ELSEVIER
Diagnostic & Interventional Imaging
Original article/Cancer imaging
Differentiation between adrenocortical carcinoma and lipid-poor adrenal adenoma using a multiparametric MRI-based diagnostic algorithm
Check for updates
Carmelia Oloukoja,b, Anthony Dohana,b,c, Martin Gaillardb,c,d, Christine Hoeffele, Lionel Groussin-Rouillerb,c,f, Jerome Bertheratb,c,f, Anne Jouinotb,c,f, Guillaume Assiéb,c,f, David Fuksb,d, Mathilde Sibonyb,g, Philippe Soyera,b, Anne-Sophie Jannoth,i, Maxime Barata,b,c,*
a Department of Radiology, Hôpital Cochin, AP-HP, 75014 Paris, France
b Université Paris Cité, Faculté de Médecine, 75006 Paris, France
” Génomique et Signalisation des Tumeurs Endocrines, Institut Cochin, INSERM U 1016, CNRS UMR8104, Université Paris Cité, 75014 Paris, France
d Department of Pancreatic and Endocrine Surgery, Hôpital Cochin, AP-HP, 75014 Paris, France
e Department of Radiology, Hôpital Robert Debré, CRESTIC, URCA, 51000 Reims, France Department of Endocrinology, Hôpital Cochin, AP-HP, 75014 Paris, France
% Department of Pathology, Hôpital Cochin, AP-HP, 75014 Paris, France
h AP-HP.Centre- Université Paris Cité, Hôpital Européen Georges Pompidou, Medical Informatics, Biostatistics and Public Health Department, 75015, Paris, France INSERM, UMR_S1138, Cordeliers Research Center, Université Paris Cité, 75006 Paris, France
ARTICLE INFO
Keywords:
Adrenal gland disease Adrenal gland neoplasm Adrenocortical carcinoma Diagnostic algorithm Magnetic resonance imaging
ABSTRACT
Purpose: The purpose of this study was to evaluate the capabilities of multiparametric magnetic resonance imag- ing (MRI) in differentiating between lipid-poor adrenal adenoma (LPAA) and adrenocortical carcinoma (ACC). Materials and methods: Patients of two centers who underwent surgical resection of LPAA or ACC after multi- parametric MRI were retrospectively included. A training cohort was used to build a diagnostic algorithm obtained through recursive partitioning based on multiparametric MRI variables, including apparent diffu- sion coefficient and chemical shift signal ratio (i.e., tumor signal intensity index). The diagnostic performan- ces of the multiparametric MRI-based algorithm were evaluated using a validation cohort, alone first and then in association with adrenal tumor size using a cut-off of 4 cm. Performances of the diagnostic algorithm for the diagnosis of ACC vs. LPAA were calculated using pathology as the reference standard.
Results: Fifty-four patients (27 with LPAA and 27 with ACC; 37 women; mean age, 48.5 ± 13.3 [standard devi- ation (SD)] years) were used as the training cohort and 61 patients (24 with LPAA and 37 with ACC; 47 women; mean age, 49 ± 11.7 [SD] years) were used as the validation cohort. In the validation cohort, the diagnostic algorithm yielded best accuracy for the diagnosis of ACC vs. LPAA (75%; 46/61; 95% CI: 55-88) when used without lesion size. Best sensitivity was obtained with the association of the diagnostic algorithm with tumor size (96%; 23/24; 95% CI: 80-99). Best specificity was obtained with the diagnostic algorithm used alone (76%; 28/37; 95% CI: 60-87).
Conclusion: A multiparametric MRI-based diagnostic algorithm that includes apparent diffusion coefficient and tumor signal intensity index helps discriminate between ACC and LPAA with high degrees of specificity and accuracy. The association of the multiparametric MRI-based diagnostic algorithm with adrenal lesion size helps maximize the sensitivity of multiparametric MRI for the diagnosis of ACC.
@ 2024 Published by Elsevier Masson SAS on behalf of Société française de radiologie.
List of abbreviations: AA, Adrenal adenoma; ACC, Adrenocortical carcinoma; ADC, Apparent diffusion coefficient; AL, Adrenal lesion; ALR, Adrenal-to-liver ratio; APW, Absolute percentage washout; ASR, Adrenal-to-spleen ratio; CI, Confidence interval; CT, Computed tomography; DWI, Diffusion-weighted imaging; FS, Fat-saturated; GBCA, Gadolinium-based contrast agent; HASTE, Half-Fourier acquisition single shot turbo spin echo; HU, Hounsfield unit; IP, In-phase; IQR, Interquartile range; LPAA, Lipid-poor adrenal adenoma; MRI, Magnetic resonance imaging; OP, Out-of-phase; PET, Positron emission tomography; ROI, Region of interest; RPW, Relative percentage washout; SD, Standard deviation; SI, Signal intensity; SII, Signal intensity index; TSE, Turbo spin-echo
* Corresponding author.
E-mail address: maxime.barat@aphp.fr (M. Barat).
https://doi.org/10.1016/j.diii.2024.03.005
1. Introduction
Adrenal incidentaloma is depicted in 2% to 7% of patients in the gen- eral population and defined as an adrenal lesion (AL) larger than 10 mm that is incidentally discovered on an imaging examination [1-3]. At the time of initial detection, assessment of the risk of malignancy is one important concern. Approximately 80% of adrenal incidentalomas are benign adrenal adenomas (AA) [2,3]. Identification of benign AA is rela- tively straightforward for homogenous and lipid-rich AA (i.e., < 10 Houns- field units [HU]) smaller than 4 cm on non-contrast computed
tomography (CT) [1]. Apart from these benign features, additional imaging is generally recommended [1]. The characterization of lipid-poor AA (LPAA) was historically based on washout calculation on contrast- enhanced CT, that is now being debated, or using chemical-shift magnetic resonance imaging (MRI) sequences, but the accuracy for the diagnosis of LPAA remains unperfect, with reported values under 90% [4-10].
Characterization of AL also relies on the assessment of adrenal hor- monal excess, and for patients with oversecretion, the diagnosis is obtained on the basis of the results of laboratory tests. In other situations, the diagnosis of primary adrenal malignancy (i.e., adrenocortical carci- noma [ACC]) is relatively straightforward when loco-regional involve- ment (i.e., European Network for the Study of Adrenal Tumors [ENSAT] stage III or IV tumors) or distant metastases are present. Similarly, when a patient with a known cancer develops AL during the follow-up, the diagnosis of adrenal metastasis is highly likely. However, in the remain- ing situations, the assessment of malignancy of LPAA can be challenging and surgery remains necessary. In these cases, the distinction mostly concerns LPAA, estimated to represent approximately one third of benign AAs and early stage ACCs (i.e., ENSAT stage I and II) [11,12]. Although, tumor size is a sensitive criterion for the diagnosis of ACC using a cut-off of 4 cm, it yields modest specificity [13]. Similarly, labora- tory tests may help differentiate between benign and malignant AL but there is some degrees of overlap and laboratory tests are not accurate enough to discriminate between ACC and LPAA [21]. Consequently, despite improvement in non-invasive diagnosis, it can be estimated that 25% of adrenalectomies, which are performed for atypical benign ALs to rule out malignancy, are futile and could be avoided [14,15].
Despite demonstrated advantages in tissue characterization, MRI is still considered as a second line imaging modality for the assess- ment of adrenal lesions [6-9]. In this regard, a consensus on the use of MRI was obtained for chemical shift imaging sequences only. Other MRI sequences such as diffusion-weighted imaging (DWI), with apparent diffusion coefficient (ADC) calculation have led to conflict- ing results in terms of AL characterization and their use remains debated [15-17]. Accordingly, the actual capabilities of MRI for the characterization of adrenal malignancy, and more specifically of ACC, must be clarified.
The purpose of this study was to evaluate the capabilities of multi- parametric MRI in differentiating between LPAA and ACC using two independent cohorts of patients from two centers.
2. Materials and methods
2.1. Patients
This retrospective, two-center study was approved by institu- tional review board (AAA-2021-08024). The requirement for informed consent from patients was waived.
The database of the department of pathology of Center 1 was que- ried from May 2006 to October 2019 to identify patients who under- went adrenalectomy. Patients were further included when: (i), they were older than 18 years; (ii), they had available preoperative CT and MRI examinations performed less than three months before surgery; and (iii), results of histopathological analysis included a definite diag- nosis of ACC ENSAT stage I or II or LPAA (defined as typical adenoma on histological examination with a spontaneous attenuation >10 Hounsfield units [HU] on unenhanced CT examination) with Ki67 rate and histopathological Weiss score. Patients without these char- acteristics and those with AL other than ACC or LPAA were excluded. Then, an independent validation cohort of patients who underwent CT and multiparametric MRI examinations from November 2019 to January 2023 in Center 1 and from January 2011 to January 2020 in Center 2 was built using the same inclusion and exclusion criteria.
2.2. MRI protocol
CT examinations were performed using a single source helical CT equipment (Revolution HD®, General-Electric Healthcare; Somatom Sensation® 64, Siemens Healthineers; or Somatom Definition® Flash, Siemens Healthineers). Unenhanced acquisition was obtained in all patients to calculate spontaneous attenuation values of ALs [7].
Multiparametric MRI examinations were performed with a 1.5-Tesla unit and included DWI using three b values (b = 0, b = 400 and b = 800 s/ mm2) and ADC map, T2-weighted half-Fourier acquisition single shot turbo spin-echo (HASTE), fat-saturated (FS) T2-weighted multi-shot turbo spin-echo (TSE), T1-weighted images (in-phase [IP] and out-of- phase [OP]), and multiphase contrast-enhanced T1-weighted sequences (30-40 s, 65-80 s, 3 min, 5 min and 10 min after intravenous adminis- tration of a gadolinium-based contrast agent (GBCA) at a dose of 0.1 mL/ kg and a rate of 2 mL/s. MRI protocol is reported in Table 1. The same MRI protocol was used in the two centers.
2.3. Image analysis
Imaging examinations were reviewed independently using a workstation (DirectView®, 11.4.0.1253 sp1 version, Carestream Health) by two independent radiologists (C.O. and M.B., with 5- and 11 years of experience in abdominal imaging, respectively) using a standardized data collection form. The two radiologists were blinded to any clinico-biological information. MRI examinations were qualita- tively and quantitatively analyzed.
2.3.1. Qualitative analysis
MRI examinations were qualitatively analyzed in terms of AL mar- gins (well-defined vs. ill-defined), signal homogeneity and intensity on DWI, and signal intensity of AL on FS T2-weighted images and
| Sequence | Diffusion | VIBE Dixon | FS T2 BLADE | Dynamic multiphase contrast-enhanced |
|---|---|---|---|---|
| Plane | Transverse | Transverse | Transverse | Transverse |
| Acquisition type | 2D | 3D | 2D | 3D |
| Slice thickness (mm) | 5 | 3 | 5 | 2.2 |
| Gap (mm) | 0.5 | − | 1 | − |
| Repetition time (ms) | 6000 | 6.83 | 2540 | 4 |
| Echo time (ms) | 56 | 2.4 | 121 | 1.5 |
| b-value (s/mm2) | 0, 400, 800 | - | - | - |
| Flip angle (°) | 90 | 10 | 142 | 14 |
| Pixel bandwidth (Hz) | 2440 | 475 | 500 | 345 |
| Acquisition matrix | 128 × 104 | 320 x 182 | 256 x 256 | 384 × 234 |
| Respiratory control | Respiratory-triggered | Respiratory-triggered | Respiratory-triggered | Respiratory-triggered |
2D indicates two-dimensional; 3D indicates three-dimensional; FS indicates fat saturated; VIBE indicates volumetric interpolated breath hold examination.
BLADE is the name of periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) sequence of Siemens Healthinners.
categorized as hyper-, iso- or hypointense using the signal intensity of the spleen as a reference.
2.3.2. Quantitative analysis
AL dimensions were measured using calipers on MR images in the axial plane using venous phase images. Signal intensity index (SII) of AL were calculated on IP and OP images at the same level after draw- ing manually a region of interest (ROI) in the largest solid portion of the AL (at least two thirds of the AL) with exclusion of lesion borders to avoid partial volume effect, using the following equation [16-18]:
SII = (SI IP (adrenal) - SI OP (adrenal))/(SI IP (adrenal)) × 100 (1)
Lesion ADC measurements were performed using similar ROIS. Standard deviations (SD) of ADC values were also recorded. Adrenal- to-spleen ratio (ASR) and adrenal-to-liver ratio (ALR) were calculated on IP and OP images using the same ROI as for SII, in the solid portion of the ALs, in the spleen and in the liver before and at 30 to 40 s (arte- rial phase images), 65 to 80 s (venous phase images), 3 min, 5 min and 10 min after intravenous administration of GBCA using the fol- lowing two equations [16]:
ASR = [(SI OP adrenal/SI OP spleen)/(SI IP adrenal/SI IP spleen) - 1] × 100 (2)
ALR = [(SI OP adrenal/SI OP liver)/(SI IP adrenal/SI IP liver) - 1] x 100 (3)
Relative percentage wash-out (RPW) and absolute percentage wash-out (APW) were calculated using similar ROIs on MR images obtained at 3-, 5- and 10 min after intravenous administration of GBCA using the following equations in which early SI is measured on venous phase images and delayed on contrast-enhanced MR images obtained at 3, 5 and 10 min [4,5]:
APW = [(early SI - delayed SI)/(early SI - unenhanced SI)] ×100 (4)
RPW = [(early SI - delayed SI)/early SI] × 100 (5)
2.4. Statistical analysis
Statistical analysis was performed using R software (version 4.1.0, R-foundation, http://www.r-project. org/). The distribution of
quantitative variables was assessed using the Shapiro-Wilk test. Quantitative variables were reported as means ± standard deviations (SD) and ranges or medians, interquartile ranges (IQR) and ranges depending on the normality of their distribution [19]. Non-paramet- ric Mann Whitney U test was used to compare quantitative variables. Qualitative variables were reported as raw numbers, proportions, and percentages and compared between patients with ACC and those with LPAA using x2 test or Fisher exact test. Both tests were two- tailed and significance was set at P < 0.05.
The standard of reference for the diagnostic of LPAA vs. ACC was the result of histopathological examination obtained after surgical adrenalectomy for all ALs. A Weiss score ≥ 3 was considered for the diagnosis of ACC.
Using the training cohort, a supervised recursive partitioning method with LPAA vs. ACC as predictive variable was performed including all evaluated qualitative (hypersignal on DW images, hypersignal on FS T2-weighted TSE images, signal homogeneity on DW images) and quantitative (ADC, lesion SII, APW at 3, 5 and 10 min, RPW at 3, 5 and 10 min) variables in order to build an algo- rithm for the diagnosis of ACC vs. LPAA [20]. This method selected the relevant variables and selected cut-offs for quantitative one. Interob- server reproductibility for the algorithm was evaluated using Bland Altman plots only for selected features [21]. The random forest method was also applied using the same data.
Performances of the two methods (recursive partitioning and random forest) for the diagnosis of ACC vs. LPAA were compared by estimating sensitivity, specificity, and accuracy and their 95% confidence intervals (CI) for both methods. Then, sensitivity, specificity, and accuracy for the diagnostic of ACC vs. LPAA were estimated with the validation cohort for the method reaching the highest performances without and with consid- eration of the largest AL size using 4 cm as cut-off. Finally, sensitivity, specificity, and accuracy for the diagnostic of ACC vs. LPAA of the algo- rithm alone were compared to those of lesion size alone and those of the combination of algorithm and lesion size using McNemar test.
3. Results
3.1. Patients
Fifty-four patients with 27 ACCs and 27 LPAAs were included in the training cohort. There were 37 women and 17 men with a mean age of 48.5 ± 13.3 (SD) years (range: 18-74 years). Sixty-one patients
406 patients with adrenalectomy from 2006 to 2020
Excluded patients (n =175; 43%) Histopathological results other than LPAA or ACC
231 eligible patients with LPAA (n = 156) or ACC (n = 75)
Excluded patients (n = 177; 43.5%) Incomplete data (no MRI, no contrast-enhanced CT and/or MRI)
54 eligible patients with LPAA or ACC and complete data
27 ACC
27 LPAA
MRI indicates magnetic resonance imaging; DWI indicates diffusion-weighted imaging; CT indicates computed tomography; ACC indicates adrenocortical carcinoma; LPAA indicates lipid-poor adrenal adenoma.
| Variable | Patients with ACC (n = 27) | Patients with LPAA (n =27) | Pvalue |
|---|---|---|---|
| Age (year) | 46 (37, 60) | 53 (46, 61) | 0.33 |
| [18-74] | [34-70] | ||
| Female | 20 (20/27; 74) | 17 (17/27; 63) | 0.59 |
| BMI | 25.5 (23,33) | 25.1 (23,31) | 0.15 |
| [18.5-48.4] | [18.4-34.3] | ||
| Ki67 | 25 (14; 40) | 1 (1; 1) | < 0.001 |
| [5-95] | [1-2] | ||
| Weiss score | 7 (6.5, 9) | 0 (0, 1) | < 0.001 |
| [4-9] | [0-2] | ||
| Secretion syndrome | 16 (16/27; 59) | 9 (9/27; 33) | 0.10 |
| Biological oversecretion | 24 (24/27; 89) | 16 (16/27; 59) | 0.03 |
| Tumor size | |||
| Length (mm) | 89 (55, 104.5) | 40 (29, 45.5) | < 0.001 |
| [39-208] | [17-77] | ||
| Width (mm) | 65 (51.5, 96) | 32 (20.5, 40) | < 0.001 |
| [32-166] | [12-65] | ||
| Height (mm) | 75 (53.5, 107) | 40 (25, 46) | < 0.001 |
| [40-226] | [12-70] |
ACC indicates adrenocortical carcinoma; BMI indicates body mass index; F indicates female; LPAA indicates lipid-poor adrenal adenoma; M indicates male; MRI indicates magnetic resonance imaging.
Qualitative variables are expressed as raw numbers; numbers into parentheses are proportions followed by percentages. Quantitative variables are expressed as medians; numbers into parentheses are first (Q1) and third (Q3) quartiles; numbers into brackets are ranges.
with 37 ACCs and 24 LPAAs were included in the validation cohort. There were 47 women and 14 men with a mean age of 49 ± 11.7 (SD) years (range: 21-78 years). Fig. 1 shows the flow chart of the study for the training cohort. Characteristics of the 54 patients of the train- ing cohort are reported in Table 2. No differences were found between patients with LPAA and those with ACC for age, sex, body mass index and cortisol secretion but differences were found for Ki67 rate, Weiss score, and AL size.
3.2. Qualitative results
On T2-weighted images, LPAAs of the training cohort, were more frequently homogeneous (8/27; 30%) than ACCs (1/27; 4%) (P = 0.02), with more frequent well-defined margins (23/27; 85%) than ACCs (13/27; 48%) (P = 0.008) and were more frequently isointense relative to the spleen (23/27; 85%) than ACCs (15/27; 56%) (P = 0.04) on FS T2-weighted images. On DWI, ACCs were more frequently hyperin- tense relative to the spleen (25/27; 93%) than LPAAs (17/27; 63%) (P=0.02) (Table 3).
| Feature | All patients (n = 54) | ACC (n = 27) | LPAA (n = 27) | P value |
|---|---|---|---|---|
| Homogeneous on FS T2W | 0 (0/54; 17) | 1 (1/27; 4) | 8 (8/27; 30) | 0.02 |
| Isointense on FS T2W | 38(38/54; 70) | 15 (15/27; 56) | 23 (23/27; 85) | 0.04 |
| Hyperintense on FS T2W | 15 (15/54; 28) | 11 (11/27; 41) | 4 (4/27; 15) | 0.07 |
| Hypointense on FS T2W | 1 (1/54; 2) | 1 (1/27; 4) | 0 (0/27; 0) | > 0.99 |
| Isointense on DWI | 12 (12/55; 22) | 2 (2/27; 7) | 10 (10/27; 37) | 0.01 |
| Hyper intense on DWI | 42 (42/54; 78) | 25 (25/27; 93) | 17 (17/27; 63) | 0.02 |
| Well-defined borders | 36 (36/54; 67) | 13 (13/27; 48) | 23 (23/27; 85) | 0.008 |
ACC indicates adrenocortical carcinoma; DWI indicates diffusion-weighted imaging; FS indicates fat saturated; LPAA indicates lipid-poor adrenal adenoma; T2W indicates T2-weighted.
Signal intensity was compared to that of the spleen on T2-weighted images and diffu- sion-weighted imaging. Variables are expressed as raw numbers; numbers into paren- theses are proportions followed by percentages.
3.3. Quantitative results
In the training cohort, LPAAs showed greater median SII (28.3; range: - 5-74.8) than ACCs (4.2; range: - 14.3-27.5) (P < 0.01). Simi- larly, median mean ADC was greater in LPAAs (1.100 x 10-3 m2/s; range: [0.630-2.200] × 10-3 m2/s) than in ACCs (0.870 x 10-3 m2/s; range: [0.470-1.434] × 10-3 m2/s) (P < 0.02) (Figs. 2, 3, 4; Table 4).
No significant differences in APW and RPW were found between ACCs and LPAAs on MRI nor for the other quantitative variables. Results of univariable analysis for the validation cohort are reported in Table 5.
3.4. Recursive partitioning
Using recursive partitioning method, an algorithm that included AL mean ADC of AL and SII only was built using the training cohort (Fig. 5). The classification obtained with the cut-off of 28 for lesion SII and 1.452 × 10-3 m2/s for mean ADC yielded 100% sensitivity (27/27; 95% CI: 87-100), 78% specificity (21/27; 95% CI: 58-91) and 89% accuracy (48/54; 95% CI: 77-99) for the diagnostic of ACC. Bland Alt- man plots showed that features included in the algorithm (i.e., SII and mean ADC) were reproducible between readers without systematic bias (Fig. 6). Random forest method classification yielded 74% sensi- tivity (20/27; 95% CI: 54-89), 74% specificity (20/27; 95% CI: 54-89) and 74% accuracy (40/54; 95% CI: 60-85) for the diagnosis of ACC that were lower than those of recursive partitioning method (P < 0.01 for sensitivity, specificity and accuracy). Therefore, the diag- nostic algorithm obtained through recursive partitioning method was retained as a predictive model.
When applied to the validation cohort, the diagnostic algorithm obtained through recursive partitioning yielded 75% sensitivity (18/ 24; 95% CI: 55-88), 76% specificity (28/37; 95% CI: 60-87) and 75% accuracy (46/61; 95% CI: 55-88) for the diagnosis of ACC vs. LPAA (Fig. 6, Table 6).
Using 4 cm as a cut-off value, largest AL size yielded 83% sensitiv- ity (20/24; 95% CI: 64-99), 30% specificity (11/37; 95% CI: 17-46) and 51% accuracy (31/61; 95% CI: 32-69) for the diagnosis of ACC vs. LPAA. When associated to largest AL size, the diagnostic algorithm yielded 96% sensitivity (23/24; 95% CI: 80-99), 57% specificity (21/ 37; 95% CI: 41-71) and 72% accuracy (44/61; 95% CI: 52-86) for the diagnosis of ACC vs. LPAA (Table 6).
4. Discussion
Our results indicate that the differentiation between ACC and LPAA can be achieved using a diagnostic algorithm that includes AL SII and mean ADC on MRI, with good accuracies (89% in the training cohort and 75% in the validation cohort). Moreover, when associated to largest AL size using a cut-off value of 4 cm, multiparametric MRI may improve the sensitivity for the diagnosis of ACC vs. LPAA.
Previous studies found that SII on MRI was the most reliable fea- ture for discriminating between benign and malignant AL when CT findings are indeterminate [22]. Although qualitative visual evalua- tion is widely used to identify LPAA with MRI, Israel et al. showed that qualitative analysis may be less sensitive for the characterization of LPAA than quantitative analysis [9]. Several studies reported high confidence level in discriminating between LPAA and malignant AL, mostly represented by metastases and much less by ACCs [8,23]. In a prospective study involving 97 patients with 69 LPAAs and 28 non- AA lesions, including eight ACCs, Dalavia et al. found significant dif- ferences in mean lesion SII between benign and malignant ALs (P < 0.001) with 88% sensitivity and 75% specificity for the diagnosis of LPAA [24].
In our study, we did not include the findings at positron emission tomography (PET)/CT in our algorithm because the results of this examination were not available for all patients. It is well established
ROI : 7.3 cm2
ROI : 7.3 cm2
ROI : 7.3 cm2
Mean : 27 HU
Mean : 79 HU
Mean : 63 HU
SD: 20 HU
SD: 31 HU
SD: 29 HU
A
B
C
D
ROI : 7 cm2
ROI : 7 cm2
ROI : 2.2 cm3 Mean : 1.040 10 3 mm2/ SD: 0.230 10 3 mm2/5
Mean : 215
Mean : 204
SD : 15
SD : 20
E
F
G
H
A, B, C: Unenhanced (A) and contrast-enhanced CT images obtained at 60 s (B) and 10 min (C) after intravenous administration of iodinated contrast material in the axial plane show a nodule (arrow) of 46 x 38 mm in the right adrenal gland, with a mean attenuation value of 27 ± 27 (standard deviation [SD]) Hounsfield units (HU), 79 + 31 (SD) HU and 63 ± 29 (SD) HU, respectively, classified as atypical using CT criteria (absolute percentage wash out = 31% and relative percentage wash out = 20%).
D: Fat saturated T2-weighted image in the axial plane show heterogeneous nodule (arrow) harboring hypersignal compared to the spleen.
E, F: Unenhanced T1-weighted in phase (E) and out-of-phase (F) MR images in the axial plane show no drop in signal intensity of the nodule between in-phase (215 ± 15 [stan- dard deviation (SD)]) and out-of-phase (204 ± 20 [SD]) with lesion signal intensity index = 4.
G, H: On diffusion-weighted MR images (b = 1000 s/mm2) in the axial plane, the adrenal nodule shows hypersignal with a mean apparent diffusion coefficient of 1.040 ± 0.230 (SD) × 10-3m2/s.
With a cut-off of 1.452 x 10-3 m2/s for mean ADC and 28 for lesion signal intensity index, this lesion is correctly classified as adrenocortical carcinoma.
ROI : 2.9cm2
Mean : 32 HU
ROI : 2.9 cm2
ROI : 2.9cm3
SD: 24 HU
Mean : 95 HU
SD: 35 HU
Mean : 65 HU
SD 23 HU
A
B
C
D
ROI : 2.9cm2
Mean : 230
ROI : 2.9cm2
ROI : 2.9cm2
SD: 40
Mean : 200
Mean : 1/199 103 mm2/s SD: 0:253 10-3 mm2/s
SD: 37
E
F
G
H
D: Fat saturated T2-weighted image in the axial plane shows heterogeneous nodule (arrow) harboring hypersignal compared to the spleen E, F: Unenhanced T1-weighted in phase (E) and out-of-phase (F) MR images in the axial plane show no drop in signal intensity of the nodule (arrow) between in-phase (230 ± 40 [standard deviation (SD)]) and out-of-phase (200 ± 37 [SD]) images with lesion signal intensity index of 10.
G, H: Diffusion-weighted MR image (b = 1000 s/mm2) in the axial plane shows the nodule in hypersignal with a mean apparent diffusion coefficient of 1.799 ± 0.253 (SD) ×10-3 m2/s.
This lesion was correctly classified as lipid-poor adrenal adenoma.
that PET/CT imaging has now a pivotal role in the characterization of adrenal malignancies [25]. However, LPAA may show marked uptake of 2-deoxy-2-[fluorine-18] fluoro-D-glucose (18F-FDG) on PET/CT in 27 to 46% of patients, resulting in false-positive findings [11]. Conse- quently, the use of 18F-FDG PET for the differentiation of ACC vs. LPAA in association with MRI requires further evaluation.
DWI has demonstrated capabilities for differentiating between benign and malignant tumors in many organs because of the greater cellular density in malignant ones [26,27]. Considering AL, studies have reported conflicting results. In a retrospective study, Miller et al. observed a large overlap in ADC values between AAs and non-AA lesions and concluded that ADC did not help discriminate between
ROV: 1.2 cm2
ROI : 1,2 cm2
ROI : 1,2 cm2
Mean : 15 HU
Mean : 82 HU
Mean : 44 HU
SP: 10 HU
SD: 30 HU
0
SD: 30 HU
A
B
C
D
ROI : 0.8 cm2
ROI : 0.8 cm2
ROI : 1 cm.
Mean : 245
Mean : 130 HU
Mean : 1.470 103 mln2/s
SD: 23
SD: 33 HU
SD: 0.210 10-3 mm2/s
E
F
G
H
P
A, B, C: Unenhanced (A) and contrast-enhanced CT images obtained at 60 s (B) and 10 min (C) after intravenous administration of iodinated contrast material in the axial plane show a nodule (arrow) of 20 x 15 mm in the right adrenal gland. Attenuations values within the different regions of interest are 15 ± 10 (standard deviation ([SD]) Hounsfield units (HU), 62 ± 30 (SD) HU and 44 ±30(SD) HU, respectively corresponding to absolute percentage wash out [APW] of 38% and relative percentage wash out [RPW] of 29%), correspond- ing to atypical adrenal lesion using CT criteria (homogenous mass, attenuation value <10 HU, larger diameter < 4 cm, and well-defined margins)
D: Fat saturated (FS) T2-weighted images in the axial plane shows adrenal nodule (arrow), which is isointense relative to the spleen.
E, F: In-phase (E) and out-of-phase (F) unenhanced T1-weighted MR images in the axial plane show homogenous signal drop of the adrenal nodule between in-phase (245 ± 23 [standard deviation (SD)]) and out-of-phase (130 ± 33 [SD]) images and lesion signal intensity index of 45.
G, H: Diffusion-weighted MR image (b value =1000 s/mm2) in the axial plane shows hyperintense adrenal nodule with a mean apparent diffusion coefficient of 1.470 ± 210 (SD) × 10-3m2/s. With a cut-off value of 1.452 x 10-3 m2/s for mean ADC and 28 for lesion signal intensity index, this lesion was correctly classified as lipid-poor adrenal adenoma.
| Feature | All patients (n = 54) | ACC (n = 27) | LPAA (n = 27) | P value |
|---|---|---|---|---|
| ADC (10-3 m2/s) | 0.945 (0.753, 1.160) | 0.870 (0.745, 0.953) | 1.100 (0.938, 1.496) | 0.02 |
| [0.470-2.200] | [0.470-1.434] | [0.630-2.200] | ||
| ADC standard deviation (10-3 m2/s) | 0.120 (0.071, 0.168) | 0.105 (0.068, 0.145) | 0.134 (0.086, 0.190) | 0.17 |
| [0.006-0.359] | [0.050-0.359] | [0.006-0.300] | ||
| Lesion SII | 9.6 (4, 28) | 4.2 (0,9) | 28.3 (11,53) | < 0.001 |
| [-14.3-74.8] | [-14.3-27.5] | [-5-74.8] | ||
| Liver SII | 2.9 (0,8) | 2.7 (0,7) | 4.5 (0, 12) | 0.23 |
| [-27.1-51.3] | [-20-38.1] | [-27.1-51.3] | ||
| Spleen SII | 1.8 (-6, 7) | 4.2(-1,8) | -1.9 (-11,6) | 0.40 |
| [-17.5-64.8] | [-11.6-20.9] | [-17.5-64.8] | ||
| ASR | 85.7 (76, 100) | 85.7 (77, 106) | 87.3 (72,98) | 0.22 |
| [41-200] | [58.2-200] | [41-118.8] | ||
| ASR arterial phase | 63.8 (54,83) | 65 (57,82) | 60.6 (54, 82) | 0.66 |
| [29.2-118.2] | [45.5-112.2] | [29.2-118.2] | ||
| ASR venous phase | 83.3 (74, 102) | 85.7 (73, 111) | 81.7 (76,94) | 0.31 |
| [57.4-143.4] | [57.4-142.4] | [60-143.4] | ||
| ASR 3 min | 101 (85, 117) | 107 (88, 119) | 99.5 (78, 108) | 0.18 |
| [59-191.7] | [69.2-191.7] | [59-157.9] | ||
| ASR 5 min | 103.3 (86, 122) | 107.2 (94, 129) | 97.5 (76, 114) | 0.18 |
| [57.1-199] | [76.9-199] | [57.1-190.6] | ||
| ASR 10 min | 113.6 (89, 137) | 134.2 (105, 140) | 86.5 (72, 118) | 0.20 |
| [62.9-201.3] | [93-142] | [62.9-201.3] | ||
| APW 3 min | 6.9 (-5, 20) | 13.7 (6, 15) | 8.1 (0,23) | 0.22 |
| [-60-39.2] | [-60-29.3] | [-43.2-39.2] | ||
| APW 5 min | 16.7 (3,31) | 12.6 (0, 26) | 25.8 (9,35) | 0.16 |
| [-69-47.6] | [-69-44.4] | [-24-48] | ||
| APW 10 min | 31 (19, 40) | 21 (19, 28) | 37 (33, 41) | 0.15 |
| [-81.8-52.9] | [-81.8-48.8] | [7.8-52.9] | ||
| RPW 3 min | 5 (-3, 12) | 2 (-4, 11) | 6 (0, 14) | 0.21 |
| [-37.1-28.6] | [-37.1-19.4] | [-31-28.6] | ||
| RPW 5 min | 13 (2,21) | 9 (0, 16) | 16 (6, 22) | 0.16 |
| [-42.7-34.3] | [-42.7-29.8] | [-17.3-34.3] | ||
| RPW 10 min | 21 (13, 30) | 14 (12,25) | 22 (21, 32) | 0.22 |
| [-50.6-33.8] | [-50.6-33.8] | [5.6-33.3] |
ACC indicates adrenocortical carcinoma; ADC indicates apparent coefficient diffusion; APW indicates absolute percentage wash out; ASR indicates adrenal-to-spleen ratio; DWI indicates diffusion-weighted imaging; LPAA indicates lipid-poor adrenal ade- noma; ROI indicates region of interest; RPW indicates relative percentage wash out; SII indicates signal intensity index. Variables are expressed as medians, numbers into parentheses are first (Q1) and third (Q3) quartiles and numbers into brackets are ranges.
| Variable | ACC | LPAA | P value |
|---|---|---|---|
| Hypersignal on DWI | 24 (24/24; 100) | 28 (28/37; 76) | 0.04 |
| Median ADC | 864 (770; 992) | 1300 (1145; 1584) | 0.01 |
| ADC standard deviation | 90 (63; 148) | 176 (145; 210) | 0.15 |
| Lesion SII | 15.4 (3.9; 47.7) | 17.6 (0.4; 42.9) | 0.40 |
ACC indicates adrenocortical carcinoma; ADC indicates apparent coefficient diffu- sion; DWI indicates diffusion-weighted imaging; LPAA indicates lipid-poor adre- nal adenoma; SII indicates signal intensity index.
Qualitative variables are expressed as raw numbers; numbers into parentheses are proportion followed by percentages. Quantitative variables are expressed as medians; numbers into parentheses are first (Q1) and third (Q3) quartiles.
benign and malignant ALs [28]. In their study, medians of mean ADC were (1.61 x 10-3 m2/s; Q1, Q3: 1.27, 1.84 x 10-3 m2/s) for benign ALs and (1.67 x 10-3 m2/s; Q1, Q3: 1.41, 1.84 x 10-3 m2/s) for malig- nant ones [28]. Other studies comparing LPAA with malignant lesions other than ACC did not report significant differences between lesion types [23,29]. In a prospective study involving 108 patients, Halefoglu et al. compared ADC for the diagnosis of LPAA vs. adrenal metastasis and did not find any significant differences in mean ADC values between LPAA (1.35 ± 0.19 [SD] × 10-3 m2/s) and adrenal metastasis (1.32 ± 0.34 [SD] × 10-3 m2/s) (P=0.673) [23]. In our study, as in prior ones, ADC value used as a single variable shows limited capabil- ities for differentiating between benign and malignant AL, but its use in combination with SII yields high discriminating capabilities for the diagnosis of ACC vs. LPAA by meeting two criteria in favor of malig- nancy (i.e., hypercellularity evaluated with DWI and a low-fat fraction evaluated with chemical-shift imaging) in the external validation cohort. Large AL diameter is known to be a highly suggestive feature of malignancy and helps discriminate between LPAA and ACC but the classical cut-off of 4 cm only has moderate accuracy with an impor- tant overlap especially for ALs between 4 and 6 cm [14]. The improvement in specificity may be an important tool with or without the help of artificial intelligence that can help radiologists and physi- cians in their daily practice to improve patient care [18].
Lesion SII
20
O
differences
8
0
0
சி.
00
o
0
00
0
o
-20
-40
0
0
20
40
60
mean measurement
ADC
500
0
differences
Q
-500
0
o
0
-1500
0
0
500
1000
1500
2000
mean measurement
Our study has some limitations. One is the retrospective design with potential inclusion bias. In addition, the distributions of ACCs and LPAAs in the training cohort (27 ACCs and 27 LPAAs) and in the validation cohort (24 ACCs and 37 LPAAs) do not represent the actual distribution of these two conditions in the real life.
In conclusion, we propose a reproducible diagnostic algorithm to differentiate between ACC and LPAA on multiparametric MRI using mean ADC and SII of AL. We advocate the use of this diagnostic algo- rithm that may affect patient care by reducing the use of futile sur- gery. However, further prospective studies should be performed to validate our results in daily practice.
Training cohort: 27 ACC / 27 LPAA Validation cohort: 24 ACC / 37 LPAA
Lesion SII < 28
Lesion SII ≥ 28
Training cohort: 27 ACC / 13 LPAA Validation cohort: 19 ACC / 17 LPAA
Mean ADC < 1.452 x 10-3 mm2/s
Mean ADC ≥ 1.452 x 10-3 mm2/s
Training cohort: 27 ACC / 6 LPAA Validation cohort: 18 ACC / 9 LPAA
Training cohort: 0 ACC / 7 LPAA Validation cohort: 1 ACC / 8 LPAA
Training cohort: 0 ACC / 14 LPAA Validation cohort: 5 ACC / 20 LPAA
Prediction: ACC
Prediction: LPAA
Prediction: LPAA
SII indicates signal intensity index of adrenal lesion; ADC indicates apparent diffusion coefficient. ACC indicates adrenocortical carcinoma; LPAA indicates lipid-poor adrenal adenoma.
| Variable | Algorithm alone | Size alone | Algorithm + size | P value1 | Pvalue2 | P value3 |
|---|---|---|---|---|---|---|
| Sensitivity (%) | 75 (18/24) [55-88] | 83 (20/24) [64-99] | 96 (23/24) [80-99] | 0.68 | 0.13 | 0.48 |
| Specificity (%) | 76 (28/37) [60-87] | 30 (11/37) [17-46] | 57 (21/37) [41-71] | 0.066 | <0.001 | 0.023 |
| Accuracy (%) | 75 (46/61) [55-88] | 51 (31/61) [32-69] | 72 (44/61) [52-86] | 0.20 | 0.008 | 0.28 |
| Positive likelihood ratio | 3.08 | 1.2 | 2.2 | |||
| [1.7-5.7] | [0.9-1.6] | [1.5-3.2] | ||||
| Negative likelihood ratio | 0.33 | 0.5 | 0.07 | |||
| [0.2-0.7] | [0.2-1.6] | [0.01-0.5] |
Number into parentheses are proportions; numbers into brackets are 95% confidence intervals.
MRI indicates magnetic resonance imaging.
1 McNemar test for the comparison between diagnostic algorithm alone and size alone.
2 McNemar test for the comparison between diagnostic algorithm alone and the association of diagnostic algorithm and lesion size.
3 McNemar test for the comparison between size alone and the association of diagnostic algorithm and lesion size.
Ethical approval
The authors declare that the work described has been performed in accordance with the Declaration of Helsinki of the World Medical Association, revised in 2013, for experiments involving humans.
Informed consent and patient details
The authors declare that this report does not contain any personal information that could lead to the identification of the patients.
Funding
This work did not receive grants from funding agencies in the public, commercial, or not-for-profit sectors
Declaration of competing interest
The authors declare no actual or potential conflict of interest related to this study.
CRediT authorship contribution statement
Carmelia Oloukoi: Data curation, Investigation, Writing - origi- nal draft. Anthony Dohan: Conceptualization, Methodology, Visuali- zation. Martin Gaillard: Data curation, Resources, Investigation. Christine Hoeffel: Data curation, Visualization, Investigation. Lionel Groussin-Rouiller: Data curation, Visualization, Resources. Jérome Bertherat: Data curation, Visualization, Resources. Anne Jouinot: Data curation, Visualization, Resources. Guillaume Assié: Data cura- tion, Visualization, Resources. David Fuks: Data curation, Resources. Mathilde Sibony: Data curation, Resources. Philippe Soyer: Concep- tualization, Methodology, Visualization, Writing - review & editing. Anne-Sophie Jannot: Conceptualization, Methodology, Software, Formal analysis. Maxime Barat: Conceptualization, Supervision, Writing - review & editing.
Acknowledgements
Maxime Barat received a research grant from the Société Française de Radiologie and the Servier institute for its PhD.
References
[1] Fassnacht M, Tsagarakis S, Terzolo M, Tabarin A, Sahdev A, Newell-Price J, et al. European Society of Endocrinology clinical practice guidelines on the
management of adrenal incidentalomas, in collaboration with the European Net- work for the Study of Adrenal Tumors. Eur J Endocrinol 2023;189:G1-G42.
[2] Bovio S, Cataldi A, Reimondo G, Sperone P, Novello S, Berruti A, et al. Prevalence of adrenal incidentaloma in a contemporary computerized tomography series. J Endocrinol Invest 2006;29:298-302.
[3] Sherlock M, Scarsbrook A, Abbas A, Fraser S, Limumpornpetch P, Dineen R, et al. Adrenal incidentaloma. Endocr Rev 2020;41:775-820.
[4] Korobkin M, Brodeur FJ, Francis IR, Quint LE, Dunnick NR, Londy F. CT time-atten- uation washout curves of adrenal adenomas and nonadenomas. AJR Am J Roent- genol 1998; 170:747-52.
[5] Schieda N, Al Dandan O, Kielar AZ, Flood TA, McInnes MD, Siegelman ES. Pitfalls of adrenal imaging with chemical shift MRI. Clin Radiol 2014;69:1186-97.
[6] Boland GW, Lee MJ, Gazelle GS, Halpern EF, McNicholas MM, Mueller PR. Charac- terization of adrenal masses using unenhanced CT: an analysis of the CT litera- ture. AJR Am J Roentgenol 1998;171:201-4.
[7] Szolar DH, Kammerhuber FH. Adrenal adenomas and nonadenomas: assessment of washout at delayed contrast-enhanced CT. Radiology 1998;207:369-75.
[8] Fujiyoshi F, Nakajo M, Fukukura Y, Tsuchimochi S. Characterization of adrenal tumors by chemical shift fast low-angle shot MR imaging: comparison of four methods of quantitative evaluation. AJR Am J Roentgenol 2003; 180:1649-57.
[9] Israel GM, Korobkin M, Wang C, Hecht EN, Krinsky GA. Comparison of unen- hanced CT and chemical shift MRI in evaluating lipid-rich adrenal adenomas. AJR Am J Roentgenol 2004;183:215-9.
[10] Schloetelburg W, Ebert I, Petritsch B, Weng AM, Dischinger U, Kircher S, et al. Adrenal wash-out CT: moderate diagnostic value in distinguishing benign from malignant adrenal masses. Eur J Endocrinol 2021;186:183-93.
[11] Barat M, Cottereau AS, Gaujoux S, Tenenbaum F, Sibony M, Bertherat J, et al. Adre- nal mass characterization in the era of quantitative imaging: state of the art. Can- cers 2022;14:569.
[12] Lughezzani G, Sun M, Perrotte P, Jeldres C, Alasker A, Isbarn H, et al. The European Network for the Study of Adrenal Tumors staging system is prognostically supe- rior to the international union against cancer-staging system: a North American validation. Eur J Cancer 2010;46:713-9.
[13] Bancos I, Taylor AE, Chortis V, Sitch AJ, Jenkinson C, Davidge-Pitts CJ, et al. Urine steroid metabolomics for the differential diagnosis of adrenal incidentalomas in the EURINE-ACT study: a prospective test validation study. Lancet Diabetes Endo- crinol 2020;8:773-81.
[14] Gaujoux S, Aime A, Assie G, Ciuni R, Bonnet S, Tenenbaum F, et al. Adrenalectomy for incidentaloma: lessons learned from a single-centre series of 274 patients. ANZ J Surg 2018;88:468-73.
[15] Kahramangil B, Kose E, Remer EM, Reynolds JP, Stein R, Rini B, et al. A Modern assessment of cancer risk in adrenal incidentalomas: analysis of 2219 patients. Ann Surg 2022;275:e238-44.
[16] Hekimsoy İ, Güler E, Harman M, Elmas N. Characterization of adrenal lesions on chemical shift MRI: comparison of 1.5 T and 3 T MRI. Abdom Radiol 2019;4:3359-69.
[17] Ramamurthy NK, Moosavi B, McInnes MD, Flood TA, Schieda N. Multiparametric MRI of solid renal masses: pearls and pitfalls. Clin Radiol 2015;70:304-16.
[18] Barat M, Gaillard M, Cottereau AS, Fishman EK, Assie G, Jouinot A, et al. Artificial intelligence in adrenal imaging: a critical review of current applications. Diagn Interv Imaging 2023;104:37-42.
[19] Barat M, Jannot AS, Dohan A, Soyer P. How to report and compare quantitative variables in a radiology article. Diagn Interv Imaging 2022; 103:571-3.
[20] Strobl C, Malley J, Tutz G. An introduction to recursive partitioning: rationale, application, and characteristics of classification and regression trees, bagging, and random forests. Psychol Methods 2009; 14:323-48.
[21] Benchoufi M, Matzner-Lober E, Molinari N, Jannot AS, Soyer P. Interobserver agreement issues in radiology. Diagn Interv Imaging 2020;101:639-41.
[22] Savci G, Yazici Z, Sahin N, Akgoz S, Tuncel E. Value of chemical shift subtraction MRI in characterization of adrenal masses. AJR Am J Roentgenol 2006; 186:130-5.
[23] Halefoglu AM, Altun I, Disli C, Ulusay SM, Ozel BD, Basak M. A prospective study on the utility of diffusion-weighted and quantitative chemical-shift magnetic res- onance imaging in the distinction of adrenal adenomas and metastases. J Comput Assist Tomogr 2012;36:367-74.
[24] Dalavia CC, Goldman SM, Melo H, Kater CE, Szejnfeld J, Iared W, et al. The value of signal intensity on T1-weighted chemical shift magnetic resonance imaging com- bined with proton magnetic resonance spectroscopy for the diagnosis of adrenal adenomas. Radiol Bras 2020;53:86-94.
[25] Libe R, Pais A, Violon F, Guignat L, Bonnet F, Huillard O, et al. Positive correlation between 18F-FDG uptake and tumor-proliferating anti- gen Ki-67 expression in adrenocortical carcinomas. Clin Nucl Med 2023;48:381-6.
[26] Kataoka M, lima M, Miyake KK, Matsumoto Y. Multiparametric imaging of breast can- cer: an update of current applications. Diagn Interv Imaging 2022; 103:574-83.
[27] Cao Y, Wang X, Shi J, Zeng X, Du L, Li Q, Nickel D, Zhou X, Zhang J. Multiple param- eters from ultrafast dynamic contrast-enhanced magnetic resonance imaging to discriminate between benign and malignant breast lesions: comparison with apparent diffusion coefficient. Diagn Interv Imaging 2023; 104:275-83.
[28] Miller FH, Wang Y, McCarthy RJ, Yaghmai V, Merrick L, Larson A, et al. Utility of diffusion-weighted MRI in characterization of adrenal lesions. AJR Am J Roent- genol 2010;194:W179-85.
[29] Cicekçi M, Onur MR, Aydin AM, Gül Y, Ozkan Y, Akpolat N, et al. The role of appar- ent diffusion coefficient values in differentiation between adrenal masses. Clin Imaging 2014;38:148-53.