' Patients with locally advanced breast cancer are preferably managed with neoadjuvant chemotherapy (NAC) to decrease the tumour size prior to surgery and increase the possibility for breast conservation. A proportion (60-70%) of patients receiving such treatment will respond. The rest will have a worse outcome and should be identified as early as possible. The purpose of this study was therefore to evaluate the dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) characterization of locally advanced breast cancer prior to any treatment and correlate these characteristics to the clinical status (alive / deceased) of the patients 5 years after neoadjuvant chemotherapy.

This study included 24 women scheduled for neoadjuvant chemotherapy. Pixel by pixel analyses from DCE-MRI T1-weighted images were performed to evaluate changes in the distribution of contrast enhancement in the entire tumour volume. Time-intensity curves were then related to clinical outcome using both Neural Network and Support Vector Machine analyses from Tiberius.

Results from correlation of clinical outcome and therapy response (RECIST), and volume size change of more than 30% showed no distinct differences in the Kaplan-Meier analysis.

However, when focusing on the result from multivariate analyses of the DCE-MRI data before NAC, the prediction of survival is highly significant (p<0.001).

Thus, it appears that clinical outcome may be reasonably predicted from initial DCE-MRI measurements in conjunction with the appropriate multivariate analyses. '

' West Fraser operates Medium Density Fibreboard, (MDF), plants in Canada. A primary quality specific for MDF plants is a measure of the glue bond strength developed in the panels. The glue bond strength test is time consuming and only limited tests can be conducted during processing. The glue bond strength is predicated on the amount of glue added to the wood fibres processed; and the way in which the product is manufactured. The glue is the most expensive components in MDF manufacture, and is lowered if glue bond strength is above the specified target.

West Fraser invested in the development of a process data capture system which models the MDF manufacturing process and associates hundreds of process variables with each panel made; the software also associates these variables with the quality tests being conducted on panels. Tiberius has allowed the prediction of glue-bond strength from process variables and a quality test that takes only minutes to perform. Glue usage has been significantly lowered by presenting more quality information and opportunities to which operators can react. A future development involving Tiberius is to try and predict glue bond strength for every panel made in real-time using process data and on-line instrumentation. '

' Our company (Norsk Hydro ASA) has very successfully utilized Tiberius for curve reconstruction and modelling as part of our formation evaluation of oil and gas fields. '

' In banks there are a variety of products available such as homeloans, credit cards and every day accounts. Once a customer starts a bank relationship by taking up a product, there are opportuninities to grow this relationship by offering them other products they might desire.

Tiberius was used to identify the distinguishing characteristics of customers who were likely to respond to a mail out offering a credit card. This model was then used to select existing , non credit card holding customers for a marketing campaign. Not only did the model predict response rate, but also profitability and credit risk of the customers.

The models gave the marketing team prior knowledge of the expected response rates, thus enabling them to determine the size of the mail out in order to achieve the marketing objectives. Once deployed the models proved extremely accurate and robust over time. '

' The outcome of a typical sporting event is extremely difficult to assess with any degree of accuracy, which is indeed the reason it makes it an area ideal for betting purposes. To make consistent profits from betting therefore is not an easy task either, if it were everybody would be doing it.

However, thanks to modern machine learning software, making a decent income from sports predictions is not impossible. My success in the markets I play is reliant upon two crucial factors;

(a) Extracting and processing key information with a greater accuracy & consistency than both the bookmaker and my fellow punters,
(b) The ability to act quickly with late information (horse scratched from race, player injured, etc.)

Tiberius hits the spot on both counts. Ease of use, the speed it can resolve to a useful solution and its accuracy of projections are equal to anything else I've used, and far better than most. Of particular value to me is the code generator, allowing me to use the algorithms in a spreadsheet. Reassessing a late change in circumstances is now a simple process, update a cell and obtain the new prediction instantly.

My investments (I don't regard this as gambling) are now based entirely upon neural network analysis, and ten winning months from the last twelve will testify as to its effectiveness. Many suggest that to profit from betting needs a good slice of luck. But you know what, the more I rely upon machine intelligence for my selections the luckier I get! '

' Large volumes of oil, gas & coal are burned each year to fuel the economies of the world. Many facilities using these various energy inputs can be updated to be more efficient and less polluting.

We use Tiberius to model industrial processes to improve the efficiency of its energy usage whilst reducing the processes environmental impact . In other words, we can predict how each plant's energy efficiency changes during various operating scenarios and incorporate this into the daily plant operations. The less fuel that is burned means less pollution and less cost. But, that's not all, the same methodology can be used to predict the emissions profile and avoid operating in suboptimum areas . '