The unusual features of gastrointestinal stromal tumour (GIST) and advances in MRI for staging and treatment monitoring

Abstract

Gastrointestinal stromal tumor (GIST) is the most common mesenchymal tumour of the gastrointestinal tract. Our cases with unusual features were categorized into unusual tumour location, clinical manifestation, and site of metastasis. These features point to a poorer prognosis but are potentially treatable with imatinib.

Among them, we investigated the incidence of lymph node metastasis in our PET/CT database. 6 (20.7%) out of 29 were positive, surprisingly higher than previous studies. The patients had rare primary locations such as ovaries and esophagus. Results underlined the need for additional study and the potential for alternative surgical approach.

PET/CT has the drawbacks of high radiation dose, tedious procedures and low availability. We investigated if DW-MRI could be a good alternative.

5 patients with histological-proven GIST metastasis underwent DW-MRI and PET/CT at staging. 29 intra-abdominal lesions were found by both modalities (100% sensitivity). Significant inverse correlations were found among all five pairs of ADCs and SUVs using Pearson’s correlation, with the highest between ADCmean and SUVmean (r = -0.496, P = 0.006) and ADCmean and SUVmax (r = -0.471, P = 0.010).

For treatment monitoring of GISTs, 6 patients with histological-proven metastasis underwent DW-MRI and PET/CT at diagnosis and at 3-months follow-up. All 37 metastatic lesions were identifiable in MRI (100% sensitivity). Significant inverse correlations were found between ADCmean_thr and SUVmean/SUVmax at diagnosis and follow-up (P < 0.001); and ΔADCmean_thr and ΔSUVmean/ΔSUVmax (P < 0.001). Also, pre-treatment ADCmean_thr outperformed SUVmean and SUVmax in treatment response prediction with AUC of 0.706.

We concluded MRI was comparable to PET/CT in sensitivity and offering cellular information for staging, treatment monitoring and response prediction.

To improve the accuracy of volume and ADC measurements in DW-MRI, we proposed a method based on thresholding both the b0 images and ADC maps. 21 metastatic GIST lesions were manually contoured as gross tumour volume (GTV) and gross ADC (ADCg). Using kmeans clustering algorithm, the probable high-cellularity tumour tissues were selected as thresholded ADC (ADCthr) and high-cellularity tumor volume (HCTV). The metabolic tumour volume (MTV) in PET/CT was measured using 40% SUVmax as lower-threshold.

Results showed that HCTV had excellent concordance with MTV (r = 0.984, p <0.001). In contrast, GTV overestimated the volume and differed significantly from MTV (p = 0.005). ADCthr correlated strongly with SUVmean (r = - 0.807, p < 0.001) and SUVmax (r = - 0.843, p < 0.001).

We also applied this method to peritoneal metastases of other primary. In 18 homogeneous lesions, no significant difference is found between HCTVADC and MTV (p = 0.729) and they are also linearly correlated (r = 0.99, P < 0.01). Therefore, we concluded that the proposed method enables more accurate volume and ADC delineation of GISTs and other metastatic tumours in DW-MRI.

To conclude, the clinical manifestations of GISTs are better understood with advanced imaging. With new sequence and segmentation method available, MRI has the potential to provide cheaper and quality imaging needed for accurate staging, treatment monitoring and response prediction of GISTs.