Deep Learning Paradigm and Its Bias for Coronary Artery Wall Segmentation in Intravascular Ultrasound Scans: A Closer Look

Journal of Cardiovascular Development and Disease, Journal Year: 2023, Volume and Issue: 10(12), P. 485 - 485

Published: Dec. 4, 2023

Coronary artery disease (CAD) has the highest mortality rate; therefore, its diagnosis is vital. Intravascular ultrasound (IVUS) a high-resolution imaging solution that can image coronary arteries, but software via wall segmentation and quantification been evolving. In this study, deep learning (DL) paradigm was explored along with bias.Using PRISMA model, 145 best UNet-based non-UNet-based methods for were selected analyzed their characteristics scientific clinical validation. This study computed thickness by estimating inner outer borders of IVUS cross-sectional scans. Further, review bias in DL system first time when it comes to Three methods, namely (i) ranking, (ii) radial, (iii) regional area, applied compared using Venn diagram. Finally, presented explainable AI (XAI) paradigms framework.UNet provides powerful walls scans due ability extract automated features at different scales encoders, reconstruct segmented decoders, embed variants skip connections. Most research hampered lack motivation XAI pruned (PAI) models. None UNet models met criteria bias-free design. For assessment settings, necessary move from paper-to-practice approach.

Language: Английский

---

An Explainable AI System for Medical Image Segmentation With Preserved Local Resolution: Mammogram Tumor Segmentation

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 125543 - 125561

Published: Jan. 1, 2023

Medical image segmentation aims to identify important or suspicious regions within medical images. However, many challenges are usually faced while developing networks for this type of analysis. First, preserving the original resolution is utmost importance task where identifying subtle features abnormalities can significantly impact accuracy diagnosis. The introduction dilated convolution module helped preserve in deep convolutional neural networks, but it has a drawback: loss local spatial due increased kernel sparsity checkboard patterns. To address this, work, double-dilated proposed maintain achieving large receptive field. This approach applied tumor breast cancer mammograms as proof-of-concept. Additionally, study tackles issue pixel-level class imbalance mammogram screenings by comparing various functions find best one mass segmentation. Our work also addresses "black-box" nature models performing quantitative and qualitative evaluations their interpretability using Gradient weighted Class Activation Map (Grad-CAM) with other explainable An experimental analysis on lesion performed from INBreast dataset, both before after integrating dilation into state-of-the-art network. results demonstrate effectiveness terms Dice similarity Miss Detection rate promotes Tversky Loss function training pixel-imbalanced data Grad-CAM explaining results.

Language: Английский

---

Multi-Level Training and Testing of CNN Models in Diagnosing Multi-Center COVID-19 and Pneumonia X-ray Images

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(18), P. 10270 - 10270

Published: Sept. 13, 2023

This study aimed to address three questions in AI-assisted COVID-19 diagnostic systems: (1) How does a CNN model trained on one dataset perform test datasets from disparate medical centers? (2) What accuracy gains can be achieved by enriching the training with new images? (3) learned features elucidate classification results, and how do they vary among different models? To achieve these aims, four models—AlexNet, ResNet-50, MobileNet, VGG-19—were five rounds incrementally adding images baseline set comprising 11,538 chest X-ray images. In each round, models were tested decreasing levels of image similarity. Notably, all showed performance drops when containing outlier or sourced other clinics. Round 1, 95.2~99.2% was for Level 1 testing (i.e., same clinic but apart only), 94.7~98.3% 2 an external similar). However, drastically decreased 3 rotation deformation), mean sensitivity plummeting 99% 36%. For 4 another clinic), 97% 86%, 67%. Rounds 3, 25% 50% improved average Level-3 15% 23% 56% 71% 83%). 5, increased Level-4 81% 92% 95%, respectively. Among models, ResNet-50 demonstrated most robust across five-round training/testing phases, while VGG-19 persistently underperformed. Heatmaps intermediate activation visual correlations pneumonia manifestations insufficient explicitly explain classification. heatmaps at shed light progression models’ learning behavior.

Language: Английский

---

Resnet Transfer Learning For Enhanced Medical Image Classification In Healthcare

Published: Dec. 29, 2023

This work overcomes the limitations of sparsely labeled data by optimizing ResNet transfer learning methods in medical classification images. Using a deductive approach along with interpretive philosophy, we optimize for better diagnostic performance on healthcare sets. Our team technical includes preprocessing datasets, configuring model architectures, and fine-tuning hyperparameters using secondary data. The improved as demonstrated results is confirmed metrics such precision, reliability, recall. Analyses comparisons demonstrate superiority over basic models. Upcoming tasks include working together to create standardized benchmarks, improving interpretability scalability, verifying actual clinical settings.

Language: Английский

---

Exploiting Machine Learning and LSTM for Human Activity Recognition: Using Physiological and Biological Sensor Data from Actigraph

2022 IEEE International Conference on Industrial Technology (ICIT), Journal Year: 2024, Volume and Issue: 3, P. 1 - 8

Published: March 25, 2024

Human activity recognition involves identifying the daily living activities of an individual through utilization sensor attributes and intelligent learning algorithms. The identification intricate human proves to be a labo-rious task, given inherent difficulty capturing long-term dependencies extracting efficient features from unprocessed data. For this purpose, study aims at recognizing classifying using physiological biological data generated by Actigraph, as they can accurately measure moderate-to-vigorous intensity physical which is mostly affected body composition also better suited for self-monitoring. We examined effectiveness these applying prevalent machine classifiers long short-term memory (LSTM) networks on recently publicly available data, includes accelerometer heart rate recordings. results our experiments showed that LSTM models performed than conventional ML with best result achieving accuracy 86.5%. findings confirms significance in more.

Language: Английский

---

Augmenting Radiological Diagnostics with AI for Tuberculosis and COVID-19 Disease Detection: Deep Learning Detection of Chest Radiographs

Diagnostics, Journal Year: 2024, Volume and Issue: 14(13), P. 1334 - 1334

Published: June 24, 2024

In this research, we introduce a network that can identify pneumonia, COVID-19, and tuberculosis using X-ray images of patients' chests. The study emphasizes tuberculosis, healthy lung conditions, discussing how advanced neural networks, like VGG16 ResNet50, improve the detection issues from images. To prepare for model's input requirements, enhanced them through data augmentation techniques training purposes. We evaluated performance by analyzing precision, recall, F1 scores across training, validation, testing datasets. results show ResNet50 model outperformed with accuracy resilience. It displayed superior ROC AUC values in both validation test scenarios. Particularly impressive were ResNet50's precision recall rates, nearing 0.99 all conditions set. On hand, also performed well during testing-detecting 0.93. Our highlights our deep learning method showcasing effectiveness over traditional approaches VGG16. This progress utilizes methods to enhance classification augmenting balancing them. positions approach as an advancement state-of-the-art applications imaging. By enhancing reliability diagnosing ailments such COVID-19 models have potential transform care treatment strategies, highlighting their role clinical diagnostics.

Language: Английский

---

HyEpiSeiD: a hybrid convolutional neural network and gated recurrent unit model for epileptic seizure detection from electroencephalogram signals

Brain Informatics, Journal Year: 2024, Volume and Issue: 11(1)

Published: Aug. 21, 2024

Abstract Epileptic seizure (ES) detection is an active research area, that aims at patient-specific ES with high accuracy from electroencephalogram (EEG) signals. The early of crucial for timely medical intervention and prevention further injuries the patients. This work proposes a robust deep learning framework called HyEpiSeiD extracts self-trained features pre-processed EEG signals using hybrid combination convolutional neural network followed by two gated recurrent unit layers performs prediction based on those extracted features. proposed evaluated public datasets, UCI Epilepsy Mendeley datasets. model achieved 99.01% 97.50% classification accuracy, respectively, outperforming most state-of-the-art methods in epilepsy domain.

Language: Английский

---

MultiNet 2.0: A Lightweight Attention-based Deep Learning Network for Stenosis measurement in Carotid Ultrasound scans and Cardiovascular Risk Assessment

Computerized Medical Imaging and Graphics, Journal Year: 2024, Volume and Issue: 117, P. 102437 - 102437

Published: Oct. 1, 2024

Language: Английский

---

Emerging Technology-Driven Hybrid Models for Preventing and Monitoring Infectious Diseases: A Comprehensive Review and Conceptual Framework

Diagnostics, Journal Year: 2023, Volume and Issue: 13(19), P. 3047 - 3047

Published: Sept. 25, 2023

The emergence of the infectious diseases, such as novel coronavirus, a significant global health threat has emphasized urgent need for effective treatments and vaccines. As diseases become more common around world, it is important to have strategies in place prevent monitor them. This study reviews hybrid models that incorporate emerging technologies preventing monitoring diseases. It also presents comprehensive review employed since outbreak COVID-19. encompasses integrate innovative technologies, blockchain, Internet Things (IoT), big data, artificial intelligence (AI). By harnessing these system enables secure contact tracing source isolation. Based on review, conceptual framework model proposes incorporates technologies. proposed tracing, isolation using blockchain technology, IoT sensors, data collection. A approach With continued research development model, efforts effectively combat safeguard public will continue.

Language: Английский

---

Exploring Explainable Melanoma Classification: Leveraging Pre-trained Deep Learning Model on MED-NODE Dataset

2022 9th International Conference on Computing for Sustainable Global Development (INDIACom), Journal Year: 2024, Volume and Issue: unknown, P. 737 - 740

Published: Feb. 28, 2024

Skin cancer, an extremely common and potentially fatal condition, emphasizes the critical importance of timely precise detection. This study presents a thorough examination dermatological image classification using deep learning models on Med Node dataset. Five prominent models, including InceptionV3, Xception, VGG19, EfficientNetB1, DenseNet201, were assessed for their ability to discern between melanoma naevus instances. Noteworthy variations in performance metrics observed, with Xception standing out exceptional accuracy 95.88% perfect precision recall both classes. In contrast, InceptionV3 demonstrated balanced trade-off recall. VGG19 exhibited comparatively lower performance, while EfficientNetB1 DenseNet201 showcased outstanding accuracy, leading remarkable 96.47%. A subsequent statistical analysis z-scores two-tailed p-values confirmed significant differences among top three (EfficientNetB1, DenseNet201). The compared proposed model existing PECK Ensemble model. results indicated substantial 5% improvement We have also added explainable AI (XAI) Lime visualize lesion section. Z-score is calculated check its reliability.

Language: Английский

---