0.55T qualitative and quantitative knee MRI data for iHEALTH and A4IM projects

Information for Users

Scanner

The MRI scanner used to acquire data is the MAGNETOM Free.Max 0.55T commercial system manufactured by Siemens Healthineers (Erlangen, Germany), FDA cleared in July 2021 (510(k) Number: K210611). The scanner was designed under the constraints of being compact, lightweight, cost-effective, and indicated for orthopedic practices and routine musculoskeletal imaging applications.

The scanner employs a superconducting magnet with DryCool technology to generate B₀ ≈ 0.55 T at the field of view. The magnet assembly weighs less than 3.5 metric tons with a system height of less than 80 inches. The bore diameter is 80 cm, optimized for patient comfort and accessibility, beneficial for patients with claustrophobia or anatomical challenges. The large bore also facilitates knee imaging with flexible coil positioning.

The gradient coil geometry employs actively shielded gradient technology with the following specifications: A maximum gradient amplitude of 22-26 mT/m, maximum slew rate of 45-55 T/m/s, a Field of View (FOV) up to 50 cm. The Tx/Rx RF coil used to acquire knee data is the Contour S Coil (Part Number: 11407176), a commercial 6-channel flexible receive coil manufactured by Siemens Healthineers (Erlangen/Forchheim, Germany). 

The Contour S Coil employs a blanket-like flexible design with a dimension of 45 cm × 27 cm (17.7 in × 10.6 in) that provides patient comfort by adapting to body contours without the need for straps. The coil offers ultra-lightweight construction with soft materials. The flexibility allows the coil to conform closely to the knee joint, maximizing signal-to-noise ratio (SNR) through improved coil-to-tissue coupling compared to rigid coil alternatives.

The coil operates at the proton Larmor frequency corresponding to the 0.55T field strength (≈23.6 MHz) and the 6-element phased array design enables parallel imaging acceleration for reduced scan times while maintaining diagnostic image quality.

Control system and data acquisition

The MR system is controlled through the manufacturer's proprietary software platform syngo MR XA60, which handles synchronized pulse-sequence execution and data acquisition. User interaction with the system is handled via the syngo MR user interface. The raw data is acquired in ISMRMRD format and the reconstructed images are available in DICOM format.

Subjects information

For this dataset, 20 subjects were recruited and both knees were scanned, resulting in a total of 40 knee images. Written informed consent was obtained from all volunteers prior to imaging. Anonymous information about the gender, age, height, and possible knee injuries of the volunteers scanned for this database is provided in Table 1.  The numbering of the subjects corresponds to the numbering specified in the folder and file names in this database.

Table 1. Additional anonymous information per subject.

Pulse sequences

For each volunteer, the complete exam can be separated into two groups of sequences: a) a qualitative protocol and b) a quantitative protocol. 

The qualitative protocol consists of running three consecutive 2D sequences with cartesian trajectories, and optimized parameters to obtain Proton Density (PD), T1-weighted, and T2-weighted images. Deep Resolve deep learning vendor software image enhancement was applied to each sequence, resulting in a total of 6 datasets. The quantitative protocol consists of acquiring two sets of images: (1) 2D Modified Look-Locker inversion (MOLLI) (3-3-5) to obtain a T1 map, and (2) and  T2 prepared balance steady state free precession (T2p-bSSFP) sequence was acquired to obtain a T2 map. The 2D T1 and T2 maps were obtained with the vendor software package using DL for the T1 and T2 contrasts by fitting the 3-parameter model for MOLLI images and the T2 decay model for T2p-bSSFP images. The calculated maps and their corresponding image contrasts can be found in this dataset.

In all the cases the images were acquired  in the sagittal plane for better knee anatomic visualization. Each image set is acquired for the same subject in the same session for both knees, and the protocol requires approximately 8:42 minutes of sequence execution time per knee, giving a total time of 17:24 min per volunteer. 

The acquisition time per sequence performed for both types of protocols and per knee are listed in Table 2. The main acquisition parameters of the sequences performed for both protocols are listed in Tables 3 and 4. Example representative images for the qualitative and quantitative protocols are shown in Figure 1 and Figure 2.

Table 2. Acquisition time per sequence

Table 3. Sequence parameters for quantitative protocol

Table 4. Sequence parameters for quantitative protocol

Figures can be found in the JPG files attached below:

Figure 1. Reference images for a) Proton density (PD) knee sagittal image, b) PD sagittal image with deep resolved, c) T1-weighted sagittal image, d) T1-weighted sagittal image with deep resolved, e) T2-weighted sagittal image, f) T2-weighted sagittal image with deep resolved. 

Figure 2. Reference images for a) knee sagittal 2D MOLLI T1 map, b) knee sagittal 2D T2p-bSSFP T2 map.