It is usually not preferred at an early stage of the disease.The detection involvesMagneticResonance Imaging(MRI),which is essential for evaluating the tumor.This paper aims to identify and detect brain tumors based on their location in the brain.In order to achieve this,the paper proposes a model that uses an extended deep Convolutional Neural Network(CNN)named Contour Extraction based Extended EfficientNet-B0(CE-EEN-B0)which is a feed-forward neural network with the efficient net layers;three convolutional layers and max-pooling layers;and finally,the global average pooling layer.The site of tumors in the brain is one feature that determines its effect on the functioning of an individual.Thus,this CNN architecture classifies brain tumors into four categories:No tumor,Pituitary tumor,Meningioma tumor,andGlioma tumor.This network provides an accuracy of 97.24%,a precision of 96.65%,and an F1 score of 96.86%which is better than already existing pre-trained networks and aims to help health professionals to cross-diagnose an MRI image.This model will undoubtedly reduce the complications in detection and aid radiologists without taking invasive steps.
文摘Diabetic Retinopathy (DR) is a serious hazard that can result inirreversible blindness if not addressed in a timely manner. Hence, numeroustechniques have been proposed for the accurate and timely detection ofthis disease. Out of these, Deep Learning (DL) and Computer Vision (CV)methods for multiclass categorization of color fundus images diagnosed withDiabetic Retinopathy have sparked considerable attention. In this paper,we attempt to develop an extended ResNet152V2 architecture-based DeepLearning model, named ResNet2.0 to aid the timely detection of DR. TheAPTOS-2019 datasetwas used to train the model. This consists of 3662 fundusimages belonging to five different stages of DR: no DR (Class 0), mild DR(Class 1), moderate DR (Class 2), severe DR (Class 3), and proliferativeDR (Class 4). The model was gauged based on ability to detect stage-wiseDR. The images were pre-processed using negative and positive weightedGaussian-based masks as feature engineering to further enhance the qualityof the fundus images by removing the noise and normalizing the images. Upsamplingand data augmentation methods were used to address the skewnessof the original dataset. The proposed model achieved an overall accuracyof 91% and an area under the receiver-operating characteristic curve (AUC)score of 95.1%, outperforming existing Deep Learning models by around 10%.Furthermore, the class-wise F1 score for No DR was 92%, Mild DR was 82%,Moderate DR was 66%, Severe was DR 89% and Proliferative DR was 80%.
