Transfer Learning for automated search for defects on chest X-rays

Borisov, A. A.; Semenov, Serafim S.; Arzamasov, Kirill

doi:10.24835/1607-0763-1243

Cited by 3 publications

(1 citation statement)

References 15 publications

(1 reference statement)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ResNet152V2's robustness and adaptability have led to its widespread use. ResNet152V2 has found application in the medical field, for instance, in automated defect detection on chest X-rays [37] and in the diagnosis of COVID-19 using chest X-ray and CT images [38]. ResNet152V2 can learn complex features from images, contributing to its high accuracy in these scenarios.…”

Section: A Resnet152v2mentioning

confidence: 99%

WeedFocusNet: A Revolutionary Approach using the Attention-Driven ResNet152V2 Transfer Learning

Devi,

Sandhya,

Chatrapati

2023

IJRITCC

View full text Add to dashboard Cite

The advancement of modern agriculture is heavily dependent on accurate weed detection, which contributes to efficient resource utilization and increased crop yield. Traditional methods, however, often need more accuracy and efficiency. This paper presents WeedFocusNet, an innovative approach that leverages attention-driven ResNet152V2 transfer learning addresses these challenges. This approach enhances model generalization and focuses on critical features for weed identification, thereby overcoming the limitations of existing methods. The objective is to develop a model that enhances weed detection accuracy and optimizes computational efficiency. WeedFocusNet, a novel deep-learning model, performs weed detection better by employing attention-driven transfer learning based on the ResNet152V2 architecture. The model integrates an attention module, concentrating its predictions on the most significant image features. Evaluated on a dataset of weed and crop images, WeedFocusNet achieved an accuracy of 99.28%, significantly outperforming previous methods and models, such as MobileNetV2, ResNet50, and custom CNN models, in terms of accuracy, time complexity, and memory usage, despite its larger memory footprint. These results emphasize the transformative potential of WeedFocusNet as a powerful approach for automating weed detection in agricultural fields.

show abstract

Section: A Resnet152v2mentioning

confidence: 99%

WeedFocusNet: A Revolutionary Approach using the Attention-Driven ResNet152V2 Transfer Learning

Devi,

Sandhya,

Chatrapati

2023

IJRITCC

View full text Add to dashboard Cite

show abstract

Investigation of the capabilities of algorithms for automated quality assurance of DICOM metadata of chest X-ray examinations

Borisov,

Arzamasov,

Semenov

et al. 2023

Medicinskaâ vizualizaciâ

Self Cite

View full text Add to dashboard Cite

Purpose. To evaluate the quality of filling DICOM tags responsible for the orientation, scanning area and photometric interpretation of the image, as well as to develop and test algorithms for automatically determining the true values of these tags for chest x-rays and fluorograms.Materials and methods. To assess the quality of filling DICOM tags, were used 1885 studies obtained from the Unified Radiological Information Service of the Unified Medical Information and Analysis System (ERIS EMIAS). For training and validation of algorithms for automatic determination of the true values of tags, were used datasets of radiographs in standard frontal and lateral projections, from open databases and from ERIS EMIAS (12,920 studies in total). The deep neural network architecture VGG 19 was chosen as the basis for creating algorithms.Results. We found that the frequency of missing values in DICOM tags can range from 6 to 75%, depending on the tag. At the same time, up to 70% of filled tag values have errors. We obtained next models: a model for determining the anatomical area of x-ray examination, a model for determining the projection on the chest x-ray, a model for determining the photometric interpretation of the image. All of the obtained algorithms have high classification quality indicators. The AUC for each of the obtained models was more than 0.99.Conclusions. Our study shows that a large number of studies in diagnostic practice contain incorrect values of DICOM tags, which can critically affect the implementation of software based on artificial intelligence technology in clinical practice. Our obtained algorithms can be integrated into the development process of such software and used in the preprocessing of images before their analysis.

show abstract

Using of neural networks to search for errors of patient's positioning on chest X-rays

Васильев,

Владзимирский,

Омелянская

et al. 2023

ПСТП

View full text Add to dashboard Cite

В работе представлены результаты применения трансферного обучения глубоких сверточных нейронных сетей для задачи поиска рентгенограмм органов грудной клетки с нарушениями укладки и позиционирования пациента. Оцениваемые нейросетевые архитектуры: InceptionV3, Xception, ResNet152V2, InceptionResNetV2, DenseNet201, VGG16, VGG19, MobileNetV2, NASNetLarge. Для обучения и тестирования использовались рентгенограммы грудной клетки, полученные из открытых наборов данных и Единого радиологического информационного сервиса города Москвы. Все полученные модели имели метрики диагностической точности выше 95%, при этом модели на основе архитектур ResNet152V2, DenseNet201, VGG16, MobileNetV2 имели статистически значимо лучшие метрики, чем другие модели. Наилучшие абсолютные значения метрик показала модель ResNet152V2 (AUC =0.999 ,чувствительность=0.987, специфичность=0.988, общая валидность =0.988, F1 мера = 0.988). Модель MobileNetV2 показала наилучшую скорость обработки одного исследования (67.8±5.0 ms). Широкое использование полученных нами алгоритмов способно облегчить создание больших баз данных качественных медицинских изображений, а также оптимизировать контроль качества при выполнении рентгенографических исследований органов грудной клетки.

show abstract

Transfer Learning for automated search for defects on chest X-rays

Cited by 3 publications

References 15 publications

WeedFocusNet: A Revolutionary Approach using the Attention-Driven ResNet152V2 Transfer Learning

WeedFocusNet: A Revolutionary Approach using the Attention-Driven ResNet152V2 Transfer Learning

Investigation of the capabilities of algorithms for automated quality assurance of DICOM metadata of chest X-ray examinations

Using of neural networks to search for errors of patient's positioning on chest X-rays

Contact Info

Product

Resources

About