Abstract
Background and Aims:Multiple myeloma is a bone marrow cancer, arising from plasma cells, with a poor prognosis.
Currently imaging diagnosis is made by detecting focal bone destruction (lytic lesions), a late finding. In Europe and the United States, imaging at initial presentation of patients with suspected multiple myeloma consists of whole body computed tomography (CT). In the United Kingdom, imaging with whole body magnetic resonance imaging (MRI) is preferred.
Earlier detection of disease with earlier initiation of treatment may improve clinical outcomes.
Whole-body quantitative biomarkers from imaging could provide an objective measure of bone marrow fat replacement and plasma cell burden in multiple myeloma patients. With CT, skeletal calcium-subtracted Hounsfield unit (HU) values can be quantified from dual energy CT acquisitions. HU are obtained from a linear transformation of measured attenuation coefficients. With MRI, skeletal fat signal can be quantified from T1-weighted gradient- recalled echo Dixon sequences as a fat fraction (FF). FF is defined as the signal arising from fat protons divided by the sum of the signals from fat and water protons.
This thesis aimed to address a clinical need to improve the objective evaluation of disease status in patients presenting with suspected multiple myeloma through the extraction of quantitative metrics from the skeletal marrow by employing semi-automatic/automatic pipelines.
Across three work packages, this thesis aimed to:
1) Extract skeletal calcium-subtracted HU from whole body dual energy CT and to assess how skeletal calcium-subtracted HU correlated with clinical markers of disease status in an exploratory sample of patients with suspected plasma cell disorders.
2) Extract skeletal FF derived from whole body MRI T1-weighted gradient-recalled echo Dixon images and to assess how FF correlated with clinical markers of disease status.
3) Explore how skeletal FF derived from whole body MRI T1-weighted gradient-recalled echo Dixon images might complement current methods of assessing response to therapy.
Methods:
Dual energy CT imaging & analysis
A non-contrast dual energy CT scan was obtained from 21 participants. The Chan-Vese algorithm was applied to extract a mask of the skeleton from the anatomical CT images. This mask was then applied to the scanner-generated parametric maps to obtain calcium- subtracted attenuation values of the skeleton. Correlation with bone marrow biopsy plasma cell infiltration and blood haemoglobin levels were undertaken using Spearman’s rank correlation test. In addition, calcium-subtracted attenuation values were obtained from region-of-interest analysis of focal lytic lesions and the marrow at L3 vertebra. These were compared to whole skeleton values using the Wilcoxon signed-rank test.
Skeletal segmentation and FF analysis from whole body MRI
Whole body T1-weighted gradient-recalled echo MR images were acquired from 45 participants. Whole body segmentation was performed using an automatic pipeline based on 2D nnU-Net with uncertainty estimation (UnnU-Net) to extract skeletal FF values from the scanner generated FF images. Wilcoxon signed-rank test was used to compare FF values obtained from the automatic segmentations and manual segmentations performed by radiologists, focusing on the whole skeleton and specific MR acquisition stations. This comparison was further conducted for two clinical diagnoses: monoclonal gammopathy of undetermined significance (MGUS, with 8 patients) and multiple myeloma (MM, with 11 patients) using Mann-Whitney U test. The association with bone marrow biopsy plasma cell infiltration was assessed using Spearman's rank correlation tests.
Exploratory assessment of change in FF in myeloma patients on treatment
Whole body T1-weighted gradient-recalled echo MR images were acquired before and after high-dose chemotherapy/autologous stem cell transplantation (ASCT) in 10 patients with multiple myeloma. Patients were categorised as biochemical responders or non-responders, based on serum paraprotein levels. The MY-RADS response criteria were also used to assess imaging response. Whole body segmentation was performed using an automatic pipeline based on 2D nnU-Net with uncertainty estimate (UnnU-Net) to extract skeletal FF values from the scanner generated FF images. Descriptive statistics were used for the comparison of pre- and post- therapy imaging and FF values.
Results:
Dual energy CT imaging
Skeletal calcium-subtracted attenuation values extracted from whole body dual energy CT images demonstrated a positive correlation (Spearman’s rho 0.79, p < 0.001) with the degree of marrow infiltration on bone marrow biopsy. This suggests that it holds potential as an objective measure of disease burden.
Whole skeleton segmentation and FF analysis from whole body MRI
The automatic segmentation performed by UnnU-Net demonstrated comparable results to manual segmentation in terms of FF quantification for each MR acquisition station and the whole skeleton in the training dataset. Moderately negative associations were found between the FF extracted from automatic segmentation and the infiltration percentage (skeleton: r = -0.52, p = 0.002; pelvic station: r = -0.43, p = 0.01). These findings lay the foundation for using FF as a skeletal MR imaging biomarker to assess changes in infiltrative myeloma disease burden during the course of disease and on treatment.
Assessment of change in FF in myeloma patients on treatment
Following therapy, the FF of the skeleton increased (13.53%) among biochemical responders, while it remained unchanged (0.42%) in biochemical non-responders. This establishes a foundation for the utilisation of FF in assessing treatment, complementing existing biochemical and imaging-based assessment of changes in focal lesions.
Conclusions:
Whole skeleton quantitative metrics extracted from whole body dual energy CT and whole body T1-weighted gradient-recalled echo MR images may provide complementary information to the current clinical methods of assessment. The findings in this thesis highlight the potential of advanced imaging techniques and quantitative analysis in providing a more objective measure of disease burden and treatment response. Further research and validation of these findings may lead to improvement in clinical assessment in the future.
| Date of Award | 1 Jun 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Vicky Goh (Supervisor), Isabel Dregely (Supervisor) & Michela Antonelli (Supervisor) |