INTRODUCTION
Multiple sclerosis (MS) is a chronic autoimmune inflammatory disorder of the central nervous system (CNS) characterized by demyelination, axonal damage, and neurodegeneration [1-3]. Approximately 2.8 million individuals are affected by MS worldwide [4]. It occurs predominantly in the 20-40 age range, and it affects women more often than men [2, 5, 6].
A hallmark feature of MS pathology is the formation of focal inflammatory lesions in the CNS located in the brain and spinal cord, mostly within the White Matter of the CNS [7, 8]. The inflammatory process causes interference in the normal transmission of signals along the nerve axons, which may give rise to a variety of neurological symptoms, including fatigue, altered sensation, reduced motor function, and cognitive impairment [9-11]. MRI plays an important role in diagnosing and managing patients with MS [10, 12].
Typical MRI sequences used to identify and characterise the location, extent, and morphology of MS lesions are T1, T2-weighted sequences, and Fluid Attenuated Inversion Recovery (FLAIR) sequences. [13, 14] FLAIR MRI sequences show great sensitivity for demonstrating the presence of periventricular and juxtacortical lesions, which are commonly seen in patients with MS. [10, 13, 15].
Although conventional MRI has played an essential role in detecting multiple sclerosis (MS) lesions, challenges exist with identifying and differentiating MS lesions using MRI [16]. The traditional fluid attenuated inversion recovery (FLAIR) sequence may have limited sensitivity for certain lesions, including cortical and subcortical lesions, which are increasingly recognized as important contributors to MS pathology. [16 -18] Additionally, distinguishing MS lesions from other white matter abnormalities, such as vascular lesions and nonspecific hyperintensities, can lead to delays in identifying the correct clinical diagnosis and affect how patients are managed [16, 19].
To overcome these challenges, advanced MRI techniques have been developed to enhance the detection and characterization of MS lesions. One of these advanced MRI techniques is the double inversion recovery (DIR) sequence, which will permit the selective suppression of signals from both cerebrospinal fluid (CSF) and the brain's white matter (WM), which will result in a better contrast between the lesion and the background image of the brain [16, 17, 20, 21]. However, DIR is limited by its relatively long acquisition time (an average of 7 minutes depending on the technique and parameters) and its lower signal-to-noise ratio (SNR) compared with the conventional FLAIR sequence, as the multiple 180° inversion pulses reduce the overall SNR of the image (Fig.
1) [22-25].
To address these limitations, we explored the use of fused images combining FLAIR and WAIR sequences, referred to as “MinPE FLAIR/WAIR” as a potential tool for improving the detection of MS lesions. By integrating complementary information from the FLAIR sequence, which suppresses the signal from CSF, and the WAIR sequence, which suppresses the signal from white matter, the “MinPE FLAIR/WAIR” method offers the potential for improved lesion conspicuousness and diagnostic accuracy in MS [22, 26, 27]. However, its clinical utility in routine practice remains to be fully established, and further research is needed to validate its efficacy and clarify its role in the diagnostic workup of MS [16, 28].
This study aimed to assess the diagnostic value of fused images generated using Minimum Pixel Value Extraction of FLAIR and WAIR sequences “MinPE FLAIR/WAIR”, in improving the detection of multiple sclerosis (MS) lesions compared with the conventional FLAIR sequence. Our findings demonstrate that MinPE FLAIR/WAIR provides greater detection rates for MS lesions in the cortical, juxtacortical, periventricular, and subcortical regions than the conventional FLAIR sequence.
MATERIALS AND METHODS
Study Design
A retrospective observational study was conducted to evaluate the effectiveness of fused Fluid-Attenuated Inversion Recovery (FLAIR) and white-matter-attenuated inversion recovery (WAIR) sequences (referred to as “MinPE FLAIR/WAIR”) in improving the detection of multiple sclerosis (MS) lesions. The study was performed at a single center, and the data were obtained from the institution’s MRI database. Ethical approval for this study was obtained from the Institutional Review Board of the College of Medicine, King Saud University (Ref.
No. 23/0315/IRB), and the requirement for patient consent was waived because of the retrospective study design. This article was reported in accordance with the STARD Reporting Checklist.
Study Population
This study enrolled 65 patients with multiple sclerosis (MS) who underwent brain MRI for suspected or confirmed MS at King Khalid University Hospital, Riyadh, Saudi Arabia, between January 2021 and December 2023. Using the Connors method for post-hoc sample size estimation and comparing our sample size to other similar studies, both methods established that about 63 subjects would be the minimum number of patients needed to test sufficient statistical power for detecting any difference in lesion detection rate. As a result, our final sample of 65 patients satisfied this requirement.
Eligible patients were aged ≥ 18 years and had a confirmed diagnosis of MS based on the 2017 McDonald criteria [29]. The following conditions precluded patients from participating: incomplete imaging data; poor imaging quality; comorbid conditions that might negatively affect MRI analysis, such as severe atrophy of the brain; significant cerebrovascular disease; and neurological disorders.
After applying these criteria, 65 patients met the inclusion requirements; five were excluded because of severe motion artifacts. The final cohort was predominantly female, comprising 47 women (72.3%) and 18 men (27.7%) (Fig. 2).
Cohort Characteristics
The mean age of the patients was 37.4 years (range, 18–65). The mean duration of MS since diagnosis was 7.5 years (range, 1–30). The cohort included the following MS subtypes: relapsing–remitting MS (RRMS), 40 patients (61.5%); secondary progressive MS (SPMS), 18 patients (27.7%); and primary progressive MS (PPMS), seven patients (10.8%).
The mean Expanded Disability Status Scale (EDSS) score was 3.2 (range, 1.0–8.5), indicating mild-to-moderate disability. Patients received various disease-modifying therapies (DMTs), including interferon-beta (n = 25, 38.5%), glatiramer acetate (n = 15, 23.1%), natalizumab (n = 10, 15.4%), or no treatment (n = 15, 23.1%).
Imaging Protocol
All MRI scans were acquired using a standard 12-channel head coil. The imaging protocol included conventional axial FLAIR and WAIR sequences. The total scan time was 2:29 min for the FLAIR sequence and 1:52 min for the WAIR sequence, giving a combined acquisition time of approximately 4 min and 21 s.
This is notably shorter than the approximately 7 min required for a DIR sequence, which is known for its superior lesion detection through selective suppression of white matter and cerebrospinal fluid signals. The parameters used for the FLAIR sequence were as follows: repetition time (TR) of 8000 ms, echo time (TE) of 95 ms, inversion time (TI) of 2750 ms, field of view (FOV) of 230×230 mm, slice thickness/spacing of 5/1 mm, an image matrix of 320/224, echo train length (ETL) of 12, and one signal average. The total scan time was 2:29 minutes, while the parameters used for the WAIR sequence were as follows: repetition time (TR) of 6000 ms, echo time (TE) of 52 ms, inversion time (TI) of 300 ms, field of view (FOV) of 230×230 mm, slice thickness/spacing of 5/1 mm, an image matrix of 320/224, echo train length (ETL) of 12, and one signal average, with a total scan time of 1:52 minutes.
When fusing two image sequences, the source images must share the same spatial resolution to ensure accurate voxel-wise integration. In practice, any spatial resolution may be used, provided that it is consistent across both sequences before fusion.
Image Analysis
The MinPE filter is an image processing technique based on minimum pixel-wise operations used to combine images. This operation takes two or more input images and generates an output image in which the value of each pixel is the minimum of the corresponding pixels from all input images. Mathematically, for two input images, A and B, the minimum pixel-wise operation is represented as:
C(x, y) = min(A(x, y), B(x, y))
where (C(x, y)) is the output pixel at coordinates (x, y), and (A(x, y)) and (B(x, y)) are the corresponding pixels in input images A and B, respectively. The minimum operation can be helpful in several image processing applications, such as determining common or overlapping areas between two images, reducing noise by averaging the values from two or more noisy images that are acquired, and creating masks using the results of segmentation on both images. The MinPE FLAIR/WAIR images were generated by applying the minimum function between pixel values of the FLAIR and WAIR sequences of corresponding pixels.
Therefore, the resulting pixel values of each MinPE pixel correspond to the minimum value of the two respective pixels (i.e., pixels from FLAIR and WAIR) from the two respective sequences. This operation was performed using a dedicated image combination tool on a GE Advantage Windows workstationV7.1.7. This method is independent of the scanner type because it is fundamentally a post-processing procedure.
The workflow relies on standard FLAIR and WAIR images and applies the same coregistration and analysis steps, regardless of the scanner vendor or model. Because all operations occur after image acquisition, the method does not depend on proprietary hardware or pulse-sequence implementations. Therefore, it can be replicated on a scanner as long as the corresponding image contrasts are available.
Coregistration Step
Before implementing the MinPE filter, the FLAIR and WAIR images were co-registered to ensure precise alignment. This step was performed using rigid body transformation algorithms in Functool software (version 9.4.05a) on a GE Advantage Windows workstation (v7.1.7) to correct for potential patient motion between the two sequences. Coregistration mitigated misalignment errors, ensuring accurate voxel-to-voxel correspondence between the FLAIR and WAIR images (Fig.
3). The registration pipeline is fully automated and does not require user-defined preprocessing steps or manual parameter tuning. This automated approach was deliberately selected to reflect routine clinical practice because it allows image fusion to be performed by technologists or radiologists without the need for advanced expertise in image processing.
The coregistration quality was assessed using systematic visual inspection, which is a widely used and recommended approach in clinical MRI workflows when rigid-body transformations are applied. For each dataset, we visually evaluated the alignment of the ventricular boundaries, cortical and subcortical white matter interfaces, and consistency of the lesion location across FLAIR and WAIR images. Datasets with visible misalignments were excluded, as shown in Fig.
S1.
Reader Evaluation and Quality Assessment
Two experienced neuroradiologists independently reviewed all MRI scans and were blinded to the clinical information. The two neuroradiologists had over 10 years of experience working with MRI. They evaluated whether or not MS lesions were present in pre-defined regions of the brain, such as cortical gray matter, juxtacortical white matter, subcortical white matter, periventricular white matter, corpus callosum, basal ganglia, thalamus & internal capsule.
To minimize potential reader bias and recall effects, all image datasets were anonymized and assigned numeric identifiers before evaluation. Conventional and fused images were independently presented to the neuroradiologists. Lesion detection was based on a visual assessment.
A qualified medical physicist assessed the image quality. Data sets demonstrating poor imaging detail, either caused by severe image artifacts or from motion between FLAIR and WAIR scans, are excluded from this study. To ensure the reliability and validity of the study findings, quality control measures were implemented throughout the data collection and analysis.
This included a careful review of MRI scans by trained neuroradiologists for image artifacts and consistent interpretation.
Statistical Analysis
Statistical analyses were performed using the IBM SPSS software (version 26). A priori sample size calculation was not performed because of the retrospective design of the study, which included all eligible MRI scans during the study period. However, a post-hoc power analysis was conducted to estimate the statistical power based on the observed effect size and the available sample of at least 60 patients, yielding a power of 90%.
Descriptive statistics were used to summarize patient demographics and lesion characteristics as frequencies, percentages, means, and standard deviations. The agreement between conventional FLAIR and “MinPE FLAIR/WAIR” sequences in lesion detection was assessed using the chi-square test, with Fisher’s exact test applied when more than 20% of cells had an expected count of zero. Statistical significance was set at p ≤ 0.05.
Inter-observer agreement for lesion detection was evaluated using Cohen’s kappa (κ), which quantifies the degree of agreement between the two readers. Kappa values were interpreted as follows: < 0.20 = poor, 0.21–0.40 = fair, 0.41–0.60 = moderate, 0.61–0.80 = good, and 0.81–1.00 = strong agreement. Post-hoc power analysis was performed based on the observed effect size for the primary comparison between MinPE FLAIR/WAIR and conventional FLAIR lesion detection rates in brainstem regions, where statistically significant differences were identified.
Using the observed effect size and sample size, the estimated study power was approximately 90% for detecting differences of this magnitude. This power estimate reflects the ability to detect large effect sizes in anatomically challenging regions and may not be generalizable to supratentorial regions where smaller effect sizes were observed.
RESULTS
In general, MinPE FLAIR/WAIR images showed improved detection rates compared with conventional FLAIR across several brain regions (Fig. 4). Although in supratentorial regions, MinPE FLAIR/WAIR demonstrated higher mean lesion counts compared with conventional FLAIR; however, most of these differences did not reach statistical significance.
Table 1 presents the demographic profiles of the 65 patients, including sex and age distribution. Of the total patients, 18 (27.7%) were male and 47 (72.3%) were female, indicating a higher representation of females in the group. The age distribution showed that 27.7% of patients were under 25 years of age, the largest proportion (35.4%) was between 25 and 30 years of age, 24.6% were between 30 and 40 years of age, and 12.3% were older than 40 years of age.
This reflects a relatively young patient cohort, with most patients aged 25–30 years. The overall age range was 17–57 years, with a mean age of 29.96 years and a standard deviation of 2.11, indicating minimal variation around the average.
A comparison between the FLAIR and MinPE FLAIR/WAIR images was performed using the chi-square test. The inter-observer agreement for each imaging method (FLAIR and MinPE FLAIR/WAIR) was assessed using the kappa (κ) statistic. In the cortical gray matter, both observers identified positive findings in 24.6% of FLAIR images, whereas MinPE FLAIR/WAIR showed a slightly higher rate (33.8%).
The chi-square test yielded a p-value of 0.247, indicating no statistically significant difference between the two methods of anesthesia. The inter-observer agreement was perfect, with κ = 1.00 for both FLAIR and MinPE FLAIR/WAIR. For the juxtacortical white matter, positive findings were observed in 43.1% of the FLAIR images and 46.2% of the MinPE FLAIR/WAIR images.
The second observer reported similar rates (43.1% and 46.2%). The chi-square test yielded a p-value of 0.724, again showing no significant difference. Kappa values of 0.971 and 0.973 indicated a strong inter-observer agreement.
In the subcortical white matter, positive findings were noted in 52.3% of the FLAIR images and 58.5% of the MinPE FLAIR/WAIR images. However, this difference was not statistically significant (p = 0.480). The inter-observer agreement remained high, with κ = 0.972 for the first observer and κ = 1.00 for the second, reflecting excellent reliability.
In the periventricular white matter, positive findings were observed in 87.7% of the FLAIR images, increasing slightly to 90.8% with MinPE FLAIR/WAIR. The chi-square test yielded a non-significant p-value of 0.571, indicating no meaningful difference between the two methods. The kappa (κ) values were 0.936 and 0.968, reflecting excellent inter-observer agreement between the two observers.
In the corpus callosum, the first observer reported positive findings in 12.3% of the FLAIR images, and the second observer reported positive findings in 13.8% of the FLAIR images. MinPE FLAIR/WAIR showed a slightly higher rate (15.4%). The differences between the techniques were not statistically significant (p = 0.612 and p = 0.804, respectively).
The κ value of 0.932 further highlights the high level of agreement between the observers (Table 2).
In the basal ganglia, positive findings were detected in 6.2% of FLAIR images by the first observer, increasing to 13.8% with the MinPE FLAIR/WAIR protocol. Chi-square testing yielded p-values of 0.144 and 0.258, indicating no significant difference between the two methods tested. Kappa (κ) values of 0.881 and 1.00 reflected excellent inter-observer agreement.
In the thalamus, positive findings were noted in 21.5% of FLAIR images and 32.3% of MinPE FLAIR/WAIR images, with a negligible difference (p = 0.166). The inter-observer agreement remained strong, with κ values of 0.916 and 0.939, confirming reliable lesion identification. In the internal capsule, the first observer identified positive findings in 15.4% of FLAIR images, which increased slightly to 18.5% with MinPE FLAIR/WAIR, with similar rates for the second observer.
The chi-square tests yielded p-values of 0.640 and 0.818, again showing no significant difference. Kappa values of 0.946 and 1.00 indicated high inter-observer agreement. Overall, the total number of findings across all regions was 296 for FLAIR and 287 for MinPE FLAIR/WAIR for the first observer and 310 for FLAIR and 396 for MinPE FLAIR/WAIR for the second observer.
These results suggest a consistent pattern of slightly higher detection with MinPE FLAIR/WAIR imaging. However, regional variations remained statistically insignificant, demonstrating a strong inter-observer reliability across the imaging methods (Fig. 5).
In the midbrain, the first observer recorded positive findings in 13.8% of FLAIR images, increasing to 41.5% with the MinPE FLAIR/WAIR sequence. The second observer reported the same rates (13.8% for FLAIR and 41.5% for MinPE FLAIR/WAIR). The chi-square test yielded a p-value of 0.0004, indicating a statistically significant difference, suggesting that MinPE FLAIR/WAIR detected more midbrain findings than FLAIR alone.
The inter-observer agreement was high, with kappa (κ) values of 0.814 for FLAIR and 1.00 for MinPE FLAIR/WAIR. In the pons, the first observer identified positive findings in 27.7% of FLAIR images, increasing to 72.3% with MinPE FLAIR/WAIR. The second observer noted a similar pattern, with positive findings increasing from 27.7% on FLAIR to 72.3% on MinPE FLAIR/WAIR.
The chi-square test produced a p-value of 0.0001, indicating a significant difference between the two methods tested. Kappa values of 0.932 and 0.944 reflected strong inter-observer agreement.
In the medulla, positive findings were observed in 4.6% of FLAIR images, increasing markedly to 58.5% with MinPE FLAIR/WAIR. The chi-square test showed a highly significant difference (p = 0.0001), supporting the greater effectiveness of MinPE FLAIR/WAIR in detecting abnormalities in this region. Inter-observer agreement was perfect for FLAIR (κ = 1.00) and nearly perfect for MinPE FLAIR/WAIR (κ = 0.978).
In the cerebellar peduncle, positive findings were noted in 21.5% of the FLAIR images, rising to 32.3% with MinPE FLAIR/WAIR. However, this difference was not statistically significant (p = 0.166). Kappa values of 0.874 and 0.939 indicated substantial inter-observer agreement (Table 3).
In the cerebellar hemisphere, the first observer noted positive findings in 16.9% of FLAIR images, increasing to 30.8% with MinPE FLAIR/WAIR sequences. The second observer reported similar results to the first. This difference was statistically significant (p = 0.043), suggesting that MinPE FLAIR/WAIR enhances lesion detection in this region of interest.
The inter-observer agreement was high, with kappa (κ) values of 0.896 and 0.937. Overall, across all regions of the posterior fossa, the first observer recorded 91 findings on FLAIR images and 374 on MinPE FLAIR/WAIR images. The second observer reported 107 findings on FLAIR and 386 on MinPE FLAIR/WAIR sequences.
These results consistently demonstrate higher lesion detection with MinPE FLAIR/WAIR, supporting its greater detection rates for identifying posterior fossa abnormalities, as further evidenced by the strong inter-observer agreement (Fig. 6).
DISCUSSION
Multiple sclerosis (MS) is a chronic autoimmune disease characterized by inflammation, demyelination, and axonal damage in the central nervous system (CNS) [1, 2]. The diagnosis and management of MS rely heavily on neuroimaging techniques, particularly magnetic resonance imaging (MRI), which provides essential information about lesion burden, disease progression, and treatment response [7]. However, conventional MRI sequences, such as Fluid-Attenuated Inversion Recovery (FLAIR), have limitations in detecting certain types of MS lesions, especially those located in the subcortical white matter, thalamus, and brainstem [16].
These limitations underscore the need for improved imaging techniques to enhance the sensitivity and specificity of MS-lesion detection. Advanced techniques, such as Double Inversion Recovery (DIR), Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE), susceptibility-weighted imaging (SWI), Phase-Sensitive Inversion-Recovery (PSIR), and 3D inversion recovery ultrashort echo time (IR-UTE) sequences, have significantly improved the visualization and detection of cortical lesions. Nevertheless, their use in routine diagnostic and clinical trial protocols remains limited because of the prolonged acquisition time [30-35].
In this study, we evaluated the effectiveness of fused fluid-attenuated Inversion Recovery (FLAIR) and white matter-attenuated inversion recovery (WAIR) sequences (“MinPE FLAIR/WAIR”) in improving the detection of MS lesions. This method is based on suppressing dominant background tissue signals by combining FLAIR, which nulls the cerebrospinal fluid, and WAIR, which reduces the normal white matter signal. This voxel-wise minimum selection increases lesion conspicuousness by retaining the signal from MS lesions while minimizing the surrounding tissue signal.
Our findings demonstrate promising results, indicating that “MinPE FLAIR/WAIR” images demonstrate enhanced lesion conspicuousness and higher lesion detection rates than conventional FLAIR sequences across various brain regions. The addition of a standard WAIR sequence, which is available in all MRI scanners, to the routine protocol adds only ~1.5 min to the total scan time.
One notable finding of our study was the increased detection of lesions in brain regions that are often missed or underestimated by conventional FLAIR imaging. The number and volume of lesions play significant roles in MS pathology and are associated with disease progression and cognitive impairment [2, 6, 7, 16]. Our findings suggest that MinPE FLAIR/WAIR may improve lesion detection in several brain regions compared with conventional 2D FLAIR.
The mean number of lesions in MinPE FLAIR/WAIR was generally more than that seen in standard FLAIR within the supratentorial areas (eg, cortical gray matter, juxtacortical white matter, and basal ganglia). However, most of these differences were not statistically significant. The trend toward higher lesion counts indicates a potential added value that warrants further investigation in larger cohorts.
Importantly, MinPE FLAIR/WAIR demonstrated a statistically significant improvement in detecting posterior fossa lesions, a region where subtle pathology can be challenging to visualize owing to anatomical constraints and CSF-related artifacts on standard imaging. This enhanced lesion conspicuousness in the infratentorial region highlights the clinical promise of MinPE sequences for a more comprehensive evaluation of the MS lesion burden [31, 36].
In the present study, the MinPE fusion technique substantially improved the conspicuousness of lesions across multiple brain regions, particularly in the posterior fossa. These findings are consistent with those of prior studies demonstrating that optimized image combination methods can enhance MS lesion detection beyond conventional FLAIR or T2-weighted sequences alone. Compared to other studies, our MinPE approach uniquely leverages information from both FLAIR (CSF suppression) and WAIR (white-matter signal attenuation), resulting in a synergistic enhancement of lesion-to-background contrast, especially in anatomically challenging regions, such as the midbrain, pons, and medulla.
The inclusion of 95% confidence intervals (CIs) for the sensitivity estimates provided additional insight into the precision of our findings. The CIs indicate the range within which the true sensitivity of each imaging technique is likely to fall, allowing for a more nuanced interpretation of the comparative performance of MinPE FLAIR/WAIR and conventional FLAIR imaging. This is particularly important for assessing the robustness of the results, given the sample size and variability in the readings.
The findings of our study have significant clinical implications for the diagnosis, management, and monitoring of MS [37], [18]. The improved lesion conspicuousness of MinPE FLAIR/WAIR images for detecting MS lesions may facilitate earlier identification of disease activity, more accurate assessment of disease burden, and improved monitoring of treatment response. Furthermore, the ability of MinPE FLAIR/WAIR to enhance lesion detection offers new opportunities to understand MS pathophysiology and develop targeted therapeutic interventions [26, 28, 31].
The results from our study were likewise consistent with previous studies that have evaluated how well different advanced imaging technologies, including double inversion recovery (DIR) or phase-sensitive inversion recovery (PSIR), can improve the identification of MS lesions [16, 38-40]. Many studies have also reported almost equal improvements in the ability to identify lesions, and that advanced MRI sequence imaging has added to or supported the ability to identify lesions and diagnose MS accurately. Again, there is no single best protocol, and the optimal imaging protocol for MS will depend on the stage of the disease, the type of lesion(s), and specific characteristics of the individual patient [16, 28, 38].
The addition of the WAIR sequence can be integrated into the standard MRI protocol with minimal disruption of the workflow. The primary requirement is the post-processing step required to generate the MinPE image contrast. This step is independent of the scanner and can be performed on an external workstation using routine image analysis software or can be fully automated within the institutional pipeline.
Therefore, although a small amount of post-acquisition processing is necessary, the method can be readily incorporated into clinical practice without modifying the existing hardware or pulse-sequence structure. The proposed fusion approach is particularly advantageous in clinical settings where access to advanced imaging sequences is limited. By combining complementary information from routinely acquired MRI sequences, the method enhances lesion conspicuity without relying on specialized acquisitions such as double inversion recovery (DIR).
This makes the approach especially suitable for older MRI scanners, where advanced imaging methods may not be available, and helps explain its improved performance compared with baseline techniques that depend on single-contrast information or specialized sequences.
LIMITATIONS AND CONSIDERATIONS
Despite these promising results, several limitations of our study should be considered when interpreting these findings. First, this study was conducted at a single center with a relatively small sample size, which may limit the generalizability of the results. Future multicenter studies with larger cohorts are needed to validate our findings and further assess the clinical utility of MinPE FLAIR/WAIR images in diverse patient populations.
Additionally, the independent acquisition of FLAIR and WAIR sequences may allow for patient motion between acquisitions, potentially leading to image displacement and fusion errors, which can be corrected using dedicated registration techniques. In the current study, no quantitative metrics (e.g., mutual information or Dice coefficient) were applied to validate coregistration accuracy; instead, coregistration quality was assessed using systematic visual inspection. Future studies should consider this limitation and compare our method with other established techniques, such as 2D and 3D DIR and PSIR, in the same patient cohort to enable a direct quantitative comparison of lesion conspicuity, sensitivity, and diagnostic value using quantitative metrics, such as SNR, CNR, or lesion-to-background ratio.
Inter-reader variability may also arise due to lesion size, volume, or fragmentation, which may be influenced by co-registration accuracy or reader confidence. The present study was designed as an exploratory analysis, and several methodological aspects provide opportunities for future investigations. The application of multiple-comparison–aware statistical approaches and the inclusion of lesion-level specificity and false-positive metrics would strengthen the quantitative interpretation.
The present study was designed as an exploratory analysis, and several methodological aspects provide opportunities for future investigations. The application of multiple-comparison–aware statistical approaches and the inclusion of lesion-level specificity and false-positive metrics would strengthen the quantitative interpretation. Future studies using consensus annotation or longitudinal imaging as reference standards may further clarify the clinical impact of improved lesion detection.
The post-hoc power analysis should be interpreted with caution, as it was derived from observed effect sizes in regions demonstrating statistically significant differences. Consequently, the reported power does not imply uniform sensitivity to detect smaller differences across all brain regions. This limitation underscores the importance of cautious interpretation of non-significant findings and highlights the need for larger, prospective studies to evaluate more subtle regional differences.
Finally, the retrospective study design and reliance on existing MRI datasets may have introduced selection bias and limited control over confounding variables. Prospective studies with standardized imaging protocols and longitudinal follow-ups are required to confirm the efficacy of MinPE FLAIR/WAIR in clinical practice.
CONCLUSION
In conclusion, our study demonstrates that fused FLAIR and WAIR images, known as MinPE FLAIR/WAIR, demonstrate improved lesion conspicuousness and higher MS lesion detection rates compared with conventional FLAIR sequences. Improved lesion detection across various brain regions has important implications for the diagnosis, monitoring, and treatment planning of MS. Future research should aim to validate these findings in larger cohorts and explore the potential role of MinPE FLAIR/WAIR in predicting clinical outcomes and guiding therapeutic decisions in patients with MS.