Introduction

Advanced cervical cancer is typically treated with (chemo)radiotherapy¹. Diagnostics and surveillance include whole-body [¹⁸F]fluorodeoxyglucose-(FDG)-PET/CT and pelvic MRI for metabolic and morphologic assessment, respectively, with emphasis on late surveillance FDG-PET for identification of residual avid tumor. (Chemo)radiotherapy can induce post-treatment edema, leading to false positive imaging findings. To allow for acute inflammation to subside, surveillance is conducted at earliest 3 months post-therapy². Techniques insensitive to post-treatment edema could allow for earlier surveillance with timely identification of poor-responders. Restriction spectrum imaging (RSI) enables separation of persistent malignancy from post-treatment edema, by compartmentalizing the diffusion signal into distinct water pools using a tissue-specific model. RSI has demonstrated clinical value in brain, prostate and breast^3-5, and promising preliminary results in cervical cancer⁶.  

Aim

To evaluate a cervical cancer-specific RSI model for assessment on an independent cohort, using combined PET/MRI data.

Methods

Four patients with primary cervical cancer (stage>IB2) underwent simultaneous FDG-PET/MRI (3T Signa PET/MR, GE Healthcare, Waukescha, USA) before (chemo)radiotherapy. Two of them additionally underwent surveillance with PET/MRI early (≈4 weeks) post-therapy, MRI 3 months post-therapy and PET/MRI 6 months post-therapy. Whole-body PET/MRI included routine static PET simultaneously with Dixon-MRI. Subsequently, pelvic MRI with conventional (clinical) and research sequences (e.g., multi-shell DWI for RSI) was conducted (scan parameters for relevant pelvic sequences in Table 1). Two healthy volunteers underwent pelvic MRI only. Matlab (R2020a) was used to fit the RSI model, updated from a preliminary version⁷, to the multi-shell DWI for estimation of the water pool signal contributions (C₁–C₄):   

ADCs were fixed in each compartment: D₁=0, D₂=0.0012, D₃=0.0035, D₄=0.5485 mm²/s. C₁ typically represents restricted diffusion, pronounced in malignancy. C₂ corresponds to hindered and restricted/hindered diffusion. C₃–C₄ represent even faster components (i.e., free diffusion, flow). The PET-derived standardized uptake value (SUV) maps were used for comparison of tumor conspicuity with RSI. Conspicuity was approximated by tumor contrast:⁸   

The S_T  and S_REF correspond to mean signal intensity within a single-slice region-of-interest in tumor and myometrium (reference tissue due to lack of healthy cervix in patients), respectively. Representative slices of PET and MRI were chosen. Differences were evaluated using Student’s t-test with Bonferroni correction (Matlab R2022b). As benchmark, the contrast between healthy cervix and myometrium was assessed for the volunteers. The radiologist surveillance report specified metabolic and morphologic tumor regression based on PET and conventional MRI, respectively. Inconclusive findings from conventional PET/MRI at 6-months surveillance were supplemented by histological examination from post-treatment incisional biopsy.

Results

Pre-treatment: Tumor, but not healthy cervix, was enhanced in the C₁, C₂ and C₁C₂ maps (Fig. 1).  C₂ also demonstrated pronounced signal in certain surrounding tissues (e.g., the intestines). C₃–C₄ maps showed low tumor signal and were therefore omitted in subsequent analysis. C₁ and C₁C₂ demonstrated superior tumor conspicuity to SUV in the quantitative analysis (Fig. 2). The visual correspondence between SUV and RSI is shown in Fig. 3.

Post-treatment: Pre- and post-therapy images of Patient I–II are shown in Fig. 4 alongside the radiologist report.

Patient I: Early and 3-months surveillance showed no residual tumor on conventional MRI and the C₁ and C₁C₂ maps, but persistent C₂ signal and partial metabolic regression on PET. At 6 months, inconsistent metabolic and morphologic findings were determined as post-treatment reaction by histological analysis. The C₁, C₂ and C₁C₂ maps showed local cervical signal.

Patient II: Early and 3-months surveillance showed partial metabolic and complete morphologic regression, whereas no RSI signal on C₁ and C₁C₂. At 6 months, MRI demonstrated complete imaging response concordant with complete metabolic regression assessed by PET.  

Discussion and Conclusion

This pilot study shows the potential of RSI for evaluation of cervical cancer. Preliminary results indicated improved pre-therapy tumor conspicuity using the C₁ and C₁C₂ maps, compared to conventional PET/MRI. At post-treatment surveillance, the C₁ and C₁C₂ maps agreed with conventional PET/MRI for the patient demonstrating complete metabolic and morphologic tumor regression at 6 months. Additionally, they showed no signs of tumor early post-therapy despite residual signal on PET. The patient with complete morphologic regression but suspicious metabolism at 6-months surveillance (assessed as post-treatment reaction) also showed signal on RSI. This might either reflect residual minimal malignancy or be artifact-related, which warrants further investigation. Limitations of this study include the simplified statistics for assessing tumor conspicuity based on minimal data. Future work will focus on larger-scale evaluation of RSI in differentiating malignancy from post-treatment reactions early after (chemo)radiotherapy. The potential synergistic value of RSI and PET in combined PET/MRI, with respect to locoregional, lymph node and metastatic assessment, will also be investigated.