Diffusion- and perfusion-weighted MRI in chest diseases

The overall aim of the thesis was to develop a conceptual approach for characterisation of thoracic lesions, by using dynamic contrast (DCE) MRI and diffusion-weighted magnetic resonance imaging (DWI), to diagnose and differentiate (benign versus malignancy) pleural, lung- and mediastinal lesions. For DCE MRI, a semi-quantitative method was used and for DWI, quantitative and visual methods were used.

Initially, a feasibility study was performed to validate the semi-quantitative DCE method regarding assessment of pleural lesions. 10 patients (40%) with suspected MPM were identified in a group of 25 patients, 15 patients (60%) showed benign lesions such as pleurisy, or state after talc pleurodesis. Using the combination of initial slope (IS) (above 6 / sec), and C-peak (above 1.5) values, sensitivity was 50%, specificity 100% and accuracy 80%, respectively.

For the assessment of pleural disease, we performed a prospective experiment where we quantitatively evaluated 31 patients with clinically suspected MPM lesions on computed tomography or PET. To this end, we obtained a diagnostic protocol. With DWI, the ADCmean of pleural injury was quantitatively determined. Values less than 1.52x10-³mm²/sec in the malignant zone and higher than 2.00x10-³ mm²/sec were considered to advocate benignity. Pleural lesions with a ADCmean value between 1.52 and 2.00x10-³ mm²/sec DCE MRI parameters should be calculated as in the feasibility study. With this method, sensitivity was 92.8%, specificity 94.1% and accuracy 93.5%. Given that this method is rather cumbersome in daily routine review, a visual method was developed. In a cohort of 100 patients with pleural abnormalities, three aspects were examined for their presence: thickened mediastinal pleura, pleural thickened with reduction of the hemithorax (‘shrinking lung’), and pleural foci, which exhibit a hyper-intensity transition with rising b-values ('pointillism sign'). This analysis showed that the ‘pointillism sign’ scored much better than the mediastinal pleural thickening or the well-known ‘shrinking lung sign’.In our study, the ‘pointillism sign’ had a sensitivity of 93%, a specificity of 79% and an accuracy of 88%.

For the evaluation of pulmonary nodules and pulmonary masses a third study was performed, which consisted of a first evaluation of 54 preoperative patients with a solitary lung lesion and then a validation of 54 other preoperative patients. From this, we concluded that a morphological DCE curve evaluation, followed by a DWI assessment of type B curve, was the best method. We found that lung lesions with type A curves were usually malignant, C and D curves benign and type B curves were complementary with DW-MRI assessed. The ADChigh threshold was 1.52 x 10-3 mm²/s. This methodology provides a sensitivity of 95%, specificity of 79% and accuracy of 91%.

In a fourth study we evaluated 100 consecutive preoperative mediastinal lesions. Because of the heterogeneity of these lesions, we had to add reviews of the lesions in b0, b1000 and b2000 values. This allowed the DW-MRI accuracy to rise from 64.5% to 80%. Accuracy with the aid of DCE-MRI was 83%, but coupled with a multi-parametric analysis (MPA), b-1000 value to the ADC-map and DCE-MRI curves increased accuracy up to 89%.

Finally, the findings from previous studies were applied in 34 patients (14 with MPM stage III, and 20 patients with SCLC-ED) for early assessment of chemotherapy treatment-induced changes with DWI and DCE in order to extract the parameters that are predictive for long-term PFS and/or OS. For instance, ADClow values were significant for predicting OS. This finding may be consistent with the mechanism of action of chemotherapeutic agents, which also exhibit a perfusion-related effect. As DWI was used as a biomarker during early treatment, the therapy could be optimized.