This article is a poem that even beginners can manage to do their best unexpectedly.
I am a graduate of a non-information department who usually works at a medical institution as a radiological technologist. I have no programming experience.
I decided to go on to graduate school around 2017 I was studying TOEFL as a preparation for the hospital exam. I often listened to the Radiological Society of North America podcast as one of the listening measures. Radiology Podcasts | RSNA
I am grateful that he will explain the necessary material for work in English. I passed the graduate school in the 2018 exam and took a break.
In the radiology podcast around August 2018 AI will be used for medical images Since the number of papers related to AI is increasing, it was said that radiology AI will be published in 2019.
inside that -There are various frameworks (tensorflow, caffe ..) that can be used free of charge for machine learning. ・ Results will be obtained if the images are trained. ・ It is useless if it is left as it is, so please verify it properly. NIH plans to publish a large chest XP dataset, which anyone can use.
There is a story that, I investigated immediately and it looks like Python is good Around November 2018,'Hello world!'
While exploring the chest XP dataset I found a dataset published by RSNA. I arrived at the fact that it seems that I am doing this in a competition format. kaggle rsna-pneumonia-detection-challenge.
I can't even manipulate Python, but I can't understand machine learning I just read Notebooks and Discussions.
I want to try it too! And move your hands At udacity I took a machine learning course and studied. Buy Curry's Tutorial I made my debut in Kaggle by submitting to titanium by imitating what I saw.
Curry's tutorial was really learning. Because the instructions are specific and the code is also written It is easy to understand what each one is doing. Anyway, it would be nice if you could explain it in Japanese!
Although it is a calculation environment, there is no GPU etc. and there are only old notebooks purchased about 10 years ago. I thought that machine learning was impossible locally I could do a lot on the Kaggle kernel. I learned about the existence of Google colabratory and am currently working on these two.
Using the code of the pneumonia competition I also tried to detect GGO from the chest XP of a patient in my hospital. (Of course, I get permission for ethics and so on.) I tried cases such as early stage lung cancer, interstitial pneumonia, and atypical mycobacteriosis. Only a poor result was obtained. I think this is because my model is not accurate. This was announced at an international conference called FHS in June 2019.
The qualification requirements for contributor are the following four and profile input.
-Run 1 script -Make 1 competition submission -Make 1 comment -Cast 1 upvote
How about the contributor seems to be relatively easy. I had cleared everything before I knew it. I filled in the profile and became a contributor.
Just because I became a contributor I'm still a beginner and it's not very valuable. Still knew Kaggle, I think it might give the impression that he is a person who moves his hands a little. In the pneumonia competition, there was a Japanese team at the top, and I was impressed that it was really amazing. I want to do my best so that I can get a medal someday.