Malignant tumors exhibit restricted diffusion of water molecules on DW-MRI due to their compact cellular organisation and tortuous extracellular space. Preliminary single-centre studies have demonstrated the potential role of the apparent diffusion coefficient (ADC) derived from DW-MRI as a quantitative biomarker for assessing early treatment response in epithelial ovarian (EOC) and primary peritoneal cancer (PPC).¹ In addition, tumor site-specific ADC profiles have suggested biologic heterogeneity of disease.² Therefore, an understanding of the ADC changes following neoadjuvant chemotherapy at different metastatic sites in advanced EOC and PPC is essential to establish the utility of this biomarker in site-specific response assessment.To establish anatomic site-specific sensitivity and specificity of DW-MRI for detecting presence of viable tumor after neoadjuvant chemotherapy in advanced EOC/PPC; and to establish the use of ADC histograms for distinguishing presence of macroscopic and microscopic disease.

Study protocol: 16 patients with newly-diagnosed advanced EOC or PPC who were scheduled to receive neoadjuvant platinum-based chemotherapy followed by interval debulking surgery were recruited from two institutions as part of an ongoing prospective multi-centre clinical trial (DISCOVAR, NCRN portfolio number 11182) and gave their written consent to participate in this study. Each patient underwent an MRI prior to commencement of chemotherapy and post-treatment after 3-4 cycles of platinum-based chemotherapy, before interval debulking surgery.

Imaging protocol: Imaging was performed on a 1.5T MR scanner. Hyoscine butylbromide (20 mg) i.m. was administered to patients before scanning to reduce image artefacts due to peristalsis. Stacked DW sequences (free-breathing axial single-shot echo-planar, slice thickness 6mm, pixel size 3x3mm) using four b-values (0,100,500,900smm^-2) were acquired in three stations covering the abdomen and pelvis from the symphysis pubis to the top of the diaphragm. Slice-matched T₁-weighted and T₂-weighted anatomical imaging were acquired in the abdomen and pelvis.

Analysis: Up to ten of the largest lesions that were representative of involved anatomic sites, were chosen per patient on the baseline imaging and analysed using in-house software.³ ADC maps were generated through monoexponential fitting for b-values of 100,500and900 smm^-2. Regions-of interest were drawn by region growing around areas of impeded diffusion arising from solid tumor on consecutive computed high b-value axial images (b=1000smm^-2), generating a volume of interest (VOI). Cystic or necrotic areas were excluded by visual matching with the morphological MRI sequences. Histograms were derived from the individual pixel ADC values within the VOI on the baseline and post-treatment imaging (bin width 8.08 x10^-5 mm²s^-1). Lesions that did not exhibit impeded diffusion post-treatment were classified as no viable tumor and were excluded from the ADC analysis post-treatment. Lesions containing impeded diffusion post-treatment were classified as containing viable tumor, and the percentage post-treatment change in ADC from baseline was calculated for these lesions. The orientation and location of the chosen lesions was recorded at debulking surgery to enable good anatomical matching between the histopathological specimens and lesions on imaging. Presence or absence of viable tumor in the designated lesions was recorded at surgery or on histopathology. On histopathology, lesions containing viable tumor were further sub classified as containing microscopic foci or as macroscopic disease if the lesions contained more numerous and larger areas of viable tumor and the Histopathologist noted tumor on macroscopy. Sensitivity, specificity, accuracy, positive and negative predictive values (PPV, NPV) were calculated for DW-MRI detecting viable tumor by anatomic site. For lesions with residual viable tumor, histogram parameters were compared between residual microscopic or macroscopic disease.

68 anatomically matched lesions (30 peritoneal nodules, 16 omental samples, 17 ovarian masses, and 5 lymph nodes) were included. DW-MRI depicted 35 of 46 viable tumor sites, and correctly identified 16 of 22 sites uninvolved by viable tumor. DW-MRI failed to visualize 11 lesions with residual tumor, 6 peritoneal (3 with microscopic tumor) and 5 in the omentum (4 with microscopic tumor) and incorrectly identified tumor in 5 peritoneal lesions and 1 lymph node. Overall sensitivity=76.1%, specificity=72.7%, accuracy=75%, PPV=85.4%, NPV=59.3%; site-specific data in Table 1. Histogram parameters showed no significant differences in absolute values of ADC or change in ADC between lesions containing residual microscopic vs macroscopic disease (Table 2, Figure 1).The detection accuracy of DW-MRI of viable tumor varied between disease sites. A possible explanation for this is that peritoneal and omental disease are often ill-defined⁴, which makes their analysis more difficult than other sites. The ADC is a useful adjunct to morphological imaging for the detection of viable tumor, but larger numbers of lesions are needed to interrogate differences between microscopic and non-viable tumor and residual macroscopic tumor because of the heterogeneity of the ADC response.