Evaluating Deep Neural Network Models on Ultrasound Single Image Super Resolution

Mikaeili M., BİLGE H. Ş.

2023 Medical Technologies Congress, TIPTEKNO 2023, Famagusta, Kıbrıs (Gkry), 10 - 12 Kasım 2023 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/tiptekno59875.2023.10359188
  • Basıldığı Şehir: Famagusta
  • Basıldığı Ülke: Kıbrıs (Gkry)
  • Anahtar Kelimeler: Deep neural network, single image super resolution, ultrasound imaging
  • Gazi Üniversitesi Adresli: Evet

Özet

Ultrasound imaging is widely applied within the medical field, offering numerous applications. Nevertheless, a significant limitation of this imaging modality lies in its susceptibility to artifacts and speckle noise. Thus, the provision of ultrasound images with elevated resolution holds paramount importance for accurate diagnosis and treatment. This study focuses on evaluating the efficacy of three distinct deep neural network architectures in achieving super resolution. Additionally, the obtained results are juxtaposed against those yielded by conventional methods. Based on our findings, the ESPCN network architecture emerges as the most effective, surpassing both alternative neural network models and conventional techniques. The outcomes of ESPCN are closely followed by those of the SRCNN network architecture.