Computational Intelligence and Neuroscience · 2022 · 237 citations · 20 references
ComputingBiomedical EngineeringComputer ScienceComputer VisionDiagnostic ImagingMachine LearningCancer DetectionDeep LearningInstance SegmentationMedical Image AnalysisMedical ImagingObject DetectionConvolutionConvolutional Neural Network DetectionDeep Learning DetectionDeep Learning-based DetectionDeep Learning-based DiagnosisDetectorsHistopathological Image AnalysisHistopathological Tumor Classification
Breast cancer is a lethal illness that has a high mortality rate. In treatment, the accuracy of diagnosis is crucial. Machine learning and deep learning may be beneficial to doctors. The proposed backbone network is critical for the present performance of CNN-based detectors. Integrating dilated convolution, ResNet, and Alexnet increases detection performance. The composite dilated backbone network (CDBN) is an innovative method for integrating many identical backbones into a single robust backbone. Hence, CDBN uses the lead backbone feature maps to identify objects. It feeds high-level output features from previous backbones into the next backbone in a stepwise way. We show that most contemporary detectors can easily include CDBN to improve performance achieved mAP improvements ranging from 1.5 to 3.0 percent on the breast cancer histopathological image classification (BreakHis) dataset. Experiments have also shown that instance segmentation may be improved. In the BreakHis dataset, CDBN enhances the baseline detector cascade mask R-CNN (mAP = 53.3). The proposed CDBN detector does not need pretraining. It creates high-level traits by combining low-level elements. This network is made up of several identical backbones that are linked together. The composite dilated backbone considers the linked backbones CDBN.
20
Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos et al. · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2017 · 22.1K citations
The Role of Context for Object Detection and Semantic Segmentation in the Wild
Roozbeh Mottaghi, Xianjie Chen, Xiaobai Liu et al. · 2014 · 1.5K citations
Associative hierarchical CRFs for object class image segmentation
Ľubor Ladický, Chris Russell, Pushmeet Kohli et al. · 2009 · 621 citations