Purpose: The accurate localization of bifurcation points in the brain near the circle of Willis is essential for labeling cerebral arteries and detecting potential aneurysms. Studies on the development of segmentation-based landmark localization methods for cerebral angiography are lacking. This study aimed to develop and validate a method for localizing anatomical landmarks using a three-dimensional (3D) encoder-decoder deep convolutional neural network architecture. Materials and Methods: Time-of-flight magnetic resonance angiography (MRA) images of 224 subjects obtained from publicly available datasets were used. Ten anatomical landmark points were annotated, and four different landmark subset configurations were trained and validated. For each configuration, 3D U-Net-based models were developed using the MRA images and their corresponding annotated landmarks. The deep learning prediction results were evaluated in terms of landmark localization errors. Results: Among the four configurations, the configuration with five landmark subsets produced the smallest landmark localization errors in the test dataset. Post-processing of the U-Net-predicted segmentation masks further reduced the mean localization errors across all landmark points for the configuration with five subsets. Conclusion: The proposed landmark localization method effectively identified major anatomical landmarks around the circle of Willis and showed the potential to automate segmental analyses of intracranial arterial tortuosity.
Acute lymphoblastic leukemia (ALL) primarily affects the bone marrow and peripheral blood. Extramedullary manifestations, especially in the sinonasal cavity, are rare and difficult to diagnose. We report a case of B-cell ALL in a 4-year-old girl who initially presented with a sinonasal mass mimicking a primary malignant neoplasm. The patient developed right eye protrusion upon nasal blowing over a period of 2 weeks. Imaging studies revealed an aggressive-appearing mass in the right maxillary sinus with extensive local invasion. The radiologic differential diagnosis included rhabdomyosarcoma, lymphoma, and other pediatric sinonasal malignancies. Histopathologic examination of the biopsy specimen and subsequent bone marrow evaluation established the diagnosis of B-cell ALL with BCR/ABL1 fusion. This case highlights the importance of including ALL in the differential diagnosis of sinonasal masses in pediatric patients, particularly when imaging features suggest an aggressive malignancy accompanied by hematologic abnormalities. A comprehensive evaluation, including imaging studies, routine laboratory tests, tissue biopsy, and bone marrow examination, is crucial for establishing an accurate diagnosis and initiating appropriate treatment.
Purpose: To characterize the electromagnetic (EM) behavior of a monopole and dipole hybrid (MDH) antenna and evaluate its performance relative to monopole and dipole antennas for 10.5 tesla (T) magnetic resonance imaging (MRI) applications. Materials and Methods: Full-wave EM simulations were performed to model an inductor-shortened 20-cm dipole and an MDH antenna with asymmetric 9-cm and 12-cm poles. Single elements and 16-channel arrays were arranged on cylindrical formers (inner diameter 25 cm) around a tissue-mimicking phantom (epsilon r = 49, sigma = 0.6 S/m, diameter = 18 cm, height = 30 cm) with similar to 4 cm inter-element spacing and no additional decoupling circuitry. The eight-channel monopole antenna and MDH arrays were further modeled and analyzed with and without coaxial feed cables to assess cable-induced effects on magnetic (B) field distribution. Relative individual transmit field maps of the 8-channel monopole antenna and MDH arrays were acquired on a 10.5 T MRI system using a gradient echo (GRE) sequence (repetition time [TR] = 4000 ms, echo time [TE] = 3.0 ms, acquisition time [TA] = 7:48 ms, flip angle = 60 degrees, field of view [FOV] = 354 x 354 mm(3), resolution = 3.0 x 1.5 x 3.0 mm(3)). Results: The MDH antenna demonstrated enhanced longitudinal B-field concentration and improved directivity compared with the dipole antenna element. Across the 8- and 16-channel arrays, the MDH structures consistently exhibited lower coupling, reducing S-21 by up to 10 dB relative to the monopole and dipole antenna arrays. Cable attachment significantly altered the MDH field patterns, reflecting dipole-like sensitivity to common-mode currents; however, appropriate cable management preserved field stability. Conclusion: The MDH antenna provides intrinsic decoupling advantages and robust multichannel behavior without requiring additional decoupling circuitry. These characteristics make the MDH architecture a strong candidate for next-generation high-density arrays for ultra-high field MRI.
Hepatic angiomyolipoma is a rare benign tumor that often mimics hepatocellular carcinoma on imaging. Here, we report a rare case of hepatic angiomyolipoma with extensive hemorrhagic degeneration, resulting in a large hemorrhagic component that mimicked a complicated cystic neoplasm. A 43-year-old woman presented with mild abdominal distension, and imaging revealed a 16-cm multiloculated cystic mass in the left hepatic lobe. Computed tomography revealed a thick, enhancing wall and septa with hyperattenuating internal fluid. On gadoxetic acid-enhanced magnetic resonance imaging, arterial enhancement was observed along the thick wall, and a subtle opposedphase signal drop indicated a small fat component that was initially missed. Laparoscopic left lobectomy was performed under suspicion of a malignant cystic neoplasm. Histopathological examination revealed an angiomyolipoma composed of thick-walled vessels, spindle-shaped muscle bundles, and scant adipocytes, with human melanoma black 45 positivity and absence of an epithelial lining, consistent with hemorrhagic degeneration. This case highlights how hemorrhagic degeneration can pose a significant diagnostic challenge by masking the typical features of hepatic angiomyolipoma, leading to diagnostic confusion with complicated cystic tumors.
Purpose: To assess the accuracy of supraspinatus muscle fat fraction and atrophy measured on the Y-view compared with the newly proposed fossa-view sagittal MRI plane. Materials and Methods: This study included 84 patients (36 male; mean age, 65.1 +/- 10.1 years) who underwent shoulder MRI with extended oblique sagittal T1-weighted and three-dimensional (3D) six-echo Dixon imaging between December 2020 and November 2022. The reference fat fraction was calculated by integrating voxel-wise Dixon values, while supraspinatus muscle volume was quantified using a 3D nnU-Net algorithm and normalized to the scapular volume to derive the standardized muscle index (SMI). Fat fraction and cross-sectional area were quantified on the Y-view and fossaview and compared with the reference values. Subgroup analyses were performed using fatty degeneration and retraction grades. Results: Agreement with the reference fat fraction was significantly higher for the fossa-view (intraclass correlation coefficient [ICC], 0.923) than for the Y-view (ICC, 0.822; p = 0.006). The fossa-view showed smaller deviations and narrower limits of agreement. For SMI, the Y-view (ICC, 0.782) showed higher agreement than the fossa-view (ICC, 0.694), although the difference was not statistically significant (p = 0.219). Subgroup analyses showed better Y-view performance at lower retraction grades and better fossa-view performance at higher grades, although the differences were not statistically significant (all p > 0.05). Conclusion: Both planes reliably quantified the fat fraction with greater accuracy in the fossa-view. However, single-plane assessment of muscle atrophy was less reliable, underscoring the need for MRI evaluation of the entire muscle.
Purpose: To evaluate several hybrid multi-element configurations to enhance B1+ efficiency, defined as the transmit magnetic field generated by the radiofrequency (RF) coil normalized by the net input power, and spatial coverage for 10.5 tesla (T) human brain imaging. Materials and Methods: Finite-difference time-domain simulations were performed using an anatomically detailed human head model to assess circularly polarized B1+ fields generated by 8-channel loop, dipole antenna, and monopole and dipole hybrid antenna (MDH) arrays. These baseline designs were extended to 16-channel arrays combining loops with dipole antennas or MDHs and to interleaved MDH-based arrays incorporating axially displaced dipole or monopole antennas. Results: The 8-channel MDH array yielded the highest central B1+ efficiency, exceeding loop and dipole antenna arrays by 38.5% and 33.3%, respectively, but peripheral coverage remained limited. The 16-channel loop + dipole antenna and loop + MDH arrays improved overall B1+ efficiency by up to 28.9% relative to their 8-channel counterparts. Interleaved combined structure arrays further enhanced peripheral excitation and superior-inferior coverage, with the T-shaped monopole antenna + MDH configuration achieving a high central efficiency. Conclusion: MDH-centered combined structure arrays effectively balance deep-field efficiency and peripheral coverage at 10.5 T. Interleaving monopole antenna-type elements beneath MDHs provides additional improvement and represents a promising design pathway for achieving more uniform excitation in 10.5 T brain imaging.
Purpose: Cerebellar maturation is critical for neonatal development and is often disrupted during preterm birth. This study used magnetic resonance (MR) fingerprinting to quantify age-related cerebellar changes and determine the specific impact of prematurity. Materials and Methods: This retrospective study included 31 neonates (22 preterm and 9 term; 17 males; mean gestational age, 249.0 +/- 20.7 days) who underwent brain MRI, including 2D multi-slice MR fingerprinting with 7 minutes 5 seconds scan time at term-equivalent age. The T1 and T2 relaxation times were measured in the middle cerebellar peduncle, dentate nucleus, and cerebellar cortex using MR fingerprinting-derived T1 and T2 maps. Group differences and correlations with postmenstrual age were assessed using independent t-tests and Pearson's correlation analysis. Results: Preterm neonates had longer T1 and T2 relaxation times in the middle cerebellar peduncle, dentate nucleus, and cerebellar cortex than did term neonates. In the middle cerebellar peduncle, T1 times were 1918 +/- 88 ms vs. 1760 +/- 68 ms (p < 0.001), and T2 times were 103 +/- 8 ms vs. 97 +/- 4 ms (p < 0.05). In the dentate nucleus, the T1 times were 2316 +/- 174 ms vs. 1976 +/- 144 ms (p < 0.001), and the T2 times were 141 +/- 18 ms vs. 122 +/- 11 ms (p < 0.05). In the cerebellar cortex, T1 was 2460 +/- 98 ms vs. 2192 +/- 127 ms (p < 0.001), and T2 was 152 +/- 14 ms vs. 127 +/- 23 ms (p < 0.05). Both T1 and T2 times negatively correlated with postmenstrual age across all regions (r ranged from-0.761 to-0.617, all p < 0.001). Prolonged relaxation times were associated with prematurity (odds ratio range: 1.03 to 1.16, p < 0.05). Conclusion: The MR fingerprinting-derived T1 and T2 relaxation times showed developmental changes with postmenstrual age and were higher in preterm neonates, indicating immaturity.
Extraskeletal Ewing sarcoma (EES) is a rare, malignant small round-cell tumor, and involvement of the spinal epidural space is unusual. Here, we report a case of spinal epidural EES in a 46-year-old man who presented with back pain. Magnetic resonance imaging revealed a dumbbell-shaped right paravertebral mass extending through the ipsilateral neural foramen, with focal bone marrow infiltration. After surgical resection and adjuvant chemoradiotherapy, progressive bone marrow involvement and vertebral metastasis were observed on follow-up imaging. This case highlights the diagnostic challenges of epidural EES and underscores the importance of recognizing bone marrow involvement, which is rare in EES.
Purpose: We evaluated differences in global and language-local networks in post-stroke aphasia associated with aging using diffusion tensor imaging (DTI)-based connectivity. Materials and Methods: Global and local metrics were extracted from deterministic tractography in fluent (n = 19; median age 60.0 years [interquartile range, IQR, 53.0-68.0]) and non-fluent (n = 38; median age 61.0 years [IQR, 50.0-64.8]) aphasia patients. Brain-age estimation was performed using a pre-trained deep learning model from T13D data. Chronological age and brain-age estimation were used as control factors to find the distinct patterns of network characteristics between the groups, along with other factors such as sex, time from onset, total intracranial and gray matter volume. Results: Brain structure age estimation was 66.66 years (IQR, 62.74-70.13) for fluent and 72.14 years (IQR, 66.99-76.85) for non-fluent aphasia patients. There are no significant differences in chronological age (p = 0.859), but significant differences in brain-age estimation (p = 0.004 < 0.05). Our study revealed distinct network patterns between the groups. Regarding global metrics, higher clustering coefficient values were found in fluent (median 0.0113 [IQR, 0.0101-0.0123]) compared to non-fluent individuals (median 0.0094 [IQR, 0.0090-0.0106]) with family-wise error (FWE) p-value (pFWE) < 0.05. These differences were not retained when adjusting for brain-age (pFWE > 0.05). For local metrics, there was a lower clustering coefficient at right superior temporal gyrus (STG-R) and a higher degree at left STG (STG-L) in the fluent compared to the non-fluent group (pFWE < 0.05). These differences are more pronounced when incorporating age, brain-age, and multiple comparison corrections. Conclusion: We observed the central hub role of STG-L, along with signs of neurological compensation reflected in the distribution of the STG-R in aphasia. The significant difference remained robust after correction for multiple comparisons and adjustment for age and brain-age, highlighting these nodes as a key discriminator between fluent and non-fluent aphasia.
Purpose This study aimed to determine the normative reference values of the Evans Index (EI) using magnetic resonance imaging (MRI) and to evaluate its relationship with sex and age. Materials and Methods This retrospective study reviewed brain MRI images reported as normal from a multispecialty hospital in Southwestern Nigeria, acquired between August 2024 and February 2025. A total of 320 apparently healthy individuals aged 2-83 years were included in this study. The maximum anterior horn width (MAHW) and maximum intracranial diameter (MICD) were measured using electronic calipers on the hospital's picture archiving and communication system workstation, and the EI was then calculated. All measurements were performed by a single observer. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 23.0. Results The study included 172 females (53.75%) and 148 males (46.25%), with an average age of 33.07 +/- 17.33 years. The mean values for the MAHW, MICD and EI were 31.63 +/- 2.72 mm, 127.60 +/- 5.80 mm, and 0.25 +/- 0.02, respectively. Weak and statistically non-significant correlations were observed between age and the various parameters: MAHW (p = 0.153), MICD (p = 0.267), and EI (p = 0.337). However, there was a small to moderate, statistically significant difference between the sexes, with males showing higher values for all parameters: MAHW [t(292.78) = 5.60, p < 0.001, mean diff. = 1.644], MICD [t(318) = 5.41, p < 0.001, mean diff. = 3.373], and EI [t(318) = 3.22, p = 0.001, mean diff. = 0.0063]. The intraclass correlation coefficients for all measurements demonstrated excellent reliability, with values exceeding 0.90. Conclusion This study established normative EI values for a Southwestern Nigerian population using MRI. These normative values are clinically important, as they assist in the diagnosis of hydrocephalus and neurodegenerative disorders.
Purpose: To determine the efficacy of apparent diffusion coefficient (ADC) values and ratios (lesion/liver tissue, lesion/splenic tissue, lesion/paraspinal muscle) in differentiating benign from malignant solid liver lesions. Materials and Methods: This study retrospectively analyzed data from 115 patients with solid liver lesions who underwent abdominal magnetic resonance imaging (MRI) at a single institution between January 2023 and December 2024. Lesions were classified as benign or malignant based on biochemical tests as well as radiographic and/or histopathologic findings. ADC values and ratios were determined using a 1.5 T MRI scanner. Quantitative variables are presented as mean +/- standard deviation or median (interquartile range). Receiver operating characteristic (ROC) analysis was used to determine the cut-off values for ADC value and ratio, for which associated areas under the ROC curve were calculated. Results: The present analysis included 115 lesions-36 benign and 79 malignant. The median ADC value of the benign lesions was significantly higher than that of malignant lesions: 1744.5 & times; 10-6 mm2/s vs. 1168.0 & times; 10-6 mm2/s, respectively. The average lesion-toliver ADC (rADCl), lesion-to-spleen ADC (rADCsp), and lesion-to-paraspinal muscle ADC (rADCm) ratios for the benign lesions were significantly higher than those of malignant lesions: 1.79 vs. 1.09, 2.31 vs. 1.44, and 1.19 vs 0.80, respectively. A threshold of 1416 & times; 10-6 mm2/s was used to differentiate benign vs. malignant lesions, with a sensitivity of 83.3% and a specificity of 78.5%. The cut-off values for rADCl, rADCsp, and rADCm were 1.55, 1.95, and 0.97, respectively, with sensitivities of 69.4%, 69.4%, and 83.3% and specificities of 87.3%, 91.1%, and 79.9%, respectively. Conclusion: ADC metrics obtained from diffusion-weighted MRI effectively distinguished benign from malignant solid liver lesions.
This study aimed to reproduce, through MRI simulation, the process by which gradient fields are applied during signal acquisition to obtain spatial information in MR signals. A digital phantom containing three cylinders filled with the simulated material was used to acquire MR signals by gradient echo using MRI simulations, performed using a pulse-sequence diagram. MR signals were acquired under three different conditions: a) without a gradient field applied, b) with only a gradient field applied, and c) with a dephasing lobe and gradient field applied. The MR signal acquired under (a) was time-domain data that, after Fourier transformation (FT), yielded a single spectral data corresponding to the resonance frequency. However, when gradient field is applied to obtain spatial information, phase dispersion (dephasing) inevitably occurs over time; furthermore, this dephasing is superimposed. Consequently, the MR signals acquired in b) and c) were confirmed to be frequency-domain data after FT, as they were obtained under gradient field application. Furthermore, the necessity of the dephasing lobe was confirmed, as data reflecting the nature of the k-space could not be obtained without its inclusion during gradient field application during signal acquisition.
The increasing number of spinal surgeries has made postoperative magnetic resonance imaging (MRI) essential, but metallic implants frequently cause severe artifacts due to inhomogeneity in the main magnetic field (B0). This review summarizes the underlying mechanisms and imaging features of metal artifacts, such as signal loss, signal translation, signal pile-up, and failure of fat suppression, and outlines the contributing factors, including implant material, shape, orientation, and magnetic-field strength. Reduction of metal artifacts in postoperative spine MRI is particularly challenging because of the preference for T2-weighted imaging, bilateral implant placement near critical structures, and limitations of spine coils that exacerbate field inhomogeneity and reduce image quality in comparison with extremity imaging. Reduction strategies include fast spinecho imaging, use of high receiver bandwidths, increased spatial resolution, and fat suppression with short tau inversion recovery or Dixon techniques. Special sequences, such as view angle tilting and three-dimensional (3D) multispectral imaging (Slice Encoding for Metal Artifact Correction and Multi-Acquisition with Variable Resonance Image Combination SeLective), are reviewed, including their limitations. A recommended protocol combining conventional and special sequences is proposed, emphasizing view angle tilting for axial images and 3D multispectral imaging for sagittal images, when feasible. Emerging technologies, such as low-field MRI and artificial intelligence-based reconstruction, represent promising future directions for metal-artifact reduction.
Purpose: Accurate measurement of brain volume using structural magnetic resonance imaging (MRI) is crucial for diagnosing and monitoring of Alzheimer's disease. This study aimed to clinically validate a new MRI-based brain volumetric software (the study software) against a widely used and validated reference software. Materials and Methods: Patients with cognitive impairment who underwent three-dimensional structural brain MRI between January and December 2021 were retrospectively included. The software's performance in detecting significant medial temporal lobe atrophy (MTA), along with the correlation between Scheltens' MTA scores and hippocampal volumes, was assessed using a non-inferiority test and Spearman's correlation analysis, respectively. MTA scores were determined by consensus among three neuroradiologists. Results: A total of 271 patients (mean age 79 +/- 7 years [standard deviation], 178 women) were included. The hippocampal volumes calculated by the two software programs were not equivalent. In detecting significant MTA, the study software demonstrated a non-inferior area under the receiver operating characteristic curve (AUC) (0.78 [95% confidence interval (CI): 0.73, 0.84]), accuracy (0.77 [95% CI: 0.72, 0.82]), and specificity (0.89 [95% CI: 0.84, 0.93]) compared to the reference software's AUC (0.71 [95% CI: 0.65, 0.78]), accuracy (0.72 [95% CI: 0.67, 0.78]), and specificity (0.80 [95% CI: 0.73, 0.85]) (p < 0.05) However, non-inferiority in sensitivity was not established between the study software (0.49 [95% CI: 0.38, 0.61]) and reference software (0.55 [95% CI: 0.44, 0.66]). The cor-relation between MTA scores and hippocampal volume was significantly stronger in the study software (-0.57 [95% CI: -0.64, -0.48]) than in the reference software (-0.30 [95% CI: -0.41, -0.19]) (p < 0.05) Conclusion: The new East Asian cohort-based MRI brain volumetric software demonstrated non-inferiority to the reference software in detecting significant MTA. Furthermore, the new study software showed a greater association with the neuroradiologists' MTA scores than the reference software.
Purpose: This study aims to evaluate the impact of the sheath current from coaxial cables on the transmission magnetic (B-1(+)) field efficiency of dipole antenna arrays for 10.5-tesla (T) magnetic resonance imaging (MRI). Materials and Methods: Electromagnetic simulations and phantom experiments were conducted using 8-and 16-channel dipole antenna arrays. Each configuration varied in terms of the coaxial cable gap (2 cm and 7 cm) and the presence or absence of floating cable traps. All antennas were tuned to 447 MHz/10.5 T. The simulated and experimental B-1(+) field efficiencies (B-1(+) field normalized by the square root of the input power) of multiple arrays were obtained and compared using an actual flip-angle imaging sequence in a cylindrical phantom with tissue-equivalent properties. Results: Simulated and experimental measurements of 8-and 16-channel dipole antenna arrays for ultra-high field (UHF) MRI showed excellent agreement, with less than +/- 10% deviation in B-1(+) efficiency. The highest efficiency (0.60 mu T/root W) was achieved with 7 cm coaxial cable spacing. When the spacing was reduced to 2 cm, the B-1(+) efficiency decreased by up to 35%. Trap removal and suboptimal cable routing further degraded performance. Comparisons across the loop, dipole, and loop + dipole antenna arrays confirmed that the dipole antenna arrays are more sensitive to cable interference. Conclusion: Coaxial cable spacing and trap existence play critical roles in determining the performance of dipole antenna arrays in UHF MRI. Effective cable management and optimized trap integration are essential for maintaining high B-1(+) efficiency in compact multichannel array designs. These findings highlight the importance of coaxial-cable management and trap design for achieving optimal transmission performance in practical MRI systems.
Purpose: This study aims to evaluate sleeve antennas that can reduce electromagnetic coupling compared with conventional dipole antennas while maintaining comparable performance at 10.5 tesla (T). Materials and Methods: We designed and fabricated dipole and sleeve antennas with matching resonant frequencies (447 MHz at 10.5 T) and similar geometric configurations, enabling a fair comparison. The sleeve antenna design consisted of monopole conductors with inductive shortening and 5-cm floating cable traps, allowing a total length of 20 cm. Two-element coupling was measured over a range of 4 to 14 cm on a phantom with tissue-equivalent properties (epsilon r = 49, sigma = 0.6 S/m). Additionally, eight-channel dipole and sleeve antenna arrays were assembled on 3D-printed formers with identical elliptical dimensions and an average inter-element spacing of 8.8 +/- 1.4 cm. Initially, bench measurements were conducted to evaluate the S-parameters of the dipole and sleeve antennas. Noise covariance matrices and relative B1+ field (transmit magnetic field) maps of each array were obtained through 10.5-T magnetic resonance imaging. Results: The sleeve antenna consistently showed lower coupling than that of the dipole antenna across all distances, with S21 values reaching below-7.2 dB at 4 cm, compared to-5.1 dB for the dipole. In the eight-channel array configuration, both antenna types exhibited similar S11 values, indicating adequate impedance matching. However, the sleeve array demonstrated improved adjacent-element isolation (lower S21) and reduced noise covariance (<= 0.07), suggesting greater channel independence. The relative B1+ field maps revealed independent transmit field patterns for both arrays, enabled by sufficient inter-element spacing. Conclusion: The sleeve antenna array provides superior decoupling and lower inter-element noise correlation than that of a conventional dipole antenna. Its simplified structure and performance advantages make it a promising solution for scalable, high-density, ultrahigh-field magnetic resonance imaging transmit/receive arrays.
Purpose: This study compared the accuracy of manganese-enhanced magnetic resonance imaging (MRI) using hollow manganese silicate (HMS-MRI) to gadolinium-enhanced MRI (Gd-MRI) in predicting lesion size and residual tumor cellularity after chemotherapy using in vivo and in vitro experiments. Materials and Methods: BT474 cells and orthotopic xenograft mouse models of human breast cancer, based on BT474 cell inoculation, were used for the in vivo and in vitro experiments, respectively. Eighteen female mice underwent Gd-MRI and HMS-MRI before and after chemotherapy. The tumor size and residual tumor cellularity obtained from Gd-MRI and HMS-MRI after treatment were compared with the pathological findings. Differences in residual tumor size and cellularity measured by MRIs also correlated with the pathological findings. Two sets of BT474 cell pellets treated with or without chemotherapy were imaged after the addition of manganese chloride (MnCl2) or gadolinium. Results: Agreement in both tumor size and residual tumor cellularity between HMS-MRI and pathology was stronger than that between Gd-MRI and pathology. The lower the tumor-stroma ratio in pathological specimens, the larger was the difference in residual tumor cellularity between Gd-MRI and HMS-MRI (r =-0.62, p = 0.018). BT474 cell pellets showed signal enhancement after exposure to MnCl2, and the degree of enhancement decreased when the cell pellets were exposed to chemotherapeutic agents. Conclusion: HMS-MRI reflects residual tumor cellularity more accurately than Gd-MRI, especially in cases with low residual tumor cellularity.
