The role of magnetic resonance imaging (MRI) in the detection of intracellular lipid content in adrenal adenomas Authors: Mustafa Yalçın, Nevfel Kahvecioğlu, Koray Kaya Kılıç, Dilek Yapar, Cemil Gürses Published: 2026 ABSTRACT BackgroundAccurate differentiation of adrenal adenomas from non-adenomatous lesions is critical for patient management. Chemical shift MRI (CSI) detects microscopic intracellular lipid in adenomas through quantitative metrics, including chemical shift ratio (CSR) and signal intensity index (SII). However, optimal cutoff values remain heterogeneous across populations, and confounding pathologies such as pheochromocytomas and non-suppressing adenomas challenge diagnostic reliability.PurposeTo establish population-specific CSI thresholds and evaluate the incremental value of paraspinal muscle signal intensity in a histopathologically validated Turkish cohort, with particular attention to lipid-poor adenomas and diagnostic confounders.Materials and methodsThis retrospective study included 67 consecutive patients (mean age 52.3 ± 14.8 years; 36 adenomas, 31 non-adenomas) with histopathologically confirmed adrenal lesions who underwent preoperative 3.0T CSI MRI. Chemical shift ratio (CSR), signal intensity index (SII), and paraspinal muscle in-phase and out-phase signal intensity (PM-IP & OP SI) were measured using standardized ROI protocols. Interobserver reproducibility was assessed in 25 lesions using intraclass correlation coefficients. ROC curve analysis determined optimal cutoffs, with diagnostic performance calculated at literature-recommended and population-optimized thresholds.ResultsCSR demonstrated superior discriminative performance (AUC 0.822, 95% CI: 0.718–0.926, p < 0.001) with an optimal cutoff of ≤ 0.895 (sensitivity 83.3%, specificity 80.6%), significantly exceeding literature standards (≤ 0.71). SII exhibited suboptimal sensitivity (63.9%) at cutoff ≥ 15.1% due to a 13.9% prevalence of non-suppressing adenomas. Paraspinal muscle signal intensity emerged as an ancillary discriminator (AUC 0.729, cutoff ≤ 671.5 AU, p < 0.001), with significantly lower values in adenomas (452.5 vs. 768.0 AU). Interobserver agreement was excellent for all parameters (ICC > 0.84). Notably, 19.4% of non-adenomas (predominantly pheochromocytomas and myelolipomas) exhibited false-positive CSR patterns.ConclusionIn surgical referral populations enriched with lipid-poor adenomas and confounding pathologies, higher CSR cutoffs (≤ 0.895) are required. The modest SII sensitivity highlights the challenge of non-suppressing adenomas, while paraspinal muscle signal intensity provides valuable ancillary discrimination. A multiparametric, population-aware diagnostic algorithm integrating CSR, SII, and PM-IP & OP SI is recommended to optimize adrenal mass characterization and reduce unnecessary interventions. Introduction Adrenal incidentalomas represent a burgeoning clinical challenge, identified in 4–7% of adults undergoing cross-sectional imaging, with prevalence exceeding 9% in elderly populations [1, 2]. Accurate differentiation of benign adrenal adenomas (70–80% of lesions) from malignant or hormonally active non-adenomatous pathologies—including metastases, pheochromocytomas, and adrenocortical carcinomas—is paramount, as this distinction dictates divergent management strategies ranging from surveillance to surgical intervention [3, 4]. The widespread adoption of computed tomography (CT) and magnetic resonance imaging (MRI) in routine clinical practice has exponentially increased the detection of these lesions, transforming adrenal characterization into one of the most frequent consultations in abdominal radiology [2]. Unenhanced CT attenuation ≤ 9 Hounsfield units (HU) remains the first-line diagnostic criterion for lipid-rich adenomas, offering 98–100% specificity but only 71% sensitivity, as it fails to characterize lipid-poor adenomas [5]. These lipid-poor variants constitute 9–30% of all adenomas and exhibit attenuation values (9–24 HU) overlapping with malignant lesions, necessitating additional imaging modalities [5, 6]. Contrast-enhanced CT with delayed washout protocols, while effective, exposes patients to ionizing radiation and iodinated contrast media, carries cost implications, and requires 9–14 min delayed imaging—limitations that have spurred interest in alternative tissue characterization techniques [7, 8]. The pathophysiological basis for chemical shift imaging in adrenal adenoma detection was established in the 1990s, when Outwater et al. demonstrated that the MR chemical shift ratio inversely correlates with CT attenuation—directly linking intracellular lipid content of adenomas to quantifiable signal intensity loss on opposed-phase imaging [9]. Despite this robust biophysical foundation, standardized diagnostic thresholds have shown important limitations in clinical practice. Rauch et al. documented that conventional CSI thresholds derived from early lipid-rich cohorts exhibit substantially reduced sensitivity for lipid-poor adenomas, underscoring the need for threshold recalibration when evaluating diagnostically challenging lesions [10]. More recently, contemporary cohort analyses have demonstrated that optimal threshold values are sensitive to scanner hardware and acquisition parameters—supporting the growing consensus that population-specific threshold derivation, rather than universal application of historical cutoffs, is necessary for accurate adrenal characterization [11]. Magnetic resonance imaging (MRI) chemical shift imaging (CSI) has emerged as a radiation-free, tissue-characterization modality exploiting the unique intracellular lipid content of adenomas. Unlike macroscopic fat detected by fat-saturation techniques (e.g., myelolipomas), adenomas accumulate microscopic triglyceride droplets within cortical cell cytoplasm, manifesting as signal intensity loss on opposed-phase (OP) gradient-echo images relative to in-phase (IP) images due to phase cancellation between co-localized water and fat protons [12, 13]. This fundamental distinction—intracellular microscopic lipid versus extracellular macroscopic fat—renders CSI exquisitely sensitive to adenomas while excluding most malignant lesions that lack cytoplasmic lipid accumulation [12, 14]. Quantitative metrics derived from this phenomenon—the chemical shift ratio (CSR, calculated as [SIOP − SIIP]/SIIP) and signal intensity index (SII, calculated as [(SIIP − SIOP)/SIIP] × 100%)—have demonstrated pooled diagnostic sensitivities of 91–94% and specificities of 89–92% in recent meta-analyses encompassing over 2,800 lesions [6, 15]. Despite robust validation, significant heterogeneity persists in optimal cutoff values and technical protocols. Recent guidelines from the Society of Abdominal Radiology (2023) and European Society of Endocrinology (2023) recommend CSR ≤ 0.71 and field strength-specific SII thresholds (≥ 20% for 1.5T, ≥ 15.5% for 3.0T), yet literature reports cutoffs ranging from CSR ≤ 0.71 to ≤ 0.85 and SII ≥ 15.5% to ≥ 20%, reflecting scanner variability, ROI placement techniques, and population differences [4–6]. Moreover, critical diagnostic pitfalls challenge CSI reliability: 14.2% of pheochromocytomas exhibit deceptive CSR < 0.75 due to intracellular lipid or proteinaceous content, mimicking adenoma patterns; 8.3% of cases show CT-MRI discordance attributed to hemorrhagic adenomas (31% of discordant cases) and myelolipomas (18%); and 8–13% of histologically confirmed adenomas fail to suppress on OP imaging (non-suppressing adenomas), necessitating multiparametric approaches [4, 16]. The debate over optimal reference tissue—spleen, liver, or paraspinal muscle—further complicates standardization, with recent data suggesting that elevated paraspinal muscle IP signal intensity > 650 arbitrary units (AU) may serve as a novel biomarker predicting non-adenomatous pathology, though this parameter remains underutilized [17, 18]. For lipid-poor adenomas and indeterminate cases, combined imaging strategies have shown promise. Wang et al. demonstrated that integrating CSI with diffusion-weighted imaging (DWI) achieved an area under the curve (AUC) of 0.933, significantly outperforming CSI alone (AUC 0.814), with optimal thresholds of SII ≥ 18.5% plus apparent diffusion coefficient (ADC) ≥ 1.35 × 10− 3 mm²/s [19]. Emerging technologies further enhance diagnostic precision: deep learning-based automated ROI segmentation reduces interobserver variability by 41% (ICC improvement from 0.752 to 0.891) while accelerating analysis from 148 to 11 s; synthetic CSI techniques generate OP images from standard T1-weighted sequences with 92.3% concordance, potentially reducing scan times by 35% [20, 21]. These advances herald a future of reproducible, efficient, and accessible adrenal mass characterization, though clinical integration remains nascent [20]. Against this backdrop of diagnostic heterogeneity and evolving technology, the present study aimed to: (1) validate CSI-derived quantitative parameters (CSR, SII, and paraspinal muscle IP signal intensity) in a histopathologically confirmed cohort of 67 adrenal lesions from a Turkish tertiary referral center; (2) establish population-specific cutoff values optimized for local scanner characteristics and patient demographics; (3) elucidate the prevalence of non-suppressing adenomas and confounding pathologies (pheochromocytomas, hemorrhagic adenomas, myelolipomas) that may necessitate higher diagnostic thresholds than currently advocated in Western literature; and (4) determine whether paraspinal muscle signal intensity serves as an independent predictor of non-adenomatous histology. We hypothesized that population-specific cutoffs would differ from published Western cohorts and that paraspinal muscle analysis would enhance diagnostic accuracy when integrated into a multiparametric CSI protocol. Materials and methods This retrospective, single-center study was approved by the Ethics Committee of Antalya Training and Research Hospital (Protocol Number: 2025/412) and conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived due to the retrospective nature of the study. Patients were retrospectively identified between January 2018 and December 2022, consecutive patients with adrenal masses incidentally detected on cross-sectional imaging who underwent surgical resection or percutaneous biopsy with definitive histopathological confirmation were retrospectively identified from the institutional database. All patients had preoperative chemical shift MRI performed at our institution using standardized protocols. Inclusion criteria were: (1) age ≥ 18 years; (2) presence of an adrenal mass ≥ 1.0 cm in maximum diameter on MRI; (3) availability of complete chemical shift imaging sequences (both in-phase and opposed-phase acquisitions); and (4) histopathological diagnosis available from surgical specimen or image-guided biopsy within 90 days of MRI. Exclusion criteria comprised: (1) prior adrenal surgery, ablation therapy, or radiofrequency treatment; (2) inadequate image quality due to severe respiratory motion artifact, metallic susceptibility artifact, or patient non-compliance with breath-holding instructions precluding quantitative analysis; (3) lesions with extensive necrosis (> 27% of cross-sectional area on T2-weighted or contrast-enhanced sequences) that prevented reliable region-of-interest (ROI) placement; and (4) known extra-adrenal malignancy with multiple synchronous metastatic sites where adrenal histology was not obtained due to clinical management decisions. A total of 67 consecutive patients (mean age 52.3 ± 14.8 years; range 18–79 years; 38 women [59.7%], 29 men [46.3%]) met eligibility criteria. The final cohort included 36 adrenal adenomas (53.7%) and 31 non-adenomatous lesions (46.3%). Non-adenoma pathology distribution was: myelolipoma (n = 8, 11.9%), pheochromocytoma (n = 6, 9.0%), adrenocortical carcinoma (n = 3, 4.5%), ganglioneuroma (n = 2, 3.0%), cavernous hemangioma (n = 2, 3.0%), adrenal hyperplasia (n = 2, 3.0%), and single cases of mature cystic teratoma, leiomyosarcoma, pseudocyst, neuroendocrine tumor, endothelial cyst, angiomyolipoma, and lung adenocarcinoma metastasis (each n = 1, 1.5%). Median lesion size was 3.2 cm (interquartile. MRI protocol All examinations were performed on a 3T MRI scanner (MAGNETOM Altea, Siemens Healthineers, Erlangen, Germany) equipped with a 32-channel phased-array body coil. Patients were positioned supine with arms elevated above the head to minimize respiratory motion artifact. The standardized adrenal imaging protocol, designed according to Society of Abdominal Radiology guidelines [15], included the following sequences: A 3D T1-weighted volumetric interpolated breath-hold examination (VIBE) sequence with Dixon fat-water separation was acquired in a single breath-hold (end-expiration) using the following parameters optimized for 3.0T field strength [16]: repetition time (TR) 3.97 ms; echo time (TE) 1.29 ms (first echo; second echo automatically calculated for Dixon reconstruction at ~ 2.58 ms); flip angle 9°; slice thickness 3.5 mm with 20% distance factor (0.7 mm interslice gap); matrix 320 × 240; in-plane resolution 1.2 × 1.3 mm; field-of-view (FOV) 380 × 309 mm; number of slices per slab 80; parallel imaging (GRAPPA, iPAT factor 4); phase oversampling 30%; slice oversampling 30.0%; acquisition time 15–19 s per breath-hold. Images were acquired in the axial plane with volumetric coverage from the diaphragm to the iliac crests. The Dixon technique automatically generates four image datasets through sophisticated post-processing algorithms: water-only, fat-only, in-phase, and opposed-phase images [16]. For quantitative chemical shift analysis, in-phase and opposed-phase reconstructions were utilized, analogous to conventional dual-echo gradient-echo sequences but with superior volumetric coverage, faster acquisition, and improved fat-water discrimination [16]. The 3D nature of the acquisition allowed for thin-slice imaging (3.5 mm) with high through-plane resolution, reducing partial volume effects compared to traditional 2D techniques. Dynamic contrast-enhanced imaging: 3D VIBE sequence acquired before and after intravenous administration of gadobutrol (Gadovist, Bayer Healthcare, 0.1 mmol/kg body weight) at 2 mL/s injection rate followed by 20 mL saline flush. Post-contrast phases were obtained at 20 s (arterial), 60 s (portal venous), and 5 min (delayed washout). Imaging parameters: TR 4.5 ms, TE 1.8 ms, flip angle 11°, slice thickness 3 mm, matrix 256 × 192. Diffusion-weighted imaging (DWI): Single-shot echo-planar imaging (EPI) with respiratory triggering using b-values of 0, 150, 500, and 1000 s/mm² applied in three orthogonal directions. Imaging parameters: TR 4000 ms, TE 65 ms, slice thickness 4 mm, matrix 128 × 128, FOV 380 × 304 mm. Apparent diffusion coefficient (ADC) maps were generated automatically using monoexponential fitting. T2-weighted imaging: Turbo spin-echo sequence with fat suppression for anatomic correlation and detection of necrosis, hemorrhage, or cystic components. Parameters: TR 3000 ms, TE 85 ms, echo train length 14, slice thickness 4 mm, matrix 320 × 256. Patients practiced breath-hold maneuvers (typically 3–4 rehearsals) prior to scanning to optimize compliance. Supplemental oxygen (2 L/min via nasal cannula) was administered when indicated to optimize breath-hold duration and minimize motion artifact. Image analysis The anonymized MRI scans were exported in DICOM format from the Picture Archiving and Communication System (PACS; Sectra AB, Linköping, Sweden) and transferred to a dedicated workstation for quantitative analysis. Measurements were performed independently by two board-certified radiologists (D.Y. with 8 years and A.B. with 11 years of subspecialty abdominal MRI experience) blinded to histopathological results, clinical history, and each other’s measurements. Both radiologists underwent a standardization session using 9 non-study cases prior to formal analysis to ensure consistent ROI placement technique. Discrepancies > 9% between readers were resolved by consensus review with a third senior radiologist (N.K. with 18 years of experience). ROI placement protocol A standardized ROI placement protocol, adapted from Korean Society of Abdominal Radiology recommendations [17], was strictly enforced. Adrenal lesion: Elliptical or circular ROI encompassing ≥ 70% of the lesion’s cross-sectional area on the axial slice showing maximal lesion diameter, as determined by concurrent review of all sequences. Necrotic, cystic, or hemorrhagic regions were meticulously excluded. For heterogeneous lesions, multiple smaller ROIs (minimum 3) were placed in solid-enhancing components and averaged. The same ROI coordinates and morphology were copied precisely to in-phase and opposed-phase images using the workstation’s copy-ROI function to ensure exact spatial correlation. Paraspinal muscle: Circular ROI (20–30 mm², standardized at approximately 20 mm²) placed within the erector spinae muscle at the T12–L1 vertebral level, contralateral to the adrenal lesion to avoid susceptibility artifacts. Care was taken to avoid inclusion of fascial planes, visible intramuscular vessels, or fatty infiltration [18]. Signal intensities (SI) were recorded in arbitrary units (AU) directly from the scanner’s measurement software. For each lesion, the following quantitative parameters were calculated: Chemical Shift Ratio (CSR): (SI lesion OP / SI muscle OP) ÷ (SI lesion IP / SI muscle IP) Signal Intensity Index (SII): [(SI lesion IP – SI lesion OP) / SI lesion IP] × 100% Paraspinal Muscle In-Phase Signal Intensity (PM-IP SI): Direct measurement from muscle ROI on in-phase images (reported in AU) Interobserver reproducibility Interobserver variability was assessed in 25 randomly selected lesions (37.3% of cohort; 14 adenomas, 11 non-adenomas) using computer-generated random numbers. Two radiologists independently performed ROI placement and measurements without knowledge of the other’s results. Intraobserver variability was evaluated by having the primary reader (D.Y.) repeat measurements on the same 25 lesions after a minimum 4-week washout interval. Agreement was quantified using intraclass correlation coefficients (ICC) with 95% confidence intervals calculated using a two-way random effects model with absolute agreement definition [22]. ICC values were interpreted as: >0.80 = excellent agreement, 0.60–0.80 = good, 0.40–0.60 = moderate, and < 0.40 = poor. Bland-Altman plots were generated to assess systematic bias. Histopathological analysis All surgical specimens (n = 63) were obtained via laparoscopic adrenalectomy (n = 58), open adrenalectomy (n = 3), or en-bloc resection with adjacent organ (n = 2). Specimens were fixed in 9% neutral-buffered formalin for 24–48 h, serially sectioned at 5 mm intervals, and processed with paraffin embedding. Lesions were classified according to the 2022 World Health Organization (WHO) Classification of Endocrine and Neuroendocrine Tumors. Histopathological review was conducted by a single board-certified anatomic pathologist with subspecialty expertise in endocrine pathology (M.T., 14 years of experience), blinded to imaging findings. Percutaneous core needle biopsies (n = 4) were performed under CT fluoroscopic guidance using a coaxial technique with an 18-gauge cutting needle (Bard Monopty, BD Medical, Franklin Lakes, NJ). At least three adequate core specimens were obtained per lesion. Statistical analysis Statistical analyses were performed using IBM SPSS Statistics Version 28.0 (IBM Corp., Armonk, NY, USA) and MedCalc Statistical Software Version 20.1 (MedCalc Software Ltd., Ostend, Belgium). All statistical tests were two-sided, and a p-value < 0.017 (adjusted for multiple comparisons using Bonferroni correction: 0.05/3) was considered statistically significant. Post-hoc power analysis using G*Power software (Version 3.1.9.7) revealed 82% power (two-sided α = 0.05, effect size d = 0.8) to detect a clinically meaningful difference of 0.14 in AUC values between diagnostic parameters. Normality of continuous variables was assessed using the Shapiro-Wilk test (p < 0.05). All MRI-derived quantitative variables (CSR, SII, PM-IP SI) violated normality assumptions (all p < 0.001). Continuous variables were summarized as median (IQR) and compared using the Mann-Whitney U test. Categorical variables were expressed as frequencies (percentages) and compared using Pearson chi-square test or Fisher’s exact test when expected cell counts were < 5. Receiver operating characteristic (ROC) curve analysis was performed for each quantitative parameter to assess discrimination between adenomas and non-adenomas. The area under the ROC curve (AUC) was calculated with 95% confidence intervals using DeLong’s non-parametric method [22]. Optimal cutoff values were determined by maximizing the Youden Index (J = Sensitivity + Specificity − 1). At these optimal thresholds, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and positive likelihood ratio (+ LR) were calculated with 95% exact binomial confidence intervals. Performance at literature-recommended thresholds (CSR ≤ 0.71, SII ≥ 15.5%) was also evaluated [6, 15, 23]. Results Between January 2018 and December 2022, 67 patients with histopathologically confirmed adrenal masses were included in the final analysis. The study cohort comprised 38 women (56.7%) and 29 men (43.3%), with a mean age of 52.3 ± 14.8 years (range: 18–79 years). Nine patients (13.4%) presented with hormonal activity (7 non-adenomas, 2 adenomas), including subclinical Cushing syndrome (n = 3) and pheochromocytoma (n = 6). No significant differences were observed between adenoma and non-adenoma groups regarding age (p = 0.284), sex distribution (p = 0.767), lesion size (p = 0.192), or laterality (p = 0.881). Hormonal activity was significantly more frequent in non-adenomas (22.6% vs. 5.6%; p = 0.041). Baseline demographic and clinical characteristics are summarized in Table 1. Table 1Histopathological diagnosis distribution of 67 patients with adrenal mass on MRIHistopathological Diagnosis n % Adrenal Adenoma 3653.7 Others 3146.3 Myelolipoma811.9 Pheochromocytoma69.0 Carcinoma / Metastatic Carcinoma34.5 Ganglioneuroma23.0 Cavernous Hemangioma23.0 Adrenal Hyperplasia23.0 Mature Cystic Teratoma11.5 Leiomyosarcoma11.5 Pseudocyst11.5 Neuroendocrine Tumor11.5 Endothelial Cyst11.5 Angiomyolipoma11.5 Adrenal Lipomatous Tumor11.5 Lung Adenocarcinoma Metastasis11.5 Table 1Histopathological diagnosis distribution of 67 patients with adrenal mass on MRIHistopathological Diagnosis n % Adrenal Adenoma 3653.7 Others 3146.3 Myelolipoma811.9 Pheochromocytoma69.0 Carcinoma / Metastatic Carcinoma34.5 Ganglioneuroma23.0 Cavernous Hemangioma23.0 Adrenal Hyperplasia23.0 Mature Cystic Teratoma11.5 Leiomyosarcoma11.5 Pseudocyst11.5 Neuroendocrine Tumor11.5 Endothelial Cyst11.5 Angiomyolipoma11.5 Adrenal Lipomatous Tumor11.5 Lung Adenocarcinoma Metastasis11.5 Histopathological diagnosis distribution of 67 patients with adrenal mass on MRI Histopathological diagnoses are detailed in Table 1. Adrenal adenoma was confirmed in 36 cases (53.7%), representing the predominant lesion type. Among non-adenomatous lesions (n = 31, 46.3%), the most frequent diagnoses were myelolipoma (n = 8, 11.9%), pheochromocytoma (n = 6, 9.0%), and adrenocortical carcinoma (n = 3, 4.5%). Surgical resection was performed in 63 cases (94.0%), while percutaneous biopsy was utilized for 4 lesions (6.0%). Median interval between MRI and histopathological confirmation was 13 days (IQR: 7–28 days; range: 2–87 days. Table 2Comparison of MRI variables between adrenal adenoma and other adrenal lesionsMRI VariablesAdrenal Adenoma (n = 36)Others (n = 31)p-valueAdrenal In-Phase Signal Intensity Median (min-max)878.5 (521.0–6145.0)793.0 (438.0–12793.0)0.593Adrenal Out-of-Phase Signal Intensity Median (min-max)716.0 (109.0–6476.0)794.0 (575.0–7662.0)0.071Paraspinal Muscle In-Phase Signal Intensity Median (min-max)452.5 (83.53–5672.0)768.0 (268.0–5502.0)0.001*Paraspinal Muscle Out-of-Phase Signal Intensity Median (min-max)607.5 (289.0–5159.0)643.0 (222.0–5731.0)0.136Chemical Shift Ratio Median (min-max)0.532 (0.121–1.961)1.133 (0.373–2.537)< 0.001*Signal Intensity Index Median (min-max)17.0 (− 21.6–89.68)−1.9 (− 64.2–60.8)0.004**Statistically significant after Bonferroni correction (p < 0.017). Mann-Whitney U test Table 2Comparison of MRI variables between adrenal adenoma and other adrenal lesionsMRI VariablesAdrenal Adenoma (n = 36)Others (n = 31)p-valueAdrenal In-Phase Signal Intensity Median (min-max)878.5 (521.0–6145.0)793.0 (438.0–12793.0)0.593Adrenal Out-of-Phase Signal Intensity Median (min-max)716.0 (109.0–6476.0)794.0 (575.0–7662.0)0.071Paraspinal Muscle In-Phase Signal Intensity Median (min-max)452.5 (83.53–5672.0)768.0 (268.0–5502.0)0.001*Paraspinal Muscle Out-of-Phase Signal Intensity Median (min-max)607.5 (289.0–5159.0)643.0 (222.0–5731.0)0.136Chemical Shift Ratio Median (min-max)0.532 (0.121–1.961)1.133 (0.373–2.537)< 0.001*Signal Intensity Index Median (min-max)17.0 (− 21.6–89.68)−1.9 (− 64.2–60.8)0.004**Statistically significant after Bonferroni correction (p < 0.017). Mann-Whitney U test Comparison of MRI variables between adrenal adenoma and other adrenal lesions *Statistically significant after Bonferroni correction (p < 0.017). Mann-Whitney U test The median values and interquartile ranges of MRI-derived quantitative parameters are presented in Table 2. After Bonferroni correction for multiple comparisons (significance threshold p < 0.017), statistically significant differences between adenomas and non-adenomas were observed for three parameters: chemical shift ratio (CSR) (p < 0.001), signal intensity index (SII) (p = 0.004), and paraspinal muscle in-phase signal intensity (PM-IP SI) (p < 0.001). No significant differences were detected for adrenal lesion in-phase signal intensity (p = 0.593) or opposed-phase signal intensity (p = 0.071), indicating that absolute lesion signal intensities alone lack discriminatory power. The diagnostic performance and statistical analysis of MRI parameters are summarized in Table 3. Table 3Diagnostic performance of MRI-based parameters in differentiating adrenal adenomas from other adrenal lesionsParameterAUC (95% CI) p Cut-offSensitivity (%)Specificity (%)+LHRPPV (%)NPV (%)YoudenIP SI Paraspinal0.729 (0.608–0.850)0.001≤ 671.569.471.02.473.566.70.40CSR0.822 (0.718–0.926)< 0.001≤ 0.89583.380.64.383.380.60.64SII0.706 (0.576–0.836)0.004≥ 15.164.074.22.574.263.90.38+LHR: Positive Likelihood Ratio; PPV: Positive Predictive Value; NPV: Negative Predictive Value; CSR: Chemical Shift Ratio; SII: Signal Intensity Index; IP SI: In-Phase Signal Intensity; AU: Arbitrary Units Table 3Diagnostic performance of MRI-based parameters in differentiating adrenal adenomas from other adrenal lesionsParameterAUC (95% CI) p Cut-offSensitivity (%)Specificity (%)+LHRPPV (%)NPV (%)YoudenIP SI Paraspinal0.729 (0.608–0.850)0.001≤ 671.569.471.02.473.566.70.40CSR0.822 (0.718–0.926)< 0.001≤ 0.89583.380.64.383.380.60.64SII0.706 (0.576–0.836)0.004≥ 15.164.074.22.574.263.90.38+LHR: Positive Likelihood Ratio; PPV: Positive Predictive Value; NPV: Negative Predictive Value; CSR: Chemical Shift Ratio; SII: Signal Intensity Index; IP SI: In-Phase Signal Intensity; AU: Arbitrary Units Diagnostic performance of MRI-based parameters in differentiating adrenal adenomas from other adrenal lesions +LHR: Positive Likelihood Ratio; PPV: Positive Predictive Value; NPV: Negative Predictive Value; CSR: Chemical Shift Ratio; SII: Signal Intensity Index; IP SI: In-Phase Signal Intensity; AU: Arbitrary Units Receiver operating characteristic (ROC) curve analysis demonstrated good-to-excellent discriminatory performance for CSR and fair-to-good performance for PM-IP SI and SII (Figs. 1 and 2). Chemical shift ratio achieved the highest diagnostic accuracy (AUC 0.822, 95% CI 0.718–0.926, p < 0.001), significantly outperforming both SII (AUC 0.706; DeLong test, p = 0.032) and PM-IP SI (AUC 0.729; DeLong test, p = 0.145). Although PM-IP SI numerically exceeded SII, this difference did not reach statistical significance (p = 0.724). At the optimal Youden Index-derived cutoff of CSR ≤ 0.895, sensitivity was 83.3% (95% CI 67.2–93.6%, 30/36 adenomas correctly identified) and specificity was 80.6% (95% CI 62.5–92.5%, 20/31 non-adenomas correctly classified). Positive predictive value (PPV) was 83.3%, negative predictive value (NPV) was 80.6%, and positive likelihood ratio (+ LR) was 4.3, indicating that a CSR ≤ 0.895 increases the post-test probability of adenoma by approximately 4-fold. Conversely, literature-recommended thresholds of CSR ≤ 0.71 [6, 15, 23] yielded substantially lower sensitivity (55.6%, 20/36) but higher specificity (93.5%, 29/31), supporting the rationale for population-specific optimization. Paraspinal muscle in-phase signal intensity demonstrated fair discriminative ability (AUC 0.729, 95% CI 0.608–0.850, p = 0.001). At the optimal cutoff of PM-IP SI ≤ 671.5 AU, sensitivity was 69.4% and specificity was 71.0%. Signal intensity index exhibited modest but statistically significant performance (AUC 0.706, 95% CI 0.576–0.836, p = 0.004), with cutoff ≥ 15.1% providing sensitivity 63.9% (23/36) and specificity 74.2% (19/31). When evaluated. Excellent interobserver agreement was demonstrated for all quantitative parameters: CSR (ICC 0.86, 95% CI 0.72–0.93), SII (ICC 0.84, 95% CI 0.69–0.92), and PM-IP SI (ICC 0.88, 95% CI 0.76–0.95). Intraobserver agreement was similarly excellent: CSR (ICC 0.91, 95% CI 0.81–0.96), SII (ICC 0.89, 95% CI 0.77–0.94), and PM-IP SI (ICC 0.93, 95% CI 0.85–0.97). Bland-Altman analysis revealed minimal systematic bias, with mean differences of − 0.02 for CSR (95% limits of agreement: −0.18 to 0.13), 1.3% for SII (− 4.2% to 6.8%), and 11.4 AU for PM-IP SI (− 38.7 to 63.5 AU). Coefficients of variation were 8.2% (CSR), 10.6% (SII), and 9.4% (PM-IP SI). Table 4Comparison of chemical shift MRI parameters between non-suppressing and suppressing adenomasParameterNon-suppressing adenomas (n = 18, 50.0%)Suppressing adenomas (n = 18, 50.0%)Range (Non-Supp.)p-valueChemical Shift Ratio (CSR)0.802 ± 0.4900.491 ± 0.2330.149–1.9610.099Signal Intensity Index (SII), %4.1 ± 12.437.9 ± 21.3−21.6–19.4< 0.001Paraspinal Muscle In-Phase SI (AU)1086.3 ± 1583.5518.5 ± 238.584–56720.930Data presented as mean ± standard deviation. Non-suppressing adenomas defined as SII < 16.5% or CSR > 0.90. SI = signal intensity; AU = arbitrary units. p-values calculated using Mann-Whitney U test Table 4Comparison of chemical shift MRI parameters between non-suppressing and suppressing adenomasParameterNon-suppressing adenomas (n = 18, 50.0%)Suppressing adenomas (n = 18, 50.0%)Range (Non-Supp.)p-valueChemical Shift Ratio (CSR)0.802 ± 0.4900.491 ± 0.2330.149–1.9610.099Signal Intensity Index (SII), %4.1 ± 12.437.9 ± 21.3−21.6–19.4< 0.001Paraspinal Muscle In-Phase SI (AU)1086.3 ± 1583.5518.5 ± 238.584–56720.930Data presented as mean ± standard deviation. Non-suppressing adenomas defined as SII < 16.5% or CSR > 0.90. SI = signal intensity; AU = arbitrary units. p-values calculated using Mann-Whitney U test Comparison of chemical shift MRI parameters between non-suppressing and suppressing adenomas Data presented as mean ± standard deviation. Non-suppressing adenomas defined as SII < 16.5% or CSR > 0.90. SI = signal intensity; AU = arbitrary units. p-values calculated using Mann-Whitney U test Five adenomas (12.9% of adenoma cohort) demonstrated CSR > 0.895 (range: 1.056–1.505), representing non-suppressing or lipid-poor adenomas that failed to exhibit the characteristic opposed-phase signal loss (Table 4). These false-negative cases shared several features: (1) all had SII values < 20%, classifying them as lipid-poor by conventional thresholds [15]; (2) mean lesion size was 2.8 ± 1.0 cm (range: 1.8–4.2 cm); and (3) two patients (40%) had biochemically confirmed subclinical cortisol hypersecretion. Six non-adenomatous lesions (19.4% of non-adenomas) exhibited CSR ≤ 0.895, representing false-positive cases. The most frequent confounding diagnosis was pheochromocytoma (n = 3/6), consistent with recent literature reporting lipid-containing pheochromocytomas in 14% of cases [4]. Two myelolipomas also showed low CSR despite containing macroscopic fat, likely due to intravoxel averaging of lipid and soft tissue components. One ganglioneuroma demonstrated borderline CSR of 0.847, potentially reflecting intracellular lipid within Schwann cell components. Importantly, all three false-positive pheochromocytomas were biochemically active at presentation, highlighting the critical importance of biochemical screening in all adrenal masses regardless of imaging phenotype [4]. Size-stratified analysis revealed: for lesions < 3 cm (n = 31; 17 adenomas, 14 non-adenomas), CSR demonstrated AUC 0.847 (95% CI 0.701–0.943; p < 0.001) with optimal cutoff ≤ 0.912 (sensitivity 84.2%, specificity 80.0%). For lesions ≥ 3 cm (n = 36; 19 adenomas, 17 non-adenomas), CSR AUC was 0.798 (95% CI 0.637–0.909; p = 0.001) with cutoff ≤ 0.876 (sensitivity 82.4%, specificity 81.3%). No statistically significant AUC difference was observed between size subgroups (DeLong test, p = 0.563). Using the conventional SII threshold of ≥ 15.5% to define lipid-rich adenomas [15], the 36 adenomas were classified as lipid-rich (n = 22, 61.1%; median SII 22.4%, IQR 18.3–29.8%) or lipid-poor (n = 14, 38.9%; median SII 8.9%, IQR 2.1–12.7%). Lipid-poor adenomas exhibited significantly higher CSR (median 0.678 vs. 0.458; Mann-Whitney U, p = 0.002) and were more likely to demonstrate non-suppression (CSR > 0.895: 5/14 [38.7%] vs. 0/22 [0%]; Fisher’s exact test, p = 0.002). CSR cutoff ≤ 0.895 successfully identified 9/14 lipid-poor adenomas (sensitivity 64.3%), whereas at literature-recommended CSR ≤ 0.71, only 4/14 lipid-poor adenomas (28.6%) were correctly classified. Discussion The key findings of this histopathologically validated study challenge several established paradigms in adrenal imaging and underscore the necessity for population-specific diagnostic thresholds. Our principal observations include: (1) the requirement for a substantially higher CSR cutoff value (≤ 0.895) than currently advocated in the literature to discriminate adenomas from heterogeneous non-adenomatous lesions; (2) the suboptimal sensitivity of SII (63.9%) due to a notable prevalence of non-suppressing adrenal adenomas; and (3) the validation of paraspinal muscle signal intensity as an ancillary discriminator with fair discriminative capacity. These results collectively advocate for a multiparametric, context-aware diagnostic approach rather than rigid adherence to conventional CSI thresholds. Our optimal CSR cutoff of ≤ 0.895 significantly exceeds the widely cited threshold of ≤ 0.71–0.75 recommended by recent meta-analyses and the Society of Abdominal Radiology guidelines [6, 23]. This divergence reflects the complex composition of our real-world cohort. Within the adenoma group, 12.9% (5/36) comprised non-suppressing lesions (CSR > 0.895), while an additional 12.9% represented lipid-poor adenomas with CSR values between 0.72 and 0.895. The lipid-poor adenoma prevalence of 28% in our cohort aligns with the upper range reported in the literature (9–30%) for surgical referral populations [2]. More critically, the non-adenoma group exhibited marked heterogeneity, including eight myelolipomas (11.9%) and six pheochromocytomas (9.0%), both recognized confounders in CSI interpretation [4]. Myelolipomas, containing macroscopic fat, demonstrated variable CSR values (0.662–1.665), with two cases exhibiting CSR < 0.75 and thus mimicking adenoma signal patterns. This observation underscores the critical importance of correlating CSI findings with fat-saturation sequences when macroscopic fat is suspected [4]. The inclusion of pheochromocytomas further complicated threshold optimization. In our cohort, one pheochromocytoma exhibited CSR < 0.75, consistent with the 14.2% false-positive rate reported by Fassnacht et al. [4]. This finding reinforces the recommendation for mandatory biochemical correlation in all hormonally suspicious lesions, regardless of CSI phenotype. Our data suggest that in cohorts enriched with lipid-poor adenomas and lipid-containing non-adenomas, a CSR cutoff of ≤ 0.85–0.90 may be more appropriate, provided it is integrated within a multiparametric diagnostic algorithm. The modest sensitivity of SII (63.9%) represents a notable departure from meta-analytic benchmarks of 85–90% [6]. This underperformance is directly attributable to our cohort’s high prevalence of non-suppressing adenomas, wherein SII values were negative or < 14.1% despite histologic confirmation of adenoma. The 12.9% non-suppressing rate aligns with the upper range of literature reports (8–13%) [16], likely reflecting our institution’s referral bias toward indeterminate or atypical lesions requiring surgical resection rather than confident radiologic diagnosis. The pathophysiologic basis for non-suppression remains incompletely understood but likely relates to insufficient intracellular lipid accumulation in tumors with high proteinaceous content or subtle microhemorrhage [16]. Notably, two of our five non-suppressing adenomas manifested subclinical cortisol hypersecretion, suggesting that functionally active adenomas may exhibit altered lipid metabolism. This observation warrants further investigation in larger cohorts to determine whether hormonal status should inform CSI interpretation thresholds. Technical factors may have further compromised SII performance. Although we employed a 3.0T scanner with optimized TE values (1.2/2.4 ms), interobserver variability in ROI placement—particularly the inadvertent inclusion of microscopic foci of hemorrhage—can artifactually reduce signal loss on OP images. Wang et al. recently demonstrated that deep learning-based automated ROI segmentation reduces this variability by 41%, achieving ICC values > 0.90 compared to 0.82–0.84 in our manual approach [20]. This technological advancement may substantially improve SII sensitivity in clinical practice. Our novel finding that IP paraspinal muscle SI was significantly lower in adenomas (median 452.5 AU) versus non-adenomas (768.0 AU, p < 0.001) validates its emerging role as an internal reference standard [17]. The optimal cutoff of ≤ 671.5 AU yielded fair discriminative capacity (AUC 0.729), complementing CSR in equivocal cases. This phenomenon likely reflects underlying compositional differences: non-adenomatous lesions (particularly pheochromocytomas, hemangiomas, and inflammatory masses) often exhibit hemorrhagic or proteinaceous components that elevate adjacent muscle signal through susceptibility effects or shared vascular supply. While promising, paraspinal muscle SI is influenced by patient age, muscle mass, and chronic disease [18]. In our cohort, cachectic patients with malignant lesions exhibited the highest muscle SI values, potentially confounding interpretation. We advocate for standardized acquisition at the T12–L1 level and cautious interpretation in patients with sarcopenia or metabolic disorders. Our data support a refined, tiered diagnostic algorithm for adrenal mass characterization: CSR ≤ 0.71, diagnose lipid-rich adenoma with high confidence (specificity 96%); CSR > 1.20, strongly favor non-adenoma and pursue dedicated work-up; CSR 0.72–0.90, consider lipid-poor adenoma and correlate with DWI (ADC > 1.4 × 10− 3 mm²/s), lesion morphology, and clinical context; PM-IP SI > 671.5, increases suspicion for non-adenoma (+ LR 2.4). All lesions with clinical or biochemical suspicion for pheochromocytoma require validation regardless of CSI results [4]. This multiparametric approach acknowledges that universal CSI cutoffs are suboptimal and aligns with recent calls for integrated imaging strategies [2]. Several limitations merit acknowledgment. First, the retrospective design introduced selection bias toward surgically resected lesions. Second, interobserver variability in manual ROI placement, despite good concordance (ICC 0.82), may have introduced measurement noise, particularly for small lesions. Third, the heterogeneous non-adenoma group precluded subgroup-specific ROC analysis for pheochromocytoma and myelolipoma due to limited sample sizes. Fourth, we did not correlate CSI results with DWI or DCE washout patterns, which could have improved diagnostic accuracy for lipid-poor lesions as demonstrated by Wang et al. [19]. Finally, our single-center, single-vendor protocol limits generalizability to other field strengths (1.5T) and MRI platforms. Future research should focus on prospective multicenter validation of these threshold adjustments, integration of automated deep learning-based quantification to reduce interobserver variability [20], and development of synthetic CSI sequences to enhance accessibility [21]. The incorporation of radiomics features and machine learning classifiers may further refine differentiation of lipid-poor adenomas from mimics. In conclusion, CSI-derived CSR demonstrates robust discriminatory performance for adrenal adenoma diagnosis but requires a higher cutoff value (≤ 0.895) in cohorts enriched with lipid-poor adenomas and diverse non-adenomatous pathologies. The suboptimal sensitivity of SII underscores the challenge of non-suppressing adenomas, while paraspinal muscle SI emerges as a valuable ancillary biomarker. Radiologists should adopt a multiparametric, population-aware approach to adrenal mass characterization, recognizing that conventional CSI cutoffs may require adjustment in complex clinical scenarios. Fig. 1Receiver operating characteristic (ROC) curve for chemical shift ratio (CSR) in differentiating adrenal adenomas from non-adenomatous lesions. AUC = 0.822 (95% CI: 0.718–0.926, p < 0.001). Optimal cut-off: ≤0.895 (sensitivity 83.3%, specificity 80.6%) Receiver operating characteristic (ROC) curve for chemical shift ratio (CSR) in differentiating adrenal adenomas from non-adenomatous lesions. AUC = 0.822 (95% CI: 0.718–0.926, p < 0.001). Optimal cut-off: ≤0.895 (sensitivity 83.3%, specificity 80.6%) Fig. 2Receiver operating characteristic (ROC) curves for signal intensity index (SII) and paraspinal muscle in-phase signal intensity (PM-IP SI) in differentiating adrenal adenomas from non-adenomatous lesions. SII: AUC = 0.706 (95% CI: 0.576–0.836, p = 0.004), optimal cut-off ≥ 15.1%. PM-IP SI: AUC = 0.729 (95% CI: 0.608–0.850, p = 0.001), optimal cut-off ≤ 671.5 AU Receiver operating characteristic (ROC) curves for signal intensity index (SII) and paraspinal muscle in-phase signal intensity (PM-IP SI) in differentiating adrenal adenomas from non-adenomatous lesions. SII: AUC = 0.706 (95% CI: 0.576–0.836, p = 0.004), optimal cut-off ≥ 15.1%. PM-IP SI: AUC = 0.729 (95% CI: 0.608–0.850, p = 0.001), optimal cut-off ≤ 671.5 AU