Purpose or Learning Objective To compare T2 and diffusion-weighted images (DWI) in upper abdomen magnetic resonance imaging (MRI) with AIR Recon Deep Learning (ARDL) algorithm with standard acquisition (non-ARDL), in terms of quantitative and qualitative image analysis and scanning time. Methods or Background Thirty consecutive healthy volunteers (23 female, mean age 55±22 years) were included and underwent unenhanced upper abdomen MRI (1.5 Tesla) from May 2021 to October 2021. Exclusion criteria were: contraindications to MRI and severe artefacts on MRI sequences. Examinations included T2 and DWI axial sequences with both standard protocol and ARDL. A radiologist evaluated objective image quality by drawing fixed regions of interest (ROIs) in the liver parenchyma, gallbladder and background to calculate signal-to-noise-ratio(SNR) and contrast to-noise-ratio(CNR). Then, subjective image quality was assessed by two radiologists independently with a five-point Likert scale including parameters as parenchyma edge sharpness, contrast, truncation and motion artefacts and overall image quality. Acquisition timing was also recorded and analysed. Results or Findings The objective analysis showed no significant differences between SNR and CNR in ARDL-T2 sequences vs non-ARDL-T2 (SNR=282.23 vs 249.6 and CNR=1122.18vs877, P= 0.5038 and 0.57 respectively) and in DWI sequences with ARDL vs non-ARDL (SNR=677.91 vs 509.35 and CNR=408.75 vs 421.30, P=0.6884 and 0.6435 respectively). For subjective analysis, ARDL sequences showed significantly better overall image quality with lower motion and truncation artefacts and higher sharpness and contrast (all p<0,0001) with the good inter-rater agreement (k=0,83925). Acquisition timing was significantly lower in both ARDL sequences compared to non-ARDL ones (T2=19.07s vs 24s and DWI=197.93s vs 498.52s, all p<0.0001). Conclusion ARDL sequences showed significantly higher overall image quality with reduced acquisition timing, in particular for DWI. Limitations Reduced number of sequences tested and small population sample.

Artificial intelligence-based VS standard acquisition in upper abdomen MRI: quantitative and qualitative image analysis / Masci, B.; Zerunian, M.; Pucciarelli, F.; Polici, M.; Piccinni, G.; Polverari, D.; Iannicelli, E.; Caruso, D.; Laghi, A.. - (2022). (Intervento presentato al convegno ECR 2022 tenutosi a Vienna).

Artificial intelligence-based VS standard acquisition in upper abdomen MRI: quantitative and qualitative image analysis

B. Masci;M. Zerunian;F. Pucciarelli;M. Polici;G. Piccinni;E. Iannicelli;D. Caruso;A. Laghi
2022

Abstract

Purpose or Learning Objective To compare T2 and diffusion-weighted images (DWI) in upper abdomen magnetic resonance imaging (MRI) with AIR Recon Deep Learning (ARDL) algorithm with standard acquisition (non-ARDL), in terms of quantitative and qualitative image analysis and scanning time. Methods or Background Thirty consecutive healthy volunteers (23 female, mean age 55±22 years) were included and underwent unenhanced upper abdomen MRI (1.5 Tesla) from May 2021 to October 2021. Exclusion criteria were: contraindications to MRI and severe artefacts on MRI sequences. Examinations included T2 and DWI axial sequences with both standard protocol and ARDL. A radiologist evaluated objective image quality by drawing fixed regions of interest (ROIs) in the liver parenchyma, gallbladder and background to calculate signal-to-noise-ratio(SNR) and contrast to-noise-ratio(CNR). Then, subjective image quality was assessed by two radiologists independently with a five-point Likert scale including parameters as parenchyma edge sharpness, contrast, truncation and motion artefacts and overall image quality. Acquisition timing was also recorded and analysed. Results or Findings The objective analysis showed no significant differences between SNR and CNR in ARDL-T2 sequences vs non-ARDL-T2 (SNR=282.23 vs 249.6 and CNR=1122.18vs877, P= 0.5038 and 0.57 respectively) and in DWI sequences with ARDL vs non-ARDL (SNR=677.91 vs 509.35 and CNR=408.75 vs 421.30, P=0.6884 and 0.6435 respectively). For subjective analysis, ARDL sequences showed significantly better overall image quality with lower motion and truncation artefacts and higher sharpness and contrast (all p<0,0001) with the good inter-rater agreement (k=0,83925). Acquisition timing was significantly lower in both ARDL sequences compared to non-ARDL ones (T2=19.07s vs 24s and DWI=197.93s vs 498.52s, all p<0.0001). Conclusion ARDL sequences showed significantly higher overall image quality with reduced acquisition timing, in particular for DWI. Limitations Reduced number of sequences tested and small population sample.
2022
ECR 2022
04 Pubblicazione in atti di convegno::04d Abstract in atti di convegno
Artificial intelligence-based VS standard acquisition in upper abdomen MRI: quantitative and qualitative image analysis / Masci, B.; Zerunian, M.; Pucciarelli, F.; Polici, M.; Piccinni, G.; Polverari, D.; Iannicelli, E.; Caruso, D.; Laghi, A.. - (2022). (Intervento presentato al convegno ECR 2022 tenutosi a Vienna).
File allegati a questo prodotto
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1645006
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact