Multiparametric MRI in the Assessment of Cervical Carcinoma
Reason: The dataset contains all the patient information and their medical images, although they are anonymised. To ensure the orignial data is protected well, I shall apply for the permament embargo.
Supporting data for ‘'Multiparametric MRI in the Assessment of Cervical Carcinoma'’
This thesis explored and investigated several MRI techniques, and radiomics analysis based on multiparametric MRI in the clinicopathological assessment of cervical carcinoma (CC), specifically in discriminating between histological subtypes, tumour grades and FIGO stages.
1. Preliminary DKI study of CC (n=117) was conducted using 4 b-values of 0, 300, 1000 and 1500 s/mm2. The results demonstrated that both DKI parameters – MD and MK were able to differentiate histological subtypes. Moreover, MD was different between tumour grades, while MK was not.
2. Texture analysis based on multiparametric MRI including T2WI, ADC and T1c (n=95), and radiomics study based on T2WI and DKI were investigated in CC (n=117). Texture analysis showed excellent performance of SVM models using combined first-order features extracted from multiple MRI sequences in the clinicopathological characterisation of CC. Furthermore, radiomics analysis also exhibited excellent performance of RF models using combined radiomic features with transforms derived from both T2WI and DKI for distinguishing histological subtypes, however, DKI-only RF model demonstrated the highest performance in distinguishing FIGO stages.
3. As a novel quantitative MRI technique, MRF exhibited excellent scan-rescan repeatability of T1 and T2 values in normal cervix in our feasibility study (n=12). Moreover, T1 and T2 estimates of MRF demonstrated the potential value for tissue differentiation between CC and normal cervical tissues (n=28).