An Enhanced Diagnostic Model for Incidental Hashimoto’s Thyroiditis Using T2-Weighted and Diffusion-Weighted Imaging

Introduction

Hashimoto’s thyroiditis (HT) is the most prevalent autoimmune disease affecting the thyroid gland and is also the leading cause of hypofunction. The prevalence of HT ranges from 5.6% to 14.2% [1], with discrepancies observed among regions with varying economic levels. The incidence of the disease in females is 4-5 times that in males [2]. Patients typically present with a significant increase in thyroid peroxidase antibodies (TPO-Ab) or thyroglobulin antibodies (Tg-Ab), along with normal or elevated levels of thyroid-stimulating hormone (TSH) [3]. Treatment primarily involves managing hypothyroidism through substitution therapy with levothyroxine [4]. Studies have shown, however, that TPO-Ab are associated with adverse pregnancy outcomes [5] and lower intelligence quotient in children [6]. Consequently, the use of supplements to reduce TPO-Ab or Tg-Ab levels, particularly the application of appropriate amounts of selenium, is also being considered by clinicians for treatment of HT.

Although pathology is the criterion standard for diagnosing HT, fine-needle aspiration cytology is not recommended for diagnosing HT if there are no thyroid nodules [7]. Ultrasound is a widely used imaging technique for HT, revealing a gland that is diffusely heterogeneous, with patchy, nodular hypoechoic areas interspersed with echogenic parenchymal bands, which impart a micronodular appearance to the gland. However, the sonographic imaging features of various causes of diffuse thyroid disease often overlap [8], and ultrasound lacks the ability to quantitatively assess the severity of thyroid damage, making it highly operator-dependent.

Currently, diffusion-weighted imaging (DWI) is the sole method capable of measuring water molecule diffusion movement in vivo. This technique employs multi-b-value scanning and single-exponential model fitting to ADC maps, along with double-exponential model fitting to intravoxel incoherent motion (IVIM) for obtaining multiple quantitative parameters D, D*, and f, in which D reflects genuine diffusion information, while D* signifies perfusion information and f represents blood flow, thus enabling reflection on both tissue diffusion and microcirculation perfusion. HT involves pathological changes such as lymphocyte infiltration, increased fibrosis, and augmented blood supply. IVIM, which theoretically can reflect the pathological state of HT, is commonly used in diagnosing or evaluating treatment efficacy in liver [9], prostate [10], breast [11], and other tumor lesions [12]. However, it has been rarely utilized for diagnosing thyroid nodule lesions or diffuse thyroid diseases. Our previous studies have shown that HT has higher signal intensity on T2WI than for a normal thyroid gland. The T2WI grayscale ratio has utility in diagnosis of HT [13], but as a diagnostic tool, it should have higher sensitivity and specificity. Therefore, research is needed to determine whether combining T2WI grayscale ratio with multi-b-value DWI can improve the diagnostic utility of incidental HT.

Material and Methods

ETHICS APPROVAL:

The study was retrospective, and the researchers submitted a report requesting a waiver of informed consent, which was approved by the Ethics Committee of the Friendship Hospital of the Ili Kazak Autonomous Prefecture (Approval No.RMB2024-62).

DATA COLLECTION:

A retrospective study was performed on 510 individuals who received physical examinations at Friendship Hospital of Ili Kazak Autonomous Prefecture between 2022 and 2023. The criteria for inclusion in the study on HT were as follows: (1) cervical spine magnetic resonance (MR) examination was conducted; (2) ultrasound revealed diffuse thyroid echogenicity, characterized by a gland that was diffusely heterogeneous with patchy, nodular hypoechoic areas interspersed with echogenic parenchymal bands, imparting a micronodular appearance to the gland; and (3) serum TPO-Ab and/or Tg-Ab levels were elevated (>34IU/ml), along with normal or elevated levels of TSH. The exclusion criteria were as follows: a history of prior neck radiation therapy, and MR images compromised by various artifacts that hindered the observation and measurement of the thyroid gland. For comparison purposes, individuals not diagnosed with Hashimoto’s thyroiditis (non-HT) were selected as controls, with the following criteria: (1) a normal thyroid ultrasound; (2) serum thyroid function tests for TPO-Ab and Tg-Ab were normal or greater than 34 IU/ml and less than 100 IU/ml; and (3) TSH, total triiodothyronine (TT3), total thyroxine (TT4), free triiodothyronine (FT3), and free thyroxine (FT4) levels were normal. These controls were artificially matched with the HT group based on age and sex, and were selected in a 1: 1 ratio during the same period. If a case corresponded to multiple controls, the random number method was used to randomly select controls. The experimental procedures of the study are shown in Figure 1.

MAGNETIC RESONANCE IMAGING:

We utilized a 3.0T MRI scanner (HDxt, General Electric, Milwaukee, Wisconsin, USA), equipped with an 8-channel head and neck coil, and ensured that all metal objects were removed from the patient’s body. The patient was positioned supinely, with the head placed first and immobilized. A rice-filled cotton bag, measuring roughly 30 cm×15 cm and weighing approximately 1 kg, was placed on the front of the neck to reduce magnetic susceptibility artifacts caused by the significant tissue difference between the neck soft tissues and air [14].

T2-WEIGHTED ITERATIVE DECOMPOSITION OF WATER AND FAT WITH ECHO ASYMMETRY AND LEAST SQUARE ESTIMATION (IDEAL-T2W) WATER IMAGING PROTOCOL AND IMAGE POST-PROCESSING:

IDEAL-T2W: In the axial position, the slice thickness was 4 mm with no interslice gap. The repetition time (TR) was 3140 ms, and the echo time (TE) was 86 ms. The matrix size was 512×512, and the field of view (FOV) was 18×18cm. The scan duration was about 2 minutes.

Image post-processing: We identified the largest cross-section of the thyroid gland, avoiding blood vessels and the gland’s edges. An irregular shape measurement tool within the PACS system was used to delineate the entire thyroid gland at the chosen level. We recorded the grayscale value of the gland, and similarly measured the grayscale values of the sternocleidomastoid muscle, trachea, and subcutaneous fat at the midline of the back at the same level, as illustrated in Figure 2. The grayscale ratios of thyroid-to-muscle (T/M), thyroid-to-trachea (T/Tr), and thyroid-to-lipid (T/L) on T2-weighted images (T2WI) were computed. Two senior radiologists with more than 7 years of experience in head and neck imaging independently performed image measurements, and the final result was the average of the 2 measurements.

MULTI-B-VALUE DWI SCANNING PROTOCOLS AND IMAGE POST-PROCESSING: Thyroid axial scanning utilized a single-excitation spin-echo diffusion-weighted plane wave imaging sequence with 10 b-values (b=0, 10, 20, 50, 100, 200, 400, 600, 800, 1000 s/mm2). This included 6 low b-values (≤200 s/mm2) for observing microcirculatory perfusion and 4 high b-values (>200 s/mm2) for observing diffusion. The scan duration was 2 minutes 45 seconds, with a slice thickness of 4 mm and no interslice gap. The TR was 5150 ms, and the TE was 74.9 ms. The matrix size was 256×256, and the FOV was 18×18 cm. The number of excitations (NEX) was 2 for the first and last b-values, and 1 for the intermediate values. Quantitative analysis was performed using the Functool software package of the GE AW4.5 workstation’s MADC tool, with the b-threshold set to 200s/mm2. Upon clicking on COMPUTER, the software automatically generated a single-exponential model DWI’s ADC map, a bi-exponential model IVIM’s D map, D* map, and f map. In the original image at b=10s/mm2, while avoiding blood vessels, the maximum region of interest (ROI) encompassing the entire thyroid gland was drawn along the contour of the thyroid gland at the upper and middle levels of the left and right lobes [15], as illustrated in Figure 3. The average value was subsequently computed, yielding the mean measurement results from 2 physicians for the left and right lobes’ ADC (×10−3 mm2/s), D (×10−3 mm2/s), D* (×10−2 mm2/s), and f values, as depicted in Figure 4.

STATISTICAL ANALYSIS:

The statistical analysis was conducted using SPSS version 29.0 (IBM Corporation, Armonk, NY, USA). A normality test was applied to quantitative data, with the paired samples adhering to a normal distribution being analyzed via a two-sample t-test, and those with a non-normal distribution evaluated using a two-sample rank sum test. Data were expressed as median (interquartile range), and statistical significance was set at P<0.05. An intraclass correlation coefficient (ICC) greater than 0.5 indicated good inter-observer measurement consistency. The ratios of T/M, T/Tr, T/L, as well as ADC, D, D*, and f, were subjected to univariate logistic regression analysis to identify risk factors for HT. Subsequently, a binary logistic regression analysis was conducted using the stepwise approach, selecting variables with P-values less than 0.05 to calculate the odds ratio (OR) and construct a comprehensive diagnostic model. Receiver operating characteristic (ROC) curves were produced using the R language version 4.4.3 (2024-02-29 ucrt). Spearman correlation analysis was employed to evaluate the correlation between each parameter and the levels of TPO-Ab and Tg-Ab. A correlation coefficient of 0.8 or greater signifies a high correlation, while a coefficient between 0.5 and 0.8 indicates a moderate correlation. A coefficient between 0.3 and 0.5 suggests a low correlation, and a coefficient less than 0.3 indicates a weak correlation.

Results

Clinical Data

CLINICAL DATA:

In conclusion, 25 pairs of HT patients and non-HT patients were selected, aligning in sex and age, with a maximum age discrepancy of 5 years – 20 pairs of females and 5 pairs of males. Within the HT group, 20 individuals exhibited normal serum thyroid function test results, meaning TSH, TT3, TT4, FT3, and FT4 levels were normal. Four presented with subclinical hypothyroidism, characterized by an elevated TSH level (maximum 9.53 mIU/L) alongside normal TT3, TT4, FT3, and FT4 levels. One individual had clinical hypothyroidism, which was indicated by elevated TSH and reduced FT3 and FT4 levels. The cohort consisted of 40 females and 10 males, aged 14–54 years, with a mean age of 34.42±8.77 years and a median age of 36 years.

A RANK SUM TEST FOR THE COMPARISON OF T2WI GRAYSCALE RATIO AND MULTI-B-VALUE DWI PARAMETERS BETWEEN HT AND NON-HT MATCHED SAMPLES: Table 1 illustrates the disparities in T2WI grayscale ratios and various parameters of multi-b-value DWI between HT and non-HT subjects. The median T2WI grayscale ratios of T/M, T/Tr, and T/L in the T2W-IDEAL water image were higher in HT subjects compared to non-HT subjects, exhibiting a statistically significant difference (P<0.05). The inter-rater consistency for these measurements ranges from 0.77 to 0.88. Conversely, the median values of ADC, D, and f were lower in HT subjects than in non-HT subjects, also with a statistically significant difference (P<0.05). However, no statistically significant difference was noted in the parameter D* between HT and non-HT groups (P>0.05).

THE DIAGNOSTIC EFFICACY OF T2WI GRAYSCALE RATIO AND MULTI-B-VALUE DWI IN THE DIAGNOSIS OF HT: Figure 5 illustrates the diagnostic efficacy of T2WI grayscale ratio and multi-b-value DWI parameters specifically for the identification of HT. The area under the curve (AUC) values for T/M, T/Tr, T/L, ADC, D, D*, and f were 0.75, 0.79, 0.83, 0.90, 0.79, 0.62, and 0.71, respectively, all with a significance level of P<0.05. Table 2 presents the optimal cut-off values corresponding to the maximum Jodds index, which are 1.94, 8.50, 1.35, 1.25, 0.79, and 0.34. The AUC for D* in diagnosing HT was 0.62 (P>0.05), while the combined AUC for T/M+T/Tr+ADC in the same diagnostic context was 0.94, accompanied by a sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of 0.92. Binary logistic regression analysis ultimately included 3 variables in the model: T/M, T/Tr, and ADC. Table 3 displays the OR of parameters that are statistically significant, with the OR for T/M being the highest. Table 4 shows the correlation between T/M, T/Tr, T/L, ADC, D, D*, f, and serum titers of TPO-Ab and Tg-Ab. Notably, the correlation between TPO-Ab and ADC was the highest, with a correlation coefficient of 0.70, followed by the correlation between TPO-Ab and D, which had a correlation coefficient of 0.63.

Discussion

LIMITATIONS:

The limitations of this study are as follows: (1) The cases were not been confirmed by pathology; (2) The number of cases was relatively small, and they were not divided into groups according to normal thyroid function and hypothyroidism, nor were cases of transient hyperthyroidism in HT included; (3) Furthermore, as a retrospective analysis, there are some instances of mixed bias and selection bias, which necessitate further verification through large-sample, multi-center studies.

Conclusions

The incidental discovery of diffuse thyroid lesions during cervical spine MRI can be quantitatively and objectively distinguished as HT or non-HT to a certain degree by utilizing a combination of T2WI-IDEAL, ADC, and IVIM techniques,which may serve as a useful combined tool for diagnosing incidental HT.