Thirty-seven participants were recruited for this study, and 35 participants completed all study requirements (female: n = 13; mean age ± SD = 25.4 ± 6.8 years; mean height ± SD = 180.7 ± 7.0 cm, body mass ± SD = 82.3 ± 8.9 kg; male: n = 22, mean age ± SD = 23.2 ± 4.6 years, mean height ± SD = 168.2 ± 5.3 cm, mean body mass ± SD = 68.7 ± 11.2 kg). Participants were required to be 18 years or older and free from any lower extremity injury within the previous one month or any leg/foot surgery within the previous year. Participants were from the university community including students, staff and faculty. In addition, participants who could not safely receive the MRI scan due to the presence of ferrous-magnetic metal objects within the body, fresh tattoos, a pacemaker, or an implantable cardioverter defibrillator were excluded. Two participants were dropped from the study because of these exclusion criteria.
Each participant provided informed consent by reading, asking relevant questions, and signing the informed consent form approved by the University’s Institutional Review Board (Brigham Young University’s Institutional Review Board of the Human Research Protection Program, study protocol, IRB2019-375). This Institutional Review Board approved all experimental protocols. All methods were carried out in accordance with relevant guidelines and regulations (Federal regulation: 45 CFR 46.111). Each participant completed a safety screen before any MRI testing. Each participant attended one MRI session and one US session. The testing session order was randomized, with each session completed within an hour of one another.
Imaging preparation
Muscle imaging included scanning the flexor hallucis brevis (FHB), abductor hallucis (ABDH), flexor digitorum brevis (FDB), quadratus plantae (QP), and abductor digiti minimi (ADM). To identify the FHB muscle, a reference mark was made proximal to the head of the first metatarsal, at 10% of the truncated foot length [33]. To image the other muscles, four reference marks were made perpendicular to the longitudinal axis of the foot’s medial, lateral, dorsal, and plantar surface. The medial reference mark was positioned on the navicular tuberosity for both the MRI and US imaging. The other three reference marks were made in the coronal plane transversely across the foot in alignment with this medial mark. Each mark was made on the skin using a marker (Fig. 1). Thus, four muscles were measured in line with the navicular tuberosity and one muscle was measured at 10% of the truncated foot length. All muscles were measured using the short axis (coronal) image for MRI and US to provide two-dimensional data.
MRI scans
A 3 Tesla magnet (TIM-Trio 3.0T MRI, Siemens, Erlangen, Germany) was employed using a 3-dimensional spoiled gradient echo sequence prescribed with the through-plane direction perpendicular to the long axis of the entire foot. Slice thickness was 1mm, slices were contiguous with no gap or overlap between them. Scan parameters were: TE/TR = 10.8/4.9 ms; matrix size: 416 × 300 × 288 pixels; resolution: 0.4 × 0.4 × 1 mm; Field of view: 150 × 108 × 288 mm; flip angle: 15 degrees; acceleration factor: 2; bandwidth: 130 Hz/pixel; total acquisition time: 4:42.
Participants completed an MRI safety screening before entering the magnet room. Before testing, fish oil capsules (Member’s Mark, Sam’s West Inc., Bentonville, Arkansas) were attached to the participant’s skin via double sided Velcro over each reference mark. The capsules were positioned with the long axis parallel to the coronal plane (Fig. 1) and were visible on the MRI images. Each capsule served as a fiducial marker so each MRI scan could be taken at the correct location. The right foot was scanned first. To do so, it was placed in an 8-channel foot/ankle coil (ScanMed, Omaha, NE, USA). The left leg was scanned second, each foot was scanned once. Light sandbags and wedges were employed to minimize foot movement during imaging.
The MRI data were captured as a large-block and formatted to appear as multiple images. The image that best intersected the center of the reference fish oil capsules was selected for data analysis (Fig. 2). Due to the 1mm slice thickness the center of the fish oil capsules on MRI could only be identified with a precision of ± 0.5mm.
US imaging
Participants were seated in a comfortable back-supported position on a treatment table. During imaging, the participant’s hip joint was externally rotated slightly to allow access to the foot’s plantar surface. A bolster was placed underneath the participant’s knee, with the ankle set at approximately zero degrees dorsiflexion throughout imaging.
All images were taken by the same imager who has two years of scanning experience with specialized training of intrinsic foot muscle imaging. Each US muscle image was collected using a ML6-15‐D matrix linear transducer probe (LOGIQ S8; GE Healthcare, Chicago, IL). Settings were set to optimize image quality. Scanning depth (3 cm), frequency (8 MHz), focal position, and time‐gain‐compensation were kept constant.
For each US CSA measurement, two separate cine loops were recorded per muscle per foot with the transducer probe removed from the participant’s foot and repositioned between cine loop recordings. The transducer probe was placed transversely to the long axis of the foot to obtain a 2-dimensional slice within the coronal plane on the medial-plantar side of the foot over each corresponding mark (Fig. 3). For the FHB, the muscle body was first found by locating the flexor hallucis longus tendon, sesamoid bones and first metatarsal head. The probe was then moved proximally over the FHB reference mark. The ABDH muscle was imaged at the medial midfoot reference mark using the navicular tuberosity. For the FDB, and QP the transducer probe was positioned on the foot’s plantar midfoot surface and aligned with the reference marks. The ADM muscle was imaged using the same short axis plane as the ABDH, FDB and QP with the transducer probe positioned at the lateral-plantar midfoot reference mark.
During imaging, participants were asked to contract specific muscles using toe flexion and toe spreading movements. Cine loops recorded the isolated contractions to help identify the muscle’s fascial borders and highlight any conformational muscle shape changes [34]. Each muscle was imaged in a relaxed state, in a contracted state, and then again in a relaxed state. All data were collected from the images in the relaxed state. Two separate recordings of the relax-contract-relax cycle were captured for each muscle. The transducer probe was briefly removed from the foot between recordings. Figure 4 presents a side-by-side comparison of muscle measurements via US and MRI.
Data processing
Two researchers performed the image processing to obtain CSA muscle measurements for each muscle. One researcher performed all of the MR image processing, while another performed all of the US image measurements. These researchers did not compare results during the image processing, thus they were blinded to each other’s results.
MRI scans for all 35 participants were loaded into Osirix (Pixmeo, Geneva, Switzerland) to determine the muscle’s CSA. CSA measurements were taken in the coronal plane from two adjacent slices for each of the five muscles assessed in this study in the foot as shown in Fig. 4 at the reference marks shown in Fig. 3, and averaged. Measurements were repeated twice consecutively for each of the five specific muscles within each slice for a total of two measurements per muscle as shown in Fig. 4. This procedure was repeated for the contralateral foot.
Ultrasound CSA measurements for all participants were obtained from two separate cine loop recordings for each of the five muscles for each foot. A single image was chosen from each cine loop while the muscle was at rest. Images of the muscle’s fascial border were captured and analyzed using software on the LOGIQ S8 machine via previously established methods [34, 35]. Muscles were imaged on the inside of the muscle fascia border. Measurements were repeated twice for each of the five specific muscles. See Fig. 4. One measurement was taken from the first cine loop and another measurement from the second cine loop. These two measurements were averaged and then compared to the average MR muscle measurements to determine inter-method validity. The two measurements were recorded separately to evaluate intratester reliability.
Statistical analysis
Pearson product moment correlations were employed to determine whether the US system generated valid mean CSA results as compared with the MRI system. Intraclass-correlation coefficients (ICC3,1) were calculated to establish reliability using CSA measurements from each MRI and US image. To assess image segmentation repeatability, we chose the ICC model with fixed raters and random subjects. To identify testing error inherent to each imaging modality, we calculated the standard error of the measurement (SEm), a 95% confidence interval, and the minimum detectable difference (MDD) for both MRI and US using the following equations:
$$\mathrm{SEm}=\mathrm{SD}\;\left(\mathrm{Sq}\;\mathrm{rt}\;1-\;{\mathrm r}_{\mathrm{ICC}}\right)$$
$$95\%\;\mathrm{CI}\;\mathrm{SEm}\;=\;\mathrm{muscle}\;\mathrm{mean}\;\pm\;\left(1.96\;\ast\;\mathrm{SEm}\right)$$
$$\mathrm{MDD}\;=\;\mathrm{SEm}\;\ast\;1.96\;\ast\;\surd2$$
Bland-Altman plots were also generated to graphically highlight CSA difference and mean CSA between the MRI and US data, and to visualize any potential systematic error pattern or trends. The x-axis (mean CSA) on the Bland-Alman plots represents the average CSA from the MRI and US data (mean CSA = [MRI CSA + US CSA] ÷ 2) for each participant (n = 35). The y-axis (CSA difference) on the Bland-Alman plots represent the absolute difference between the CSA values (CSA difference = MRI CAS - US CSA) for each participant. We then calculated the percent muscle size based on the limits of agreement to help understand the variation between US and MRI derived CSA. Statistical analyses were performed using SPSS version 27.0 statistical software (IBM Corporation, Armonk, NY). An alpha of 0.05 was employed to determine statistical significance.