Fine-Grained Multi-Instance Classification in Microscopy Through Deep Attention

Abstract

Fine-grained classification of microscopic image data with limited samples is an open problem in computer vision and biomedical imaging. Deep learning based vision systems mostly deal with high number of low-resolution images, whereas subtle detail in biomedical images require higher resolution. To bridge this gap, we propose a simple yet effective deep network that performs two tasks simultaneously in an end-to-end manner. First, it utilises a gated attention module that can focus on multiple key instances at high resolution without extra annotations or region proposals. Second, the global structural features and local instance features are fused for final image level classification. The result is a robust but lightweight end-to-end trainable deep network that yields state-of-the-art results in two separate fine-grained multi-instance biomedical image classification tasks: a benchmark breast cancer histology dataset and our new fungi species mycology dataset. In addition, we demonstrate the interpretability of the proposed model by visualising the concordance of the learned features with clinically relevant features.

Evaluation of Fuzzy Measures Using Dempster–Shafer Belief Structure: A Classifier Fusion Framework

“…40 of the 400 images were used as the test set for image classification while the patch was extracted from the rest of the 360 images, 20% of which were used for testing patch-wise accuracy, 10% for validation while the rest 70% for patch-wise training. As can be observed from Table VI, the proposed method beats almost all existing methods barring [25] which used a complex attention-based module. 2) Novel Covid-19 Chestxray Repository: We have evaluated our proposed DS-based fuzzy ensemble approach on the Novel COVID-19 Chestxray Repository proposed in [40].…”

“…Here, we briefly explain the existing breast histology image classification methods which we consider for comparison and which were applied to the said dataset. Fan et al [25] employed a deep attention network and trained it with a batch size of 16, SGD optimizer, learning rate of 0.1, and recorded the results up to 50 epochs. DenseNet-161 was utilized with considerable offline data augmentation by the authors of the work [26].…”

Synthetic-to-real: instance segmentation of clinical cluster cells with unlabeled synthetic training

“…In the second stage, the instance cells were refined through up-down sampling and residuals. Similarly, Fan et al (2020) constructed an attention map to learn each instance end-to-end, it can effectively suppress the background and improve the recognition ability of overlapping instances.…”

Scaled Dilation of DropBlock Optimization in Convolutional Neural Network for Fungus Classification

“…The efficiency of the experiment results was 98.03% (accuracy). In [22], the research proposed fine-grained multi-instance-based deep attention. The model utilized 2,000 images in the training and experiment.…”