scholarly journals Spectral Detector CT-Derived Pulmonary Perfusion Maps And Pulmonary Parenchyma Characteristics For The Semiautomated Classification of Pulmonary Hypertension

2021 ◽  
Author(s):  
Roman Johannes Gertz ◽  
Felix Gerhardt ◽  
Jan Robert Kröger ◽  
Rahil Shahzad ◽  
Liliana Caldeira ◽  
...  
Keyword(s):  
Pulmonary Hypertension ◽  
Iodine Content ◽  
Lung Parenchyma ◽  
High Accuracy ◽  
Pulmonary Perfusion ◽  
Spectral Detector Ct ◽  
Perfusion Defects ◽  
Pulmonary Parenchyma ◽  
Perfusion Maps

Abstract Objectives: To evaluate the usefulness of spectral detector CT (SDCT)-derived pulmonary perfusion maps and pulmonary parenchyma characteristics for the semiautomated classification of pulmonary hypertension (PH).Methods: A total of 162 consecutive patients with right heart catheter (RHC)-proven PH of different etiologies as defined by the Nice classification who underwent CT pulmonary angiography (CTPA) on SDCT and 20 patients with an invasive rule-out of PH were included in this retrospective study. Semiautomatic lung segmentation into normal and malperfused areas based on iodine content as well as automatic, virtual noncontrast-based emphysema quantification were performed. Corresponding volumes, histogram features and the ID SkewnessPerfDef-Emphysema-Index (O-index) accounting for the ratio of ID distribution in malperfused lung areas and the proportion of emphysematous lung parenchyma were computed and compared between groups.Results: Patients with PH showed a significantly greater extent of malperfused lung areas as well as stronger and more homogenous perfusion defects. In Nice class 3 and 4 patients, ID skewness revealed a significantly more homogenous ID distribution in perfusion defects than in all other subgroups. The b-index allowed for further subclassification of subgroups 3 and 4 (p < 0.001), identifying patients with chronic thromboembolic PH (CTEPH, subgroup 4) with high accuracy (AUC: 0.92, 95%-CI, 0.85-0.99).Conclusion: Abnormal pulmonary perfusion in PH can be detected and quantified by semiautomated SDCT-based pulmonary perfusion maps. ID skewness in malperfused lung areas, and the j-index allow for a classification of PH subgroups, identifying Nice class 3 and 4 patients with high accuracy, independent of reader expertise.

Segmentation of Pulmonary Parenchyma from Pulmonary CT Based on ResU-Net++ Model

2021 ◽  
Vol 11 (3) ◽  
pp. 760-766
Author(s):  
Xuantong Yan ◽  
Yuejiang Wu ◽  
Wenjun Tan
Keyword(s):  
Neural Network ◽  
Neural Network Model ◽  
Lung Parenchyma ◽  
High Accuracy ◽  
Basic Step ◽  
Feature Parameters ◽  
Deep Feature ◽  
Lung Structure ◽  
Pulmonary Parenchyma ◽  
Pulmonary Ct

Pulmonary parenchyma segmentation is an important basic step in CT detection, which requires high accuracy and speed. According to the characteristics of lung structure, we present a method for segmenting CT images of lung parenchyma based on the ResU-net++ neural network model. The model preserves the deep feature parameters extracted from the residual blocks while paying attention to preserving the feature parameters obtained by transferring the shallow convolution blocks. The experimental results show that compared with U-net and U-net++ models, the lung parenchyma segmentation results using this method are more accurate, the accuracy of lung parenchyma edge identification is significantly improved, and it is more suitable for the diverse structures at both ends of the lung.

Optimization of K Value in KNN Algorithm for Spam and Ham Email Classification

2020 ◽  
Vol 4 (2) ◽  
pp. 377-383
Author(s):  
Eko Laksono ◽  
Achmad Basuki ◽  
Fitra Bachtiar
Keyword(s):  
Frequency Distribution ◽  
Confusion Matrix ◽  
High Accuracy ◽  
K Value ◽  
A Value ◽  
Optimal Value ◽  
Classification Evaluation ◽  
Email Spam ◽  
Email Classification

There are many cases of email abuse that have the potential to harm others. This email abuse is commonly known as spam, which contains advertisements, phishing scams, and even malware. This study purpose to know the classification of email spam with ham using the KNN method as an effort to reduce the amount of spam. KNN can classify spam or ham in an email by checking it using a different K value approach. The results of the classification evaluation using confusion matrix resulted in the KNN method with a value of K = 1 having the highest accuracy value of 91.4%. From the results of the study, it is known that the optimization of the K value in KNN using frequency distribution clustering can produce high accuracy of 100%, while k-means clustering produces an accuracy of 99%. So based on the results of the existing accuracy values, the frequency distribution clustering and k-means clustering can be used to optimize the K-optimal value of the KNN in the classification of existing spam emails.

scholarly journals A Machine Vision Approach for Bioreactor Foam Sensing

2021 ◽  
pp. 247263032110088
Author(s):  
Jonas Austerjost ◽  
Robert Söldner ◽  
Christoffer Edlund ◽  
Johan Trygg ◽  
David Pollard ◽  
...  
Keyword(s):  
Machine Learning ◽  
Machine Vision ◽  
State Of The Art ◽  
Low Cost ◽  
High Accuracy ◽  
Consumer Electronics ◽  
Learning System ◽  
Automotive Applications ◽  
Fine Grained

Machine vision is a powerful technology that has become increasingly popular and accurate during the last decade due to rapid advances in the field of machine learning. The majority of machine vision applications are currently found in consumer electronics, automotive applications, and quality control, yet the potential for bioprocessing applications is tremendous. For instance, detecting and controlling foam emergence is important for all upstream bioprocesses, but the lack of robust foam sensing often leads to batch failures from foam-outs or overaddition of antifoam agents. Here, we report a new low-cost, flexible, and reliable foam sensor concept for bioreactor applications. The concept applies convolutional neural networks (CNNs), a state-of-the-art machine learning system for image processing. The implemented method shows high accuracy for both binary foam detection (foam/no foam) and fine-grained classification of foam levels.

scholarly journals Dual-Energy Computed Tomography of the Lung in COVID-19 Patients: Mismatch of Perfusion Defects and Pulmonary Opacities

Diagnostics ◽  
2020 ◽  
Vol 10 (11) ◽  
pp. 870
Author(s):  
Saif Afat ◽  
Ahmed E. Othman ◽  
Konstantin Nikolaou ◽  
Sebastian Gassenmaier
Keyword(s):  
Computed Tomography ◽  
Dual Energy ◽  
Lung Damage ◽  
Rating System ◽  
Pulmonary Perfusion ◽  
Imaging Studies ◽  
Dual Energy Computed Tomography ◽  
Contrast Enhanced ◽  
Perfusion Defects

To evaluate contrast-enhanced dual-energy computed tomography (DECT) chest examinations regarding pulmonary perfusion patterns and pulmonary opacities in patients with confirmed COVID-19 disease. Fourteen patients with 24 DECT examinations performed between April and May 2020 were included in this retrospective study. DECT studies were assessed independently by two radiologists regarding pulmonary perfusion defects, using a Likert scale ranging from 1 to 4. Furthermore, in all imaging studies the extent of pulmonary opacities was quantified using the same rating system as for perfusion defects. The main pulmonary findings were ground glass opacities (GGO) in all 24 examinations and pulmonary consolidations in 22 examinations. The total lung scores after the addition of the scores of the single lobes showed significantly higher values of opacities compared to perfusion defects, with a median of 12 (9–18) for perfusion defects and a median of 17 (15–19) for pulmonary opacities (p = 0.002). Furthermore, mosaic perfusion patterns were found in 19 examinations in areas with and without GGO. Further studies will be necessary to investigate the pathophysiological background of GGO with maintained perfusion compared to GGO with reduced perfusion, especially regarding long-term lung damage and prognosis.

scholarly journals Fully Automated Segmentation of Lung Parenchyma Using Break and Repair Strategy

2019 ◽  
Vol 28 (2) ◽  
pp. 275-289 ◽  
Author(s):  
S. Pramod Kumar ◽  
Mrityunjaya V. Latte
Keyword(s):  
Active Contour ◽  
Lung Parenchyma ◽  
Threshold Selection ◽  
Chain Code ◽  
Repair Strategy ◽  
Automated Method ◽  
Segmentation Methods ◽  
Segmentation Accuracy ◽  
Pulmonary Parenchyma ◽  
Optimal Threshold Selection

Abstract The traditional segmentation methods available for pulmonary parenchyma are not accurate because most of the methods exclude nodules or tumors adhering to the lung pleural wall as fat. In this paper, several techniques are exhaustively used in different phases, including two-dimensional (2D) optimal threshold selection and 2D reconstruction for lung parenchyma segmentation. Then, lung parenchyma boundaries are repaired using improved chain code and Bresenham pixel interconnection. The proposed method of segmentation and repairing is fully automated. Here, 21 thoracic computer tomography slices having juxtapleural nodules and 115 lung parenchyma scans are used to verify the robustness and accuracy of the proposed method. Results are compared with the most cited active contour methods. Empirical results show that the proposed fully automated method for segmenting lung parenchyma is more accurate. The proposed method is 100% sensitive to the inclusion of nodules/tumors adhering to the lung pleural wall, the juxtapleural nodule segmentation is >98%, and the lung parenchyma segmentation accuracy is >96%.

scholarly journals Deep Learning for the Classification of Non-Hodgkin Lymphoma on Histopathological Images

Cancers ◽  
2021 ◽  
Vol 13 (10) ◽  
pp. 2419
Author(s):  
Georg Steinbuss ◽  
Mark Kriegsmann ◽  
Christiane Zgorzelski ◽  
Alexander Brobeil ◽  
Benjamin Goeppert ◽  
...  
Keyword(s):  
Deep Learning ◽  
Hodgkin Lymphoma ◽  
Expert Knowledge ◽  
B Cell Lymphoma ◽  
High Accuracy ◽  
Lymphocytic Leukemia ◽  
Non Hodgkin Lymphoma ◽  
Image Patches ◽  
Histopathological Images

The diagnosis and the subtyping of non-Hodgkin lymphoma (NHL) are challenging and require expert knowledge, great experience, thorough morphological analysis, and often additional expensive immunohistological and molecular methods. As these requirements are not always available, supplemental methods supporting morphological-based decision making and potentially entity subtyping are required. Deep learning methods have been shown to classify histopathological images with high accuracy, but data on NHL subtyping are limited. After annotation of histopathological whole-slide images and image patch extraction, we trained and optimized an EfficientNet convolutional neuronal network algorithm on 84,139 image patches from 629 patients and evaluated its potential to classify tumor-free reference lymph nodes, nodal small lymphocytic lymphoma/chronic lymphocytic leukemia, and nodal diffuse large B-cell lymphoma. The optimized algorithm achieved an accuracy of 95.56% on an independent test set including 16,960 image patches from 125 patients after the application of quality controls. Automatic classification of NHL is possible with high accuracy using deep learning on histopathological images and routine diagnostic applications should be pursued.

Sign in / Sign up

Export Citation Format

Share Document