A review on deep-learning algorithms for fetal ultrasound-image analysis

Medical Image Analysis, Journal Year: 2022, Volume and Issue: 83, P. 102629 - 102629

Published: Oct. 14, 2022

Language: Английский

---

A Novel Machine Learning-Based Prediction Method for Early Detection and Diagnosis of Congenital Heart Disease Using ECG Signal Processing

Technologies, Journal Year: 2024, Volume and Issue: 12(1), P. 4 - 4

Published: Jan. 2, 2024

Congenital heart disease (CHD) represents a multifaceted medical condition that requires early detection and diagnosis for effective management, given its diverse presentations subtle symptoms manifest from birth. This research article introduces groundbreaking healthcare application, the Machine Learning-based Heart Disease Prediction Method (ML-CHDPM), tailored to address these challenges expedite timely identification classification of CHD in pregnant women. The ML-CHDPM model leverages state-of-the-art machine learning techniques categorize cases, taking into account pertinent clinical demographic factors. Trained on comprehensive dataset, captures intricate patterns relationships, resulting precise predictions classifications. evaluation model’s performance encompasses sensitivity, specificity, accuracy, area under receiver operating characteristic curve. Remarkably, findings underscore ML-CHDPM’s superiority across six pivotal metrics: precision, recall, false positive rate (FPR), negative (FNR). method achieves an average accuracy 94.28%, precision 87.54%, recall 96.25%, specificity 91.74%, FPR 8.26%, FNR 3.75%. These outcomes distinctly demonstrate effectiveness reliably predicting classifying cases. marks significant stride toward diagnosis, harnessing advanced within realm ECG signal processing, specifically

Language: Английский

---

Deep learning on medical image analysis

CAAI Transactions on Intelligence Technology, Journal Year: 2024, Volume and Issue: unknown

Published: June 24, 2024

Abstract Medical image analysis plays an irreplaceable role in diagnosing, treating, and monitoring various diseases. Convolutional neural networks (CNNs) have become popular as they can extract intricate features patterns from extensive datasets. The paper covers the structure of CNN its advances explores different types transfer learning strategies well classic pre‐trained models. also discusses how has been applied to areas within medical analysis. This comprehensive overview aims assist researchers, clinicians, policymakers by providing detailed insights, helping them make informed decisions about future research policy initiatives improve patient outcomes.

Language: Английский

---

The severity prediction of the binary and multi-class cardiovascular disease − A machine learning-based fusion approach

Computational Biology and Chemistry, Journal Year: 2022, Volume and Issue: 98, P. 107672 - 107672

Published: March 31, 2022

Language: Английский

---

Use of artificial intelligence and deep learning in fetal ultrasound imaging

Ultrasound in Obstetrics and Gynecology, Journal Year: 2022, Volume and Issue: 62(2), P. 185 - 194

Published: Nov. 27, 2022

Deep learning is considered the leading artificial intelligence tool in image analysis general. Deep-learning algorithms excel at recognition, which makes them valuable medical imaging. Obstetric ultrasound has become gold standard imaging modality for detection and diagnosis of fetal malformations. However, relies heavily on operator's experience, making it unreliable inexperienced hands. Several studies have proposed use deep-learning models as a to support sonographers, an attempt overcome these problems inherent ultrasound. many clinical applications field imaging, including identification normal abnormal anatomy measurement biometry. In this Review, we provide comprehensive explanation fundamentals deep with particular focus its applicability. © 2022 International Society Ultrasound Obstetrics Gynecology.

Language: Английский

---

TransFSM: Fetal Anatomy Segmentation and Biometric Measurement in Ultrasound Images Using a Hybrid Transformer

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2023, Volume and Issue: 28(1), P. 285 - 296

Published: Nov. 6, 2023

Biometric parameter measurements are powerful tools for evaluating a fetus's gestational age, growth pattern, and abnormalities in 2D ultrasound. However, it is still challenging to measure fetal biometric parameters automatically due the indiscriminate confusing factors, limited foreground-background contrast, variety of anatomy shapes at different ages, blurry anatomical boundaries ultrasound images. The performance standard CNN architecture these tasks restricted receptive field. We propose novel hybrid Transformer framework, TransFSM, address multi-anatomy segmentation measurement tasks. Unlike vanilla based on single-scale input, TransFSM has deformable self-attention mechanism so can effectively process multi-scale information segment with irregular sizes. devised BAD capture more intrinsic local details using boundary-wise prior knowledge, which compensates defects extracting features. In addition, auxiliary head designed improve mask prediction by learning semantic correspondence same pixel categories feature discriminability among categories. Extensive experiments were conducted clinical cases benchmark datasets experiment results indicate that our method achieves state-of-the-art seven evaluation metrics compared CNN-based, Transformer-based, approaches. By Knowledge distillation, proposed create compact efficient model high deploying potential resource-constrained scenarios. Our study serves as unified framework estimation across multiple regions monitor practice.

Language: Английский

---

Deep Learning for Improving the Effectiveness of Routine Prenatal Screening for Major Congenital Heart Diseases

Journal of Clinical Medicine, Journal Year: 2022, Volume and Issue: 11(21), P. 6454 - 6454

Published: Oct. 31, 2022

Early prenatal screening with an ultrasound (US) can significantly lower newborn mortality caused by congenital heart diseases (CHDs). However, the need for expertise in fetal cardiologists and high volume of cases limit practically achievable detection rates. Hence, automated to support clinicians is desirable. This paper presents analyses potential deep learning (DL) techniques diagnose CHDs USs. Four convolutional neural network architectures were compared select best classifier satisfactory results. dense (DenseNet) 201 architecture was selected classification seven CHDs, such as ventricular septal defect, atrial atrioventricular Ebstein's anomaly, tetralogy Fallot, transposition great arteries, hypoplastic left syndrome, a normal control. The sensitivity, specificity, accuracy DenseNet201 model 100%, respectively, intra-patient scenario 99%, 97%, 98%, inter-patient scenario. We used DL prediction validate our proposed against results three expert cardiologists. produces result, which means that interpret decision improve CHD diagnostics. work represents step toward goal assisting front-line sonographers diagnoses at population level.

Language: Английский

---

TransFusionNet: Semantic and Spatial Features Fusion Framework for Liver Tumor and Vessel Segmentation Under JetsonTX2

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2022, Volume and Issue: 27(3), P. 1173 - 1184

Published: Sept. 16, 2022

Liver cancer is one of the most common malignant diseases worldwide. Segmentation and reconstruction liver tumors vessels in CT images can provide convenience for physicians preoperative planning surgical intervention. In this paper, we introduced a TransFusionNet framework, which consists semantic feature extraction module, local spatial an edge multi-scale fusion module to achieve fine-grained segmentation vessels. addition, applied transfer learning approach pre-train using public datasets then fine-tune model further improve fitting effect. Furthermore, proposed intelligent quantization scheme compress weights achieved high performance inference on JetsonTX2. The framework mean IoU 0.854 vessel task, 0.927 tumor task. When profiling Computational Performance quantized inference, our 4TFLOPs Node with NVIDIA RTX3090 132GFLOPs This unprecedented effect solves accuracy bottleneck automated certain extent.

Language: Английский

---

A progressive growing generative adversarial network composed of enhanced style-consistent modulation for fetal ultrasound four-chamber view editing synthesis

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108438 - 108438

Published: April 13, 2024

Language: Английский

---

SPReCHD: Four-Chamber Semantic Parsing Network for Recognizing Fetal Congenital Heart Disease in Medical Metaverse

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2022, Volume and Issue: 28(6), P. 3672 - 3682

Published: Nov. 4, 2022

Echocardiography is essential for evaluating cardiac anatomy and function during early recognition screening congenital heart disease (CHD), a widespread complex malformation. However, fetal CHD still faces many difficulties due to instinctive movements, artifacts in ultrasound images, distinctive structures. These factors hinder capturing robust discriminative representations from resulting CHD's low prenatal detection rate. Hence, we propose multi-scale gated axial-transformer network (MSGATNet) capture four-chamber semantic information. Then, SPReCHD: parsing recognizing the clinical treatment of medical metaverse, integrating MSGATNet segment locate arbitrary contours, further distinguished heart. Comprehensive experiments indicate that our SPReCHD sufficient CHD, achieving precision 95.92%, recall 94%, an accuracy 95%, $F_{1}$ score 94.95% on test set, dramatically improving

Language: Английский

---