Clinical Usefulness of 2D Thin-Slice Images Made Possible by 3T DLR-MRI-
- Maximizing Utility in the Field of Orthopedics -
Takahide Kakigi, MD, PhD
Vantage Galan 3T / ZGO*, a high-end MRI system manufactured by Canon Medical Systems Corporation, has been in operation at Kyoto University Hospital, Japan, since 2019. The system was updated to the latest system in 2020. This lecture focuses on the clinical usefulness of high-resolution 2D thin-slice images and MPR images obtained using a 3T Deep Learning Reconstruction (DLR) MRI system, in the field of orthopedics.
* Vantage Galan 3T / ZGO is not commercially available in all country.
Importance of high-resolution images in the field of orthopedics
1. Overview of AiCE
Advanced intelligent Clear-IQ Engine (AiCE) is a noise reduction technique based on advanced Deep Learning Reconstruction technology developed by Canon Medical Systems Corporation. Using this Deep Learning technology, AiCE can create high-SNR images from low-SNR images.
A deep convolutional neural network (DCNN) is trained in advance to transform images with high noise levels into images with low noise levels. The algorithm obtained by training the DCNN is then installed in an MRI system, allowing noise in newly acquired low-SNR images to be removed to obtain high-SNR images. As the DCNN was trained using only the high-frequency noise components, noise reduction (denoising) can be achieved regardless of sequence.
When AiCE is used in combination with a rapid imaging technique, such as SPEEDER or Compressed SPEEDER, 2D images with a standard slice thickness can be acquired in a shorter scan time, and image quality can be improved compared to images acquired using conventional methods. In addition, by applying AiCE to images acquired with slightly longer scan time, image quality can be further improved and high-resolution and thin-slice 2D images can be obtained.
2. Clinical usefulness of 2D 1-mm images and MPR images
In the field of orthopedics, it can be difficult to detect and evaluate pathologies, such as rotator cuff tears, cartilage injuries, and labral tears using images with a standard resolution and slice thickness. Ideally, such injuries should be evaluated using high-resolution 2D thin-slice images (2D 1-mm images). 3D images can also provide detailed views, but 2D 1-mm images allow us to obtain a clearer understanding of the patterns and locations of these injuries.
Some of the key advantages of 2D 1-mm images are a reduction in the partial volume effect and the ability to evaluate pathology using multi-sectional images generated by multiplanar reconstruction (MPR). The tradeoff between a higher SNR and a longer scan time can be largely avoided by employing AiCE and a rapid imaging technique in combination. As for scan time, it may be shorter to scan 2D 1-mm images than 2D 3-mm images in three image planes. A further advantage of 2D 1-mm images over 3D images is that high-resolution imaging can be performed with excellent tissue contrast and with relatively less motion artifacts.
Figure 1 shows a comparison of 2D 3-mm images (a) and 2D 1-mm images (b) of a tear of the subscapularis tendon obtained using fat-suppressed proton density-weighted imaging (FS-PDWI) (coronal slices). In the 2D 3-mm images, there appears to be a tear in the most cranial part of the subscapularis tendon (blue arrow). However, it is not possible to confidently identify it as a tear, because the tear is seen in only a single image slice and the image may have been affected by the partial volume effect. On the other hand, in the 2D 1-mm images, the tear in the subscapularis tendon can be evaluated in multiple image slices acquired in a direction parallel to the fibers (yellow arrows). It is also possible to confirm that there is no dislocation of the long head biceps tendon, and it is judged that the slightly higher intensity observed in the images is due to the magic angle effect.
Figure 2 shows a tear of the subscapularis tendon (yellow arrows), which is clearly depicted in the axial image (b) and sagittal MPR image (c). The presence of soft tissues, such as the superior glenohumeral ligament which prevent dislocation of the long head biceps tendon (green arrow) can also be confirmed. The findings seen in the 2D 1-mm images were consistent with the findings of arthroscopic examination.
Advanced intelligent Clear-IQ Engine (AiCE) is a noise reduction technique based on advanced Deep Learning Reconstruction technology developed by Canon Medical Systems Corporation. Using this Deep Learning technology, AiCE can create high-SNR images from low-SNR images.
A deep convolutional neural network (DCNN) is trained in advance to transform images with high noise levels into images with low noise levels. The algorithm obtained by training the DCNN is then installed in an MRI system, allowing noise in newly acquired low-SNR images to be removed to obtain high-SNR images. As the DCNN was trained using only the high-frequency noise components, noise reduction (denoising) can be achieved regardless of sequence.
When AiCE is used in combination with a rapid imaging technique, such as SPEEDER or Compressed SPEEDER, 2D images with a standard slice thickness can be acquired in a shorter scan time, and image quality can be improved compared to images acquired using conventional methods. In addition, by applying AiCE to images acquired with slightly longer scan time, image quality can be further improved and high-resolution and thin-slice 2D images can be obtained.
2. Clinical usefulness of 2D 1-mm images and MPR images
In the field of orthopedics, it can be difficult to detect and evaluate pathologies, such as rotator cuff tears, cartilage injuries, and labral tears using images with a standard resolution and slice thickness. Ideally, such injuries should be evaluated using high-resolution 2D thin-slice images (2D 1-mm images). 3D images can also provide detailed views, but 2D 1-mm images allow us to obtain a clearer understanding of the patterns and locations of these injuries.
Some of the key advantages of 2D 1-mm images are a reduction in the partial volume effect and the ability to evaluate pathology using multi-sectional images generated by multiplanar reconstruction (MPR). The tradeoff between a higher SNR and a longer scan time can be largely avoided by employing AiCE and a rapid imaging technique in combination. As for scan time, it may be shorter to scan 2D 1-mm images than 2D 3-mm images in three image planes. A further advantage of 2D 1-mm images over 3D images is that high-resolution imaging can be performed with excellent tissue contrast and with relatively less motion artifacts.
Figure 1 shows a comparison of 2D 3-mm images (a) and 2D 1-mm images (b) of a tear of the subscapularis tendon obtained using fat-suppressed proton density-weighted imaging (FS-PDWI) (coronal slices). In the 2D 3-mm images, there appears to be a tear in the most cranial part of the subscapularis tendon (blue arrow). However, it is not possible to confidently identify it as a tear, because the tear is seen in only a single image slice and the image may have been affected by the partial volume effect. On the other hand, in the 2D 1-mm images, the tear in the subscapularis tendon can be evaluated in multiple image slices acquired in a direction parallel to the fibers (yellow arrows). It is also possible to confirm that there is no dislocation of the long head biceps tendon, and it is judged that the slightly higher intensity observed in the images is due to the magic angle effect.
Figure 2 shows a tear of the subscapularis tendon (yellow arrows), which is clearly depicted in the axial image (b) and sagittal MPR image (c). The presence of soft tissues, such as the superior glenohumeral ligament which prevent dislocation of the long head biceps tendon (green arrow) can also be confirmed. The findings seen in the 2D 1-mm images were consistent with the findings of arthroscopic examination.
Figure 1: Comparison of 2D 3-mm images and 2D 1-mm images of the subscapularis tendon tear
Figure 2: MPR images (axial and sagittal images) generated from the original 2D 1-mm images (coronal image) shown in Figure 1
(Data courtesy of Dr. Ryuzo Arai, Department of Orthopedics, Kyoto Katsura Hospital)
Clinical cases in which 2D 1-mm images were found to be particularly helpful
Case 1 (Figure 3) is a patient with a complete tear of the supraspinatus tendon and subscapularis tendon and dislocation of the long head biceps tendon. In 2D 1-mm FS-PDWI, the subscapularis tendon appears to be completely torn (white arrows) and displaced toward the medial side of the long head biceps tendon (purple arrow). Full-thickness tear of the supraspinatus tendon can be confirmed (green arrow), and the image findings were consistent with the surgical findings. The coronal image shows the swelling of the long head biceps tendon (red arrows) and degenerative changes were found at surgery.
Figure 3. Case 1: Complete tear of the supraspinatus tendon and subscapularis tendon and dislocation of the long head of biceps tendon
Case 2 (Figure 4) is superior labrum from anterior to posterior (SLAP) leison. High signal intensity (orange arrows) are observed in the superior labrum in the coronal 2D 1-mm FS-PDWI (Figure 4a), and the sagittal MPR image (Figure 4b) clearly shows the extent of the labral injury in the anteroposterior direction (orange arrowheads).
In addition to the above cases, 2D 1-mm images were also found to be extremely useful in a patient with an injury of the anterosuperior part of the labrum and a paralabral cyst in the hip joint. The lesions could be clearly visualized and the continuity of the labral injury and the paralabral cyst could be clearly seen in three image planes, allowing us to make a diagnosis with confidence.
In addition to the above cases, 2D 1-mm images were also found to be extremely useful in a patient with an injury of the anterosuperior part of the labrum and a paralabral cyst in the hip joint. The lesions could be clearly visualized and the continuity of the labral injury and the paralabral cyst could be clearly seen in three image planes, allowing us to make a diagnosis with confidence.
Figure 4. Case 2: SLAP lesion
Case 3 (Figure 5) is a tear of the radial collateral ligament of the metacarpophalangeal joint of the little finger. High signal intensity is observed in the radial collateral ligament in 2D 1-mm FS-PDWI (blue arrows). The radial collateral ligament is composed of the proper and accessory ligaments. The sagittal image (b) clearly shows that both of these ligaments are torn. The tears at the head of the metacarpal bone were confirmed at surgery.
Figure 5. Case 3: Tear of the radial collateral ligament of the metacarpophalangeal joint of the little finger
Future prospects of 1.5T DLR-MRI
Vantage Orian, a 1.5T DLR-MRI system manufactured by Canon Medical Systems Corporation, is currently in operation at Saiseikai Ibaraki Hospital, in Ibaraki, Osaka, Japan.
By applying AiCE to 2D thin-slice images, image quality comparable to 3T MRI can be achieved and fine structures can be clearly visualized. The acquisition time is slightly longer, but labral and cartilage injuries can be evaluated more easily than with 3D images.
Case 4 (Figure 6) is a tear in the body of the lateral meniscus. In 2D 3-mm T2*-weighted images (a), because only one image slice is available, it is not possible to confidently identify the presence of a tear, but in 1.2-mm 2D FS-PDWI (b), the tear can be confirmed in multiple slices of original image and MPR images (pink arrows). The axial image demonstrates a radial tear.
As discussed above, the use of AiCE can dramatically improve the diagnostic capabilities of 1.5T MRI systems, although 3T systems still offer some advantages in terms of resolution.
By applying AiCE to 2D thin-slice images, image quality comparable to 3T MRI can be achieved and fine structures can be clearly visualized. The acquisition time is slightly longer, but labral and cartilage injuries can be evaluated more easily than with 3D images.
Case 4 (Figure 6) is a tear in the body of the lateral meniscus. In 2D 3-mm T2*-weighted images (a), because only one image slice is available, it is not possible to confidently identify the presence of a tear, but in 1.2-mm 2D FS-PDWI (b), the tear can be confirmed in multiple slices of original image and MPR images (pink arrows). The axial image demonstrates a radial tear.
As discussed above, the use of AiCE can dramatically improve the diagnostic capabilities of 1.5T MRI systems, although 3T systems still offer some advantages in terms of resolution.
Figure 6. Case 4: Tear of the body of the lateral meniscus at 1.5T MRI
Summary
The ability to obtain high-resolution 2D thin-slice images (2D 1-mm images) is extremely useful in the field of orthopedics. The partial volume effect (which is a problem in images acquired with a standard slice thickness) can be minimized, the resolution limits of 3D images can be improved, and the effects of blurring can also be minimized. In addition, by applying MPR reconstruction to 2D thin-slice images, anatomical structures can be evaluated using multi-sectional images, as in 3D imaging.
Understanding anatomy and patterns of injury using 2D thin-slice images makes it easier to evaluate pathologies when using images with a standard slice thickness. Fatsuppressed proton density-weighted and T2-weighted 2D thin-slice images are extremely useful for the diagnosis of a wide range of orthopedic diseases and injuries. //
Understanding anatomy and patterns of injury using 2D thin-slice images makes it easier to evaluate pathologies when using images with a standard slice thickness. Fatsuppressed proton density-weighted and T2-weighted 2D thin-slice images are extremely useful for the diagnosis of a wide range of orthopedic diseases and injuries. //
* The contents of this report include the personal opinions of the author based on his clinical experience and knowledge.
* Deep learning technology is used in the design stage of image reconstruction processing in AiCE. The MRI system itself does not have self-learning capabilities.
Takahide Kakigi, MD, PhD
Department of Diagnostic Imaging and Nuclear Medicine, Kyoto University Graduate School of Medicine, Japan
Department of Diagnostic Imaging and Nuclear Medicine, Kyoto University Graduate School of Medicine, Japan
