Multidomain CT Metal Artifacts Reduction Using Partial Convolution Based Inpainting

Artem Pimkin, Alexander Samoylenko, Natalia Antipina, Anna Ovechkina, Andrey Golanov, Alexandra Dalechina, Mikhail Belyaev

Результат исследований: Глава в книге, отчете, сборнике статейМатериалы для конференциирецензирование

7 Цитирования (Scopus)

Аннотация

Recent Metal Artifacts Reduction (MAR) methods for Computed Tomography are often based on image-to-image convolutional neural networks for adjustment of corrupted sinograms or images themselves. In this paper, we are exploring the capabilities of a multidomain method, which consists of both sinogram correction (projection domain step) and restored image correction (image-domain step). We formulate the first step problem directly as sinogram inpainting, which allows us to use methods of this specific field, such as partial convolutions. Moreover, we propose a synthetic data generation pipeline to avoid problems with overfitting to metal shapes set and an artifacts formation technique. The proposed method achieves state-of-the-art (-75% MSE) improvement in comparison with a classic benchmark - Li-MAR.

Язык оригиналаАнглийский
Название основной публикации2020 International Joint Conference on Neural Networks, IJCNN 2020 - Proceedings
ИздательInstitute of Electrical and Electronics Engineers Inc.
ISBN (электронное издание)9781728169262
DOI
СостояниеОпубликовано - июл. 2020
Событие2020 International Joint Conference on Neural Networks, IJCNN 2020 - Virtual, Glasgow, Великобритания
Продолжительность: 19 июл. 202024 июл. 2020

Серия публикаций

НазваниеProceedings of the International Joint Conference on Neural Networks

Конференция

Конференция2020 International Joint Conference on Neural Networks, IJCNN 2020
Страна/TерриторияВеликобритания
ГородVirtual, Glasgow
Период19/07/2024/07/20

Fingerprint

Подробные сведения о темах исследования «Multidomain CT Metal Artifacts Reduction Using Partial Convolution Based Inpainting». Вместе они формируют уникальный семантический отпечаток (fingerprint).

Цитировать