scholarly journals Spatial Pyramid Pooling in Deep Convolutional Networks for Automatic Tuberculosis Diagnosis

2020 ◽  
Vol 37 (6) ◽  
pp. 1075-1084
Author(s):  
Pike Msonda ◽  
Sait Ali Uymaz ◽  
Seda Soğukpınar Karaağaç
Keyword(s):  
Neural Networks ◽  
Developing Countries ◽  
Convolutional Neural Networks ◽  
Automated System ◽  
Machine Learning Techniques ◽  
World Health ◽  
Deep Convolutional Neural Networks ◽  
Automatic Diagnosis ◽  
Spatial Pyramid Pooling ◽  
Spatial Pyramid

In recent decades, automatic diagnosis using machine-learning techniques have been the focus of research. Mycobacterium Tuberculosis (TB) is a deadly disease that has plagued most developing countries presents a problem that can be tackled by automatic diagnosis. The World Health Organization (WHO) set years 2030 and 2035 as milestones for a significant reduction in new infections and deaths although lack of well-trained professionals and insufficient or fragile public health systems (in developing countries) are just some of the major factors that have slowed the eradication of the TB endemic. Deep convolutional neural networks (DCNNs) have demonstrated remarkable results across problem domains dealing with grid-like data (i.e., images and videos). Traditionally, a methodology for detecting TB is through radiology combined with previous success DCNN have achieved in image classification makes them the perfect candidate to classify Chest X-Ray (CXR) images. In this study, we propose three types of DCNN trained using two public datasets and another new set which we collected from Konya Education and Research Hospital, Konya, Turkey. Also, the DCNN architectures were integrated with an extra layer called Spatial Pyramid Pooling (SPP) a methodology that equips convolutional neural networks with the ability for robust feature pooling by using spatial bins. The result indicates the potential for an automated system to diagnose tuberculosis with accuracies above a radiologist professional.

scholarly journals Deep convolutional neural networks for COVID‐19 automatic diagnosis

2021 ◽  
Author(s):  
Heba M. Emara ◽  
Mohamed R. Shoaib ◽  
Mohamed Elwekeil ◽  
Walid El‐Shafai ◽  
Taha E. Taha ◽  
...  
Keyword(s):  
Neural Networks ◽  
Convolutional Neural Networks ◽  
Deep Convolutional Neural Networks ◽  
Automatic Diagnosis

scholarly journals Automatic diagnosis of disc bulge and disc herniation based on lumbar MR images using deep convolutional neural networks (Preprint)

2019 ◽  
Author(s):  
Qiong Pan ◽  
Xiyang Liu ◽  
Kai Zhang ◽  
Lin He ◽  
Zhou Dong ◽  
...  
Keyword(s):  
Neural Networks ◽  
Magnetic Resonance ◽  
Convolutional Neural Networks ◽  
Disc Herniation ◽  
Lumbar Disc ◽  
Diagnostic Efficiency ◽  
Mr Images ◽  
Deep Convolutional Neural Networks ◽  
Automatic Diagnosis ◽  
Disc Bulge

BACKGROUND Lumbar abnormalities often lead to the lower back pain which has keep plaguing people’s lives. Magnetic resonance imaging (MRI) is one of the most efficient techniques to detect lumbar diseases and widely used in clinic. How to interpret massive amounts of magnetic resonance (MR) images quickly and accurately is an urgent problem. OBJECTIVE The aim of this study is to present an automatic system to diagnosis of disc bulge and herniation which is time-saving and effective so that can reduce radiologists’ workload. METHODS The diagnosis of disorders of lumbar vertebral is highly dependent on medical images, therefore, we choose two most common diseases disc bulge and herniation as the research objects. The study is mainly about classification of the axial lumbar disc MR images using deep convolutional neural networks. RESULTS This system comprises of four steps. First step, automatic localizes vertebral bodies (including L1, L2, L3, L4, L5, and S1, L: Lumbar, S: Sacral) in sagittal images using the Faster R-CNN and 4-fold cross-validations show 100% accuracy respectively. Second step, automatically determine the corresponding disc of each axial lumbar disc MR images with 100% accuracy. In the third step, the accuracy to automatic localizes intervertebral disc region of interest (ROI) in axial MR images is 100%. The three classification (disc normal, disc bulge and disc herniation) accuracies in the last step in five groups (L1-L2, L2-L3, L3-L4, L4-L5, and L5-S1) are 92.7%, 84.4%, 92.1%, 90.4% and 84.2% respectively. CONCLUSIONS The automatic diagnosis system was successful built which can classify images of disc normal, disc bulge and disc herniation. This system provides an online test to interpret lumbar disc MR images which will be very helpful in improving the diagnostic efficiency and standardizing diagnosis reports, also, the system can be promoted to detect other lumbar abnormalities and cervical spondylosis.

scholarly journals Segmentation and classification of colon glands with deep convolutional neural networks and total variation regularization

PeerJ ◽  
2017 ◽  
Vol 5 ◽  
pp. e3874 ◽  
Author(s):  
Philipp Kainz ◽  
Michael Pfeiffer ◽  
Martin Urschler
Keyword(s):  
Neural Networks ◽  
Total Variation ◽  
Convolutional Neural Networks ◽  
Machine Learning Techniques ◽  
Superior Performance ◽  
Total Variation Regularization ◽  
Deep Convolutional Neural Networks ◽  
Large Variability ◽  
Medical Diagnoses

Segmentation of histopathology sections is a necessary preprocessing step for digital pathology. Due to the large variability of biological tissue, machine learning techniques have shown superior performance over conventional image processing methods. Here we present our deep neural network-based approach for segmentation and classification of glands in tissue of benign and malignant colorectal cancer, which was developed to participate in the GlaS@MICCAI2015 colon gland segmentation challenge. We use two distinct deep convolutional neural networks (CNN) for pixel-wise classification of Hematoxylin-Eosin stained images. While the first classifier separates glands from background, the second classifier identifies gland-separating structures. In a subsequent step, a figure-ground segmentation based on weighted total variation produces the final segmentation result by regularizing the CNN predictions. We present both quantitative and qualitative segmentation results on the recently released and publicly available Warwick-QU colon adenocarcinoma dataset associated with the GlaS@MICCAI2015 challenge and compare our approach to the simultaneously developed other approaches that participated in the same challenge. On two test sets, we demonstrate our segmentation performance and show that we achieve a tissue classification accuracy of 98% and 95%, making use of the inherent capability of our system to distinguish between benign and malignant tissue. Our results show that deep learning approaches can yield highly accurate and reproducible results for biomedical image analysis, with the potential to significantly improve the quality and speed of medical diagnoses.

Sign in / Sign up

Export Citation Format

Share Document