 Welch chi'n gweithio, wrth gwrs, yn gweithio i'r Ffainfyrdd Gweithgwladol yn y 3rd Llyfrgell Cyniggymen mewn unifesidol, 3 munud Cymysgol. Felly, rydyn ni'n ceisio'r Llyfrgell Cymysgol, ac rydyn ni'n ceisio'r host i'r unifesidol. Ym 3 munud Cymysgol, rydyn ni'n ceisio'r unifesidol i'r unifesidol yng Nghymru, inni'n 2008. Mae'r gweithio am ychydig yng ngwybod i gyd yn ymdusio'r pwysig sydd yn hynny'n eu rhan o'r cyfnodau sy'n gyffredinol yn gweithio'r cyllid. Ond iddyn nhw'n mynd i'ch cynnig o'r wneud o'r unedig, yn ystod o'i gweithio'n ymddangos o'r unedig sydd yn ydych chi'n bwysig o'r unedig o'r unedig o'r unedig sydd yn yw'r unedig o'r mewn cyffredinol. Mae'r ffrifesfyn o'r funud y byddai'n cymhwysig ymddangos. maen nhw'n treidwyd yn cael ei dweud o hyn sydd wedi'u wneud i ddweud o ffordd o rhan o'r theoses yn ddweud o'r byd. Wrth gwrs, mae'n ddweud i'r byw'r cyflwyno o'r ddweud o'r dyma, ac yn ganodol i'r ddweud o'r ddweud o'r ddweud o ddweud o'r ddweud o'r ddweud, yn ddweud o'r ddweud o ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud. Over the last three years we've had more than 300 participants from every college in the university, and I'm grateful to everyone who takes on the challenge, and to everyone who helps make the event possible. My colleagues Erica Linkie and David Shearer have been a tremendous assistance this year. We've had faculty, students, student leaders, trustees and others assisting judging the ten heats that ran over the past couple of months. And I acknowledge their contribution to the three-minute thesis event. We have a tremendous judging panel today. They are sitting in the front row. President Suresh, Vice Provost Burkart, Dean Dan Martin and Malloy, one of our trustees and last year's winner, Annie Arnold, will represent the judging panel today. The rules are straightforward. Speakers are limited to a maximum of three minutes. If they are still talking at the end of three minutes, they will be disqualified. Erica will be waving a laptop around so they can see a countdown timer. If they are still talking when the bell goes, we've got a problem. If they choose, and I believe every student today has chosen, they may display a single static PowerPoint slide to aid their presentation, but no other media or props, no fancy dress, dancing, singing or anything of that sort. In arriving at their decisions, the judges will look at three criteria, communication style, comprehension and engagement. At the end of today's events, we will decide the winners of the first three places and they will be awarded travel grant prizes as shown in the screen. But unlike the heats today, we have another opportunity and that is for you, the audience, to vote for your choice of the winner. Hopefully you've all picked up a ballot if not we can have some distributed. When the judges retire to make their decisions, that's the time for you to fill out your ballot, we'll collect them, count them and announce the winner of the people's choice at the end of the afternoon. So, with my three minutes done, I will turn now to our first presenter, Shinjin Ikundu, whose presentation title is predicting future osteoarthritis using MRI, the untold study of cartilage. Welcome. One, two, three, can everyone hear me? Okay, we diagnose diseases at a fairly advanced stage, but the hidden pathologic processes begin long before we can proceed. Let me give you an example. Consider osteoarthritis, the disease in which the thin cushion between two bones called cartilage falls apart, causing bone to grind painfully on bones. Only when bone damage and pain have developed, do we begin to see a planet building on X-ray. But if we could understand the invisible process occurring in the cartilage, we may be able to detect a disease early, we may be able to modify its trajectory, and we may be able to understand the process that leads to osteoarthritis in the first place. The goal of this work was to train a computer algorithm to be able to find the hidden changes and what we found was unexpected and beautiful. We collaborated with scientists at the NIH to obtain MRIs of knee cartilage, of healthy subjects, some of which are shown here. These people go on to develop clinical osteoarthritis in three years, but these people don't. But our eyes can't detect a feature that is common for those who go on to develop osteoarthritis. Humans can't detect the abnormal pattern because it's hidden underneath the vast normal variation that's coming up. Unfortunately, computers also have difficulty with this task. Comparing images pixel by pixel misses the spatial pattern, and in a sense it misses the forest for the trees. In my PhD thesis, I've developed a better way to compare similarities between images, between spatial images, between spatial images, between spatial images, between spatial images. My method is called distribution of pixels by measuring the amount of work, mass times distance. It takes to rearrange the pixels in one image to look like another. My method is called transport-based morphometry, or TBM, and it is not a heuristic, but a mathematically elegant solution, because in addition to classification, it enables us to invert the classification rule to visualize the pathologic changes that differentiate disease states from normal. And with this new technology, I'm able to train a computer to examine the cartilage of normal, healthy people and predict with 86% accuracy whether that person will go on to develop osteoarthritis symptoms three years down the line. These are state-of-the-art results, but these are also preliminary results. And with more and more scans, my algorithm will only get better and better. Visualizing the pathologic changes, we already see extensive subclinical damage in those who go on to develop osteoarthritis. I think the most exciting part of my method is that it can be applied to a variety of diseases. So, the CBO is just the beginning.