We performed a prospective preclinical study, in which treatment response of lung cancer-bearing mice assessed by clinical CT and clinical MRI was compared to the results of micro-CT measurements.
Mouse model
Thirty-one mice with lung tumours were included in the study. Tumour development was driven by the expression of oncogenic Kras and the deletion of Tp53 and Ercc1, as described in detail by Jokic et al. [11]. Tumour-bearing mice were divided into three arms: mice receiving three cycles of cisplatin therapy (intraperitoneal injections of 7.5 mg/kg of body weight equating 7.5 mL/kg of body weight of cisplatin diluted in phosphate-buffered saline (PBS), one injection per week in three consecutive weeks) (n = 10), mice receiving three cycles of sham treatment (intraperitoneal injections of 7.5 mL/kg of body weight of PBS, one injection per week in three consecutive weeks) (n = 12), and mice which did not receive any therapy at all (n = 9). All mice were examined via micro-CT, clinical CT, and clinical MRI 2 days before and 24 days after treatment initiation. All animal experiments were performed in accordance with the national and European regulations and were approved by the local authorities (Landesamt für Natur, Umwelt und Verbraucherschutz NRW, reference number NRW 84-02.04.2013.A136).
Imaging
All mice underwent inhalation anaesthesia with isoflurane (2.0 to 2.5%) in air during image acquisition. Clinical MRI, clinical CT, and micro-CT scans were performed 2 days before and 24 days after treatment initiation, respectively. In order to keep the imaging protocols simple and to minimise stress and health risks for the mice, we decided to forego application of intravenous contrast media and respiratory gating.
Micro-CT
The LaTheta LCT-100A micro-CT-scanner (Aloka Co., Tokyo, Japan) was used with the following parameters: 50 kVp x-ray source with a focal spot size of 50 μm, 1 mA current, 4,340 ms exposure time per projection, 480 × 480 matrix covering a 49.48 × 49.48 mm2 field of view, and 0.3 mm voxel spacing in the z-axis, resulting in a voxel size of 0.1 × 0.1 × 0.3 mm3. Phantom measurements conducted by Stiller et al. [12] in the LaTheta LCT-100A revealed a radiation exposure of about 5 mGy in scans consisting of 58 slices. Micro-CT examinations performed in our study consisted of 40 to 60 slices, depending on the murine lung size.
Clinical CT
Imaging was performed on a Brilliance iCT 256-slice scanner (Philips Healthcare, Best, The Netherlands) in prone position using the following parameters: tube voltage 120 kV, tube current 100 mAs, slice thickness 0.67 mm, matrix 1,024 × 1,024, field of view as small as possible to include the murine thorax, rotation time 0.4 s, and pitch 0.4. Volume CT dose index for one slice was 13.6 mGy (calculated using a 16-cm human head phantom). The scan length fluctuated between approximately 1.5 and 2 cm, depending on the murine lung size, resulting in a dose length product of approximately 20.4 to 27.2 mGy × cm. Images were reconstructed using a hard reconstruction kernel.
Clinical MRI
Scans were acquired on an Ingenia 3.0-T system (Philips Healthcare, Best, The Netherlands) combined with a commercially available small animal coil (Philips Research, Hamburg, Germany) with heating function to preserve body temperature during the examination. The protocol consisted of one axial multi-shot T2-weighted turbo-spin-echo (TSE) sequence with the following parameters: echo time 65 ms, repetition time 1,540 ms, flip angle 90°, field of view 40 × 40 mm, matrix 256 × 256, slice thickness 1 mm without gap, acquired voxel size 0.19 × 0.27 × 1 mm, and reconstructed voxel size 0.15 × 0.15 × 1 mm, number of signal acquisitions 6. Standard scan time for 15 slices was 5:05 min. The detailed scan protocol of the T2-weighted sequence is available at the supplement for all vendors, as well as an ExamCard for Philips 3.0T Achieva and Ingenia MRI systems ready to use.
Image analysis
Total tumour burden was quantified by semiautomated, threshold-based volumetry using the postprocessing software IntelliSpace Discovery Imalytics Fundamentals (Philips Healthcare, Best, The Netherlands). At first, mouse lungs were manually defined excluding the mediastinal structures. In a second step, semiautomated tumour segmentation was performed using a threshold-based algorithm. For definition of the thresholds, two radiologists with 13-year (T.P.) and 2-year (J.E.S.) experience in rodent imaging visually evaluated the segmentation results of different threshold settings. They found the optimal threshold values between ventilated lung and soft tissue: −200 HU on micro-CT and clinical CT scans and 200 arbitrary units on MRI. Ventilated lung tissue showed lower and tumours showed higher HU and arbitrary units than the threshold, respectively. Figure 1 gives an overview of the segmentation process.
On micro-CT and clinical CT images, tumours show higher density values than ventilated lung tissue but similar density to intrapulmonary vessels. Hence, intrapulmonary vessels were included in the tumour volume on micro-CT and clinical CT datasets. In contrast, on clinical T2-weighted MRI scans, ventilated lung tissue and vessels appear hypointense due to low proton density and flow void effects, respectively, whereas tumours show high signal intensity. Non-malignant pulmonary processes, such as oedema, atelectasis, and pneumonia, were included in the tumour volume on micro-CT and clinical CT as well as on MRI scans.
Statistical analysis
Statistical analysis was performed using SPSS (IBM Corporation, Armonk, NY, USA) version 25. Normal distribution of data was assessed by the Shapiro-Wilk test. To assess differences of tumour volumes before and after treatment as measured by the three modalities, paired t tests or Wilcoxon’s signed-rank tests were performed depending on whether data was normally distributed or not. We used micro-CT as a reference standard for assessing tumour size and created the Bland-Altman plots comparing tumour volumes measured by CT and MRI, respectively, to compare the different measurements [13]. In order to evaluate the mean relative difference of tumour volumes measured by both clinical scanners and micro-CT, respectively, pre- and posttreatment data were combined. To evaluate the change of lung tumour burden, the ratio of posttreatment tumour volume/pretreatment tumour volume was calculated for clinical CT, MRI, and micro-CT.
Receiver operating characteristic (ROC) analyses were performed in a step-wise manner. First, we examined how well changes of tumour volume measured by all three modalities were able to discriminate the treatment group (cisplatin) from the non-treatment groups (PBS and no treatment, respectively). Mice with sham and without treatment were combined in order to increase the number of individuals in this group and thereby make the analysis more robust and clear. Second, we measured how well clinical CT and MRI could differentiate between treatment responders and non-responders as determined by micro-CT volume reductions of 10%, 15%, 20%, and 30%. Relative tumour sizes (i.e, posttreatment volume/pretreatment volume) determined by CT and MRI, respectively, were used as test variables. To investigate the correlation of tumour volume changes between all three modalities, Spearman’s r was calculated. All tests were two-sided. The alpha level was at 0.05.