Search

Summary mean DCE-MRI variables show a clear dependency between signal and noise variance, which can be shown to reduce the effectiveness of difference assessments. Appropriate transformation of these variables supports statistically efficient and robust comparisons. The capabilities of DCE-MRI based descriptions of hepatic colorectal tumour classification was assessed, with regard to their potential for use as imaging biomarkers. Four DCE-MRI parameters were extracted from 102 selected tumour regions. A multi-dimensional statistical distance metric was assessed for the challenging task of comparing intra- and inter- subject tumour differences. Statistical errors were estimated using bootstrap resampling. The potential for tumour classification was assessed via Monte Carlo simulation. Transformation of the variables and fusion into a single chi-squared statistic shows that inter subject variation in hepatic tumours is measurable and significantly greater than intra-subject variation at the group level. However, reliability analysis shows that, at current noise levels, individual tumour assessment is not possible. Appropriate data transforms for DCE-MRI derived parameters produce an improvement in statistical sensitivity compared to conventional approaches. Reliability analysis shows, that even with data transformation, DCI-MRI variables do not currently facilitate good tumour discrimination and a doubling of SNR is needed to support non-trivial levels of classification, Comment: 18 pages and 6 figures

Imaging demonstrates that preclinical and human tumors are heterogeneous, i.e. a single tumor can exhibit multiple regions that behave differently during both normal development and also in response to treatment. The large variations observed in control group tumors can obscure detection of significant therapeutic effects due to the ambiguity in attributing causes of change. This can hinder development of effective therapies due to limitations in experimental design, rather than due to therapeutic failure. An improved method to model biological variation and heterogeneity in imaging signals is described. Specifically, Linear Poisson modelling (LPM) evaluates changes in apparent diffusion co-efficient (ADC) before and 72 hours after radiotherapy, in two xenograft models of colorectal cancer. The statistical significance of measured changes are compared to those attainable using a conventional t-test analysis on basic ADC distribution parameters. When LPMs were applied to treated tumors, the LPMs detected highly significant changes. The analyses were significant for all tumors, equating to a gain in power of 4 fold (i.e. equivelent to having a sample size 16 times larger), compared with the conventional approach. In contrast, highly significant changes are only detected at a cohort level using t-tests, restricting their potential use within personalised medicine and increasing the number of animals required during testing. Furthermore, LPM enabled the relative volumes of responding and non-responding tissue to be estimated for each xenograft model. Leave-one-out analysis of the treated xenografts provided quality control and identified potential outliers, raising confidence in LPM data at clinically relevant sample sizes.

Citizen science has become a popular tool for preliminary data processing tasks, such as identifying and counting Lunar impact craters in modern high-resolution imagery. However, use of such data requires that citizen science products are understandable and reliable. Contamination and missing data can reduce the usefulness of datasets so it is important that such effects are quantified. This paper presents a method, based upon a newly developed quantitative pattern recognition system (Linear Poisson Models) for estimating levels of contamination within MoonZoo citizen science crater data. Evidence will show that it is possible to remove the effects of contamination, with reference to some agreed upon ground truth, resulting in estimated crater counts which are highly repeatable. However, it will also be shown that correcting for missing data is currently more difficult to achieve. The techniques are tested on MoonZoo citizen science crater annotations from the Apollo 17 site and also undergraduate and expert results from the same region.