Researchers at the Australian National University have developed a tool using artificial intelligence that can classify a brain tumour in an hour. Current technology takes weeks while the method developed by researchers at ANU utilises deep learning and demographic modelling to classify tumour types.
Category
📺
TVTranscript
00:00 Identifying specific subtypes of brain tumor is very important.
00:07 That information will help doctor to select the best treatment for individual patient.
00:14 So practically, we call that as personalized cancer medicine,
00:19 meaning that we select the best treatment for individual patient.
00:27 Basically to do that, the gold standard method is based on DNA methylation profiling.
00:34 However, there are some limitations of this method.
00:41 First, this method is only available in some certain laboratory or hospital.
00:48 Second, this method is very time consuming.
00:54 Usually it takes several weeks to obtain a result.
00:59 However, as you know, patients with advanced brain tumor cannot wait.
01:06 They need to receive treatment as soon as possible.
01:10 So to overcome the limitation, in collaboration with experts from national cancer institutes
01:20 from the US, we developed an artificial intelligence framework called DIPLOI
01:27 that classifies brain tumor into 10 major categories.
01:33 Remarkably, we obtained an accuracy of 95%.
01:39 So that is very important and significant.
01:44 How did you train the AI to recognize the differences between the tumors?
01:51 So our model was trained on a large dataset from a wide medical center in the US.
01:59 That consists of about 1,800 patients.
02:10 Once the model was trained on, we applied that model to be done on three different datasets,
02:17 mostly from Europe.
02:20 And that dataset had totaled around 2,200 patients.
02:27 And can it be used for other cancers?
02:30 Of course, and that is what we are working hard on this one now.
02:36 We are trying to expand the work on other cancer types.
02:41 Do you think it will take much work to adapt it for other cancers?
02:46 We can use the same technology, I mean the same framework.
02:50 However, the effort or the time comes mostly from the data collection,
02:57 because collecting data is very time consuming.
03:01 And Danto Kwang, how does it feel for you personally to have been involved in such a
03:07 development which will help people facing cancer?
03:12 That is a very good question.
03:14 And that is why I have been working on this field for over five years.
03:21 I feel we are very helpful to build or to make something contribute to our community.
03:29 our community.
03:29 [BLANK_AUDIO]