European Radiology: Deep learning enables 3D MRCPs for IPMN patients to be "more selective, more exciting"

Several studies have shown that intraductal papillary myxoma (IPMNs) have a certain malignant potential and are pathologically graded, ranging from low-grade to high-grade dysplasia, and some even Can progress to invasive carcinoma
.
At present, magnetic resonance imaging (MRI) is the preferred clinical diagnostic and follow-up imaging method for IPMNs, among which Three-dimensional (3D) magnetic resonance cholangiopancreatography (MRCP) is a key technique
.

To improve the temporal and spatial resolution of MRI, parallel imaging (PI) can be used in 1.
5 or 3T images or in the K domain, although there are limits to the increase in temporal and spatial resolution of the examination.

Recently, compressed sensing (CS) technology has been introduced into the clinic to reduce the number of
K-space samples.
However, one of the disadvantages of CS compared to PI is its lower
signal-to-noise ratio (SNR).
The combination of CS and PI provides better image quality
compared to CS alone.
In 2020, multiple K spatial data acquisition (fast 3Dm) technologies using each repetition time (TR) technology were introduced clinically.

Deep learning reconstruction (DLR) improves imaging quality in various anatomical fields and has been shown in several studies to improve the quality
of MRI images acquired by CS and PI.

Recently, a study published in the journal European Radiology compared the difference between acquisition time and image quality obtained by IPMN patients using various techniques, and further clarified the clinical practical value of DLR, providing technical support
for clinical accurate IPMN diagnosis and follow-up evaluation.

This review retrospectively included 32 IPMN patients, each of whom received 3D MRCPs obtained using PI, Fast 3Dm, and CS and PI, and reconstructed
with and without DLR.
Tukey's HSD test was used to compare acquisition time, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR)
obtained by all protocols.
Results from endoscopic ultrasound, ERCP, surgery, or pathology were determined as standard references, and the distribution classification
of all 3D MRCP protocols was compared by McNemar testing.

The acquisition time of Fast 3Dm and CS with and without DLR was significantly shorter than that of PI with and without DLR (P < 0.
05).

The SNR and CNR of each MRCP sequence with DLR were significantly higher than those without DLR (P < 0.
05).

The IPMN distribution accuracy of PI with and without DLR and fast 3Dm with DLR was significantly higher than that of fast 3Dm without DLR and PI without DLR (p < 0.
05).

Figure SNRs, CRs and CNRs (with or without DLR) on CBD and MPD on three-dimensional MRCPs obtained with PI, CS and Fast 3Dm

This study found that DLR improves image quality and IPMN evaluation of
3D MRCPs obtained using PI, Fast 3Dm, or CS and PI.
In addition, rapid 3Dm and CS using PI can be used as an alternative technology for PI in MRCP of IPMN patients, which provides a technical reference and basis
for the accurate diagnosis and evaluation of IPMN.

Original source:

Takahiro Matsuyama,Yoshiharu Ohno,Kaori Yamamoto,et al.
Comparison of utility of deep learning reconstruction on 3D MRCPs obtained with three different k-space data acquisitions in patients with IPMN.
DOI:10.
1007/s00330-022-08877-2