Endoscopic Image Classification and Retrieval using Clustered Convolutional Features

Semi-Supervised Bladder Tissue Classification in Multi-Domain Endoscopic Images

OBJECTIVE

Accurate visual classification of bladder tissue during Trans-Urethral Resection of Bladder Tumor (TURBT) procedures is essential to improve early cancer diagnosis and treatment. During TURBT interventions, White Light Imaging (WLI) and Narrow Band Imaging (NBI) techniques are used for lesion detection. Each imaging technique provides diverse visual information that allows clinicians to identify and classify cancerous lesions. Computer vision methods that use both imaging techniques could improve endoscopic diagnosis. We address the challenge of tissue classification when annotations are available only in one domain, in our case WLI, and the endoscopic images correspond to an unpaired dataset, i.e. there is no exact equivalent for every image in both NBI and WLI domains.

METHOD

We propose a semi-surprised Generative Adversarial Network (GAN)-based method composed of three main components: a teacher network trained on the labeled WLI data; a cycle-consistency GAN to perform unpaired image-to-image translation, and a multi-input student network. To ensure the quality of the synthetic images generated by the proposed GAN we perform a detailed quantitative, and qualitative analysis with the help of specialists.

CONCLUSION

The overall average classification accuracy, precision, and recall obtained with the proposed method for tissue classification are 0.90, 0.88, and 0.89 respectively, while the same metrics obtained in the unlabeled domain (NBI) are 0.92, 0.64, and 0.94 respectively. The quality of the generated images is reliable enough to deceive specialists.

SIGNIFICANCE

This study shows the potential of using semi-supervised GAN-based bladder tissue classification when annotations are limited in multi-domain data.

Prognosis Prediction in COVID-19 Patients through Deep Feature Space Reasoning

The COVID-19 pandemic has presented a unique challenge for physicians worldwide, as they grapple with limited data and uncertainty in diagnosing and predicting disease outcomes. In such dire circumstances, the need for innovative methods that can aid in making informed decisions with limited data is more critical than ever before. To allow prediction with limited COVID-19 data as a case study, we present a complete framework for progression and prognosis prediction in chest X-rays (CXR) through reasoning in a COVID-specific deep feature space. The proposed approach relies on a pre-trained deep learning model that has been fine-tuned specifically for COVID-19 CXRs to identify infection-sensitive features from chest radiographs. Using a neuronal attention-based mechanism, the proposed method determines dominant neural activations that lead to a feature subspace where neurons are more sensitive to COVID-related abnormalities. This process allows the input CXRs to be projected into a high-dimensional feature space where age and clinical attributes like comorbidities are associated with each CXR. The proposed method can accurately retrieve relevant cases from electronic health records (EHRs) using visual similarity, age group, and comorbidity similarities. These cases are then analyzed to gather evidence for reasoning, including diagnosis and treatment. By using a two-stage reasoning process based on the Dempster-Shafer theory of evidence, the proposed method can accurately predict the severity, progression, and prognosis of a COVID-19 patient when sufficient evidence is available. Experimental results on two large datasets show that the proposed method achieves 88% precision, 79% recall, and 83.7% F-score on the test sets.

A comprehensive review on acquisition of phenotypic information of Prunoideae fruits: Image technology

Fruit phenotypic information reflects all the physical, physiological, biochemical characteristics and traits of fruit. Accurate access to phenotypic information is very necessary and meaningful for post-harvest storage, sales and deep processing. The methods of obtaining phenotypic information include traditional manual measurement and damage detection, which are inefficient and destructive. In the field of fruit phenotype research, image technology is increasingly mature, which greatly improves the efficiency of fruit phenotype information acquisition. This review paper mainly reviews the research on phenotypic information of Prunoideae fruit based on three imaging techniques (RGB imaging, hyperspectral imaging, multispectral imaging). Firstly, the classification was carried out according to the image type. On this basis, the review and summary of previous studies were completed from the perspectives of fruit maturity detection, fruit quality classification and fruit disease damage identification. Analysis of the advantages and disadvantages of various types of images in the study, and try to give the next research direction for improvement.

Dexterous Identification of Carcinoma through ColoRectalCADx with Dichotomous Fusion CNN and UNet Semantic Segmentation

Human colorectal disorders in the digestive tract are recognized by reference colonoscopy. The current system recognizes cancer through a three-stage system that utilizes two sets of colonoscopy data. However, identifying polyps by visualization has not been addressed. The proposed system is a five-stage system called ColoRectalCADx, which provides three publicly accessible datasets as input data for cancer detection. The three main datasets are CVC Clinic DB, Kvasir2, and Hyper Kvasir. After the image preprocessing stages, system experiments were performed with the seven prominent convolutional neural networks (CNNs) (end-to-end) and nine fusion CNN models to extract the spatial features. Afterwards, the end-to-end CNN and fusion features are executed. These features are derived from Discrete Wavelet Transform (DWT) and Vector Support Machine (SVM) classification, that was used to retrieve time and spatial frequency features. Experimentally, the results were obtained for five stages. For each of the three datasets, from stage 1 to stage 3, end-to-end CNN, DenseNet-201 obtained the best testing accuracy (98%, 87%, 84%), ((98%, 97%), (87%, 87%), (84%, 84%)), ((99.03%, 99%), (88.45%, 88%), (83.61%, 84%)). For each of the three datasets, from stage 2, CNN DaRD-22 fusion obtained the optimal test accuracy ((93%, 97%) (82%, 84%), (69%, 57%)). And for stage 4, ADaRDEV2-22 fusion achieved the best test accuracy ((95.73%, 94%), (81.20%, 81%), (72.56%, 58%)). For the input image segmentation datasets CVC Clinc-Seg, KvasirSeg, and Hyper Kvasir, malignant polyps were identified with the UNet CNN model. Here, the loss score datasets (CVC clinic DB was 0.7842, Kvasir2 was 0.6977, and Hyper Kvasir was 0.6910) were obtained.

Precision medical image hash retrieval by interpretability and feature fusion

BACKGROUND AND OBJECTIVE

To address the problem of low accuracy of medical image retrieval due to high inter-class similarity and easy omission of lesions, a precision medical image hash retrieval method combining interpretability and feature fusion is proposed, taking chest X-ray images as an example.

METHODS

Firstly, the DenseNet-121 network is pre-trained on a large dataset of medical images without manual annotation using the comparison to learn (C2L) method to obtain a backbone network model containing more medical representations with training weights. Then, a global network is constructed by using global image learning to acquire an interpretable saliency map as attention mechanisms, which can generate a mask crop to get a local discriminant region. Thirdly, the local discriminant regions are used as local network inputs to obtain local features, and the global features are used with the local features by dimension in the pooling layer. Finally, a hash layer is added between the fully connected layer and the classification layer of the backbone network, defining classification loss, quantization loss and bit-balanced loss functions to generate high-quality hash codes. The final retrieval result is output by calculating the similarity metric of the hash codes.

RESULTS

Experiments on the Chest X-ray8 dataset demonstrate that our proposed interpretable saliency map can effectively locate focal regions, the fusion of features can avoid information omission, and the combination of three loss functions can generate more accurate hash codes. Compared with the current advanced medical image retrieval methods, this method can effectively improve the accuracy of medical image retrieval.

CONCLUSIONS

The proposed hash retrieval approach combining interpretability and feature fusion can effectively improve the accuracy of medical image retrieval which can be potentially applied in computer-aided-diagnosis systems.

Proposing Novel Data Analytics Method for Anatomical Landmark Identification from Endoscopic Video Frames

Background

The anatomical landmarks contain the characteristics that are used to guide the gastroenterologists during the endoscopy. The expert can also ensure the completion of examination with the help of the anatomical landmarks. Automatic detection of anatomical landmarks in endoscopic video frames can be helpful for guiding the physicians during screening the gastrointestinal tract (GI).

Method

This study presents an automatic novel method for anatomical landmark detection of GI tract from endoscopic video frames based on semisupervised deep convolutional neural network (CNN) and compares the results with supervised CNN model. We consider the anatomical landmarks from Kvasir dataset that includes 500 images for each class of Z-line, pylorus, and cecum. The resolution of these images varies from 750 × 576 up to 1920 × 1072 pixels.

Result

Experimental results show that the supervised CNN has highly desirable performance with accuracy of 100%. Also, our proposed semisupervised CNN can compete with a slight difference similar to the CNN model. Our proposed semisupervised model trained using 1, 5, 10, and 20 percent of training data records as labeled training dataset has the average accuracy of 83%, 98%, 99%, and 99%, respectively.

Conclusion

The main advantage of our proposed method is achieving the high accuracy with small amount of labeled data without spending time for labeling more data. The strength of our proposed method saves the required labor, cost, and time for data labeling.

GASTRO-CADx: a three stages framework for diagnosing gastrointestinal diseases

Gastrointestinal (GI) diseases are common illnesses that affect the GI tract. Diagnosing these GI diseases is quite expensive, complicated, and challenging. A computer-aided diagnosis (CADx) system based on deep learning (DL) techniques could considerably lower the examination cost processes and increase the speed and quality of diagnosis. Therefore, this article proposes a CADx system called Gastro-CADx to classify several GI diseases using DL techniques. Gastro-CADx involves three progressive stages. Initially, four different CNNs are used as feature extractors to extract spatial features. Most of the related work based on DL approaches extracted spatial features only. However, in the following phase of Gastro-CADx, features extracted in the first stage are applied to the discrete wavelet transform (DWT) and the discrete cosine transform (DCT). DCT and DWT are used to extract temporal-frequency and spatial-frequency features. Additionally, a feature reduction procedure is performed in this stage. Finally, in the third stage of the Gastro-CADx, several combinations of features are fused in a concatenated manner to inspect the effect of feature combination on the output results of the CADx and select the best-fused feature set. Two datasets referred to as Dataset I and II are utilized to evaluate the performance of Gastro-CADx. Results indicated that Gastro-CADx has achieved an accuracy of 97.3% and 99.7% for Dataset I and II respectively. The results were compared with recent related works. The comparison showed that the proposed approach is capable of classifying GI diseases with higher accuracy compared to other work. Thus, it can be used to reduce medical complications, death-rates, in addition to the cost of treatment. It can also help gastroenterologists in producing more accurate diagnosis while lowering inspection time.

Deep Learning for Multigrade Brain Tumor Classification in Smart Healthcare Systems: A Prospective Survey

Brain tumor is one of the most dangerous cancers in people of all ages, and its grade recognition is a challenging problem for radiologists in health monitoring and automated diagnosis. Recently, numerous methods based on deep learning have been presented in the literature for brain tumor classification (BTC) in order to assist radiologists for a better diagnostic analysis. In this overview, we present an in-depth review of the surveys published so far and recent deep learning-based methods for BTC. Our survey covers the main steps of deep learning-based BTC methods, including preprocessing, features extraction, and classification, along with their achievements and limitations. We also investigate the state-of-the-art convolutional neural network models for BTC by performing extensive experiments using transfer learning with and without data augmentation. Furthermore, this overview describes available benchmark data sets used for the evaluation of BTC. Finally, this survey does not only look into the past literature on the topic but also steps on it to delve into the future of this area and enumerates some research directions that should be followed in the future, especially for personalized and smart healthcare.

Residual LSTM layered CNN for classification of gastrointestinal tract diseases

nowadays, considering the number of patients per specialist doctor, the size of the need for automatic medical image analysis methods can be understood. These systems, which are very advantageous compared to manual systems both in terms of cost and time, benefit from artificial intelligence (AI). AI mechanisms that mimic the decision-making process of a specialist increase their diagnosis performance day by day, depending on technological developments. In this study, an AI method is proposed to effectively classify Gastrointestinal (GI) Tract Image datasets containing a small number of labeled data. The proposed AI method uses the convolutional neural network (CNN) architecture, which is accepted as the most successful automatic classification method of today, as a backbone. According to our approach, a shallowly trained CNN architecture needs to be supported by a strong classifier to classify unbalanced datasets robustly. For this purpose, the features in each pooling layer in the CNN architecture are transmitted to an LSTM layer. A classification is made by combining all LSTM layers. All experiments are carried out using AlexNet, GoogLeNet, and ResNet to evaluate the contribution of the proposed residual LSTM structure fairly. Besides, three different experiments are carried out with 2000, 4000, and 6000 samples to determine the effect of sample number change on the proposed method. The performance of the proposed method is higher than other state-of-the-art methods.

Language-based translation and prediction of surgical navigation steps for endoscopic wayfinding assistance in minimally invasive surgery

Purpose

In the context of aviation and automotive navigation technology, assistance functions are associated with predictive planning and wayfinding tasks. In endoscopic minimally invasive surgery, however, assistance so far relies primarily on image-based localization and classification. We show that navigation workflows can be described and used for the prediction of navigation steps.

Methods

A natural description vocabulary for observable anatomical landmarks in endoscopic images was defined to create 3850 navigation workflow sentences from 22 annotated functional endoscopic sinus surgery (FESS) recordings. Resulting FESS navigation workflows showed an imbalanced data distribution with over-represented landmarks in the ethmoidal sinus. A transformer model was trained to predict navigation sentences in sequence-to-sequence tasks. The training was performed with the Adam optimizer and label smoothing in a leave-one-out cross-validation study. The sentences were generated using an adapted beam search algorithm with exponential decay beam rescoring. The transformer model was compared to a standard encoder-decoder-model, as well as HMM and LSTM baseline models.

Results

The transformer model reached the highest prediction accuracy for navigation steps at 0.53, followed by 0.35 of the LSTM and 0.32 for the standard encoder-decoder-network. With an accuracy of sentence generation of 0.83, the prediction of navigation steps at sentence-level benefits from the additional semantic information. While standard class representation predictions suffer from an imbalanced data distribution, the attention mechanism also considered underrepresented classes reasonably well.

Conclusion

We implemented a natural language-based prediction method for sentence-level navigation steps in endoscopic surgery. The sentence-level prediction method showed a potential that word relations to navigation tasks can be learned and used for predicting future steps. Further studies are needed to investigate the functionality of path prediction. The prediction approach is a first step in the field of visuo-linguistic navigation assistance for endoscopic minimally invasive surgery.

Disease Detection in Plum Using Convolutional Neural Network under True Field Conditions

The agriculture sector faces crop losses every year due to diseases around the globe, which adversely affect food productivity and quality. Detecting and identifying plant diseases at an early stage is still a challenge for farmers, particularly in developing countries. Widespread use of mobile computing devices and the advancements in artificial intelligence have created opportunities for developing technologies to assist farmers in plant disease detection and treatment. To this end, deep learning has been widely used for disease detection in plants with highly favorable outcomes. In this paper, we propose an efficient convolutional neural network-based disease detection framework in plum under true field conditions for resource-constrained devices. As opposed to the publicly available datasets, images used in this study were collected in the field by considering important parameters of image-capturing devices such as angle, scale, orientation, and environmental conditions. Furthermore, extensive data augmentation was used to expand the dataset and make it more challenging to enable robust training. Investigations of recent architectures revealed that transfer learning of scale-sensitive models like Inception yield results much better with such challenging datasets with extensive data augmentation. Through parameter quantization, we optimized the Inception-v3 model for deployment on resource-constrained devices. The optimized model successfully classified healthy and diseased fruits and leaves with more than 92% accuracy on mobile devices.

HyperKvasir, a comprehensive multi-class image and video dataset for gastrointestinal endoscopy

Artificial intelligence is currently a hot topic in medicine. However, medical data is often sparse and hard to obtain due to legal restrictions and lack of medical personnel for the cumbersome and tedious process to manually label training data. These constraints make it difficult to develop systems for automatic analysis, like detecting disease or other lesions. In this respect, this article presents HyperKvasir, the largest image and video dataset of the gastrointestinal tract available today. The data is collected during real gastro- and colonoscopy examinations at Bærum Hospital in Norway and partly labeled by experienced gastrointestinal endoscopists. The dataset contains 110,079 images and 374 videos, and represents anatomical landmarks as well as pathological and normal findings. The total number of images and video frames together is around 1 million. Initial experiments demonstrate the potential benefits of artificial intelligence-based computer-assisted diagnosis systems. The HyperKvasir dataset can play a valuable role in developing better algorithms and computer-assisted examination systems not only for gastro- and colonoscopy, but also for other fields in medicine.

Endoscopic Image Clustering with Temporal Ordering Information Based on Dynamic Programming

In this paper, we propose a clustering method with temporal ordering information for endoscopic image sequences. It is difficult to collect a sufficient amount of endoscopic image datasets to train machine learning techniques by manual labeling. The clustering of endoscopic images leads to group-based labeling, which is useful for reducing the cost of dataset construction. Therefore, in this paper, we propose a clustering method where the property of endoscopic image sequences is fully utilized. For the proposed method, a deep neural network was used to extract features from endoscopic images, and clustering with temporal ordering information was solved by dynamic programming. In the experiments, we clustered the esophagogastroduodenoscopy images. From the results, we confirmed that the performance was improved by using the sequential property.

Development of a Strategy to Predict and Detect Falls Using Wearable Sensors

Falls are a prevalent problem in actual society. Some falls result in injuries and the cost associated with their treatment is high. This is a complex problem that requires several steps in order to be tackled. Firstly, it is crucial to develop strategies that recognize the locomotion mode, indicating the state of the subject in various situations. This article aims to develop a strategy capable of identifying normal gait, the pre-fall condition, and the fall situation, based on a wearable system (IMUs-based). This system was used to collect data from healthy subjects that mimicked falls. The strategy consists, essentially, in the construction and use of classifiers as tools for recognizing the locomotion modes. Two approaches were explored. Associative Skill Memories (ASMs) based classifier and a Convolutional Neural Network (CNN) classifier based on deep learning. Finally, these classifiers were compared, providing for a tool with a good accuracy in recognizing the locomotion modes. Results have shown that the accuracy of the classifiers was quite acceptable. The CNN presented the best results with 92.71% of accuracy considering the pre-fall step different from normal steps, and 100% when not considering.

Medical Image Retrieval with Compact Binary Codes Generated in Frequency Domain Using Highly Reactive Convolutional Features

Efficient retrieval of relevant medical cases using semantically similar medical images from large scale repositories can assist medical experts in timely decision making and diagnosis. However, the ever-increasing volume of images hinder performance of image retrieval systems. Recently, features from deep convolutional neural networks (CNN) have yielded state-of-the-art performance in image retrieval. Further, locality sensitive hashing based approaches have become popular for their ability to allow efficient retrieval in large scale datasets. In this paper, we present a highly efficient method to compress selective convolutional features into sequence of bits using Fast Fourier Transform (FFT). Firstly, highly reactive convolutional feature maps from a pre-trained CNN are identified for medical images based on their neuronal responses using optimal subset selection algorithm. Then, layer-wise global mean activations of the selected feature maps are transformed into compact binary codes using binarization of its Fourier spectrum. The acquired hash codes are highly discriminative and can be obtained efficiently from the original feature vectors without any training. The proposed framework has been evaluated on two large datasets of radiology and endoscopy images. Experimental evaluations reveal that the proposed method significantly outperforms other features extraction and hashing schemes in both effectiveness and efficiency.