Automatic detection of retinopathy with optical coherence tomography images via a semi-supervised deep learning method

Identifying retinopathy in optical coherence tomography images with less labeled data via contrastive graph regularization

Retinopathy detection using optical coherence tomography (OCT) images has greatly advanced with computer vision but traditionally requires extensive annotated data, which is time-consuming and expensive. To address this issue, we propose a novel contrastive graph regularization method for detecting retinopathies with less labeled OCT images. This method combines class prediction probabilities and embedded image representations for training, where the two representations interact and co-evolve within the same training framework. Specifically, we leverage memory smoothing constraints to improve pseudo-labels, which are aggregated by nearby samples in the embedding space, effectively reducing overfitting to incorrect pseudo-labels. Our method, using only 80 labeled OCT images, outperforms existing methods on two widely used OCT datasets, with classification accuracy exceeding 0.96 and an Area Under the Curve (AUC) value of 0.998. Additionally, compared to human experts, our method achieves expert-level performance with only 80 labeled images and surpasses most experts with just 160 labeled images.

A new computer-aided diagnosis tool based on deep learning methods for automatic detection of retinal disorders from OCT images

PURPOSE

Early detection of retinal disorders using optical coherence tomography (OCT) images can prevent vision loss. Since manual screening can be time-consuming, tedious, and fallible, we present a reliable computer-aided diagnosis (CAD) software based on deep learning. Also, we made efforts to increase the interpretability of the deep learning methods, overcome their vague and black box nature, and also understand their behavior in the diagnosis.

METHODS

We propose a novel method to improve the interpretability of the used deep neural network by embedding the rich semantic information of abnormal areas based on the ophthalmologists' interpretations and medical descriptions in the OCT images. Finally, we trained the classification network on a small subset of the online publicly available University of California San Diego (UCSD) dataset with an overall of 29,800 OCT images.

RESULTS

The experimental results on the 1000 test OCT images show that the proposed method achieves the overall precision, accuracy, sensitivity, and f1-score of 97.6%, 97.6%, 97.6%, and 97.59%, respectively. Also, the heat map images provide a clear region of interest which indicates that the interpretability of the proposed method is increased dramatically.

CONCLUSION

The proposed software can help ophthalmologists in providing a second opinion to make a decision, and primitive automated diagnoses of retinal diseases and even it can be used as a screening tool, in eye clinics. Also, the improvement of the interpretability of the proposed method causes to increase in the model generalization, and therefore, it will work properly on a wide range of other OCT datasets.

Multi-kernel driven 3D convolutional neural network for automated detection of lung nodules in chest CT scans

The accurate position detection of lung nodules is crucial in early chest computed tomography (CT)-based lung cancer screening, which helps to improve the survival rate of patients. Deep learning methodologies have shown impressive feature extraction ability in the CT image analysis task, but it is still a challenge to develop a robust nodule detection model due to the salient morphological heterogeneity of nodules and complex surrounding environment. In this study, a multi-kernel driven 3D convolutional neural network (MK-3DCNN) is proposed for computerized nodule detection in CT scans. In the MK-3DCNN, a residual learning-based encoder-decoder architecture is introduced to employ the multi-layer features of the deep model. Considering the various nodule sizes and shapes, a multi-kernel joint learning block is developed to capture 3D multi-scale spatial information of nodule CT images, and this is conducive to improving nodule detection performance. Furthermore, a multi-mode mixed pooling strategy is designed to replace the conventional single-mode pooling manner, and it reasonably integrates the max pooling, average pooling, and center cropping pooling operations to obtain more comprehensive nodule descriptions from complicated CT images. Experimental results on the public dataset LUNA16 illustrate that the proposed MK-3DCNN method achieves more competitive nodule detection performance compared to some state-of-the-art algorithms. The results on our constructed clinical dataset CQUCH-LND indicate that the MK-3DCNN has a good prospect in clinical practice.

Wavelet scattering transform application in classification of retinal abnormalities using OCT images

To assist ophthalmologists in diagnosing retinal abnormalities, Computer Aided Diagnosis has played a significant role. In this paper, a particular Convolutional Neural Network based on Wavelet Scattering Transform (WST) is used to detect one to four retinal abnormalities from Optical Coherence Tomography (OCT) images. Predefined wavelet filters in this network decrease the computation complexity and processing time compared to deep learning methods. We use two layers of the WST network to obtain a direct and efficient model. WST generates a sparse representation of the images which is translation-invariant and stable concerning local deformations. Next, a Principal Component Analysis classifies the extracted features. We evaluate the model using four publicly available datasets to have a comprehensive comparison with the literature. The accuracies of classifying the OCT images of the OCTID dataset into two and five classes were [Formula: see text] and [Formula: see text], respectively. We achieved an accuracy of [Formula: see text] in detecting Diabetic Macular Edema from Normal ones using the TOPCON device-based dataset. Heidelberg and Duke datasets contain DME, Age-related Macular Degeneration, and Normal classes, in which we achieved accuracy of [Formula: see text] and [Formula: see text], respectively. A comparison of our results with the state-of-the-art models shows that our model outperforms these models for some assessments or achieves nearly the best results reported so far while having a much smaller computational complexity.

Distinctions between Choroidal Neovascularization and Age Macular Degeneration in Ocular Disease Predictions via Multi-Size Kernels ξcho-Weighted Median Patterns

Age-related macular degeneration is a visual disorder caused by abnormalities in a part of the eye's retina and is a leading source of blindness. The correct detection, precise location, classification, and diagnosis of choroidal neovascularization (CNV) may be challenging if the lesion is small or if Optical Coherence Tomography (OCT) images are degraded by projection and motion. This paper aims to develop an automated quantification and classification system for CNV in neovascular age-related macular degeneration using OCT angiography images. OCT angiography is a non-invasive imaging tool that visualizes retinal and choroidal physiological and pathological vascularization. The presented system is based on new retinal layers in the OCT image-specific macular diseases feature extractor, including Multi-Size Kernels ξcho-Weighted Median Patterns (MSKξMP). Computer simulations show that the proposed method: (i) outperforms current state-of-the-art methods, including deep learning techniques; and (ii) achieves an overall accuracy of 99% using ten-fold cross-validation on the Duke University dataset and over 96% on the noisy Noor Eye Hospital dataset. In addition, MSKξMP performs well in binary eye disease classifications and is more accurate than recent works in image texture descriptors.

LDADN: a local discriminant auxiliary disentangled network for key-region-guided chest X-ray image synthesis augmented in pneumoconiosis detection

Pneumoconiosis is deemed one of China's most common and serious occupational diseases. Its high prevalence and treatment cost create enormous pressure on socio-economic development. However, due to the scarcity of labeled data and class-imbalanced training sets, the computer-aided diagnostic based on chest X-ray (CXR) images of pneumoconiosis remains a challenging task. Current CXR data augmentation solutions cannot sufficiently extract small-scaled features in lesion areas and synthesize high-quality images. Thus, it may cause error detection in the diagnosis phase. In this paper, we propose a local discriminant auxiliary disentangled network (LDADN) to synthesize CXR images and augment in pneumoconiosis detection. This model enables the high-frequency transfer of details by leveraging batches of mutually independent local discriminators. Cooperating with local adversarial learning and the Laplacian filter, the feature in the lesion area can be disentangled by a single network. The results show that LDADN is superior to other compared models in the quantitative assessment metrics. When used for data augmentation, the model synthesized image significantly boosts the performance of the detection accuracy to 99.31%. Furthermore, this study offers beneficial references for insufficient label or class imbalanced medical image data analysis.

Most favorable stimulation duration in the sensorimotor cortex for fNIRS-based BCI

One of the primary objectives of the brain-computer interface (BCI) is to obtain a command with higher classification accuracy within the shortest possible time duration. Therefore, this study evaluates several stimulation durations to propose a duration that can yield the highest classification accuracy. Furthermore, this study aims to address the inherent delay in the hemodynamic responses (HRs) for the command generation time. To this end, HRs in the sensorimotor cortex were evaluated for the functional near-infrared spectroscopy (fNIRS)-based BCI. To evoke brain activity, right-hand-index finger poking and tapping tasks were used. In this study, six different stimulation durations (i.e., 1, 3, 5, 7, 10, and 15 s) were tested on 10 healthy male subjects. Upon stimulation, different temporal features and multiple time windows were utilized to extract temporal features. The extracted features were then classified using linear discriminant analysis. The classification results using the main HR showed that a 5 s stimulation duration could yield the highest classification accuracy, i.e., 74%, with a combination of the mean and maximum value features. However, the results were not significantly different from the classification accuracy obtained using the 15 s stimulation. To further validate the results, a classification using the initial dip was performed. The results obtained endorsed the finding with an average classification accuracy of 73.5% using the features of minimum peak and skewness in the 5 s window. The results based on classification using the initial dip for 5 s were significantly different from all other tested stimulation durations (p < 0.05) for all feature combinations. Moreover, from the visual inspection of the HRs, it is observed that the initial dip occurred as soon as the task started, but the main HR had a delay of more than 2 s. Another interesting finding is that impulsive stimulation in the sensorimotor cortex can result in the generation of a clearer initial dip phenomenon. The results reveal that the command for the fNIRS-based BCI can be generated using the 5 s stimulation duration. In conclusion, the use of the initial dip can reduce the time taken for the generation of commands and can be used to achieve a higher classification accuracy for the fNIRS-BCI within a 5 s task duration rather than relying on longer durations.