The person who released 'AI to detect breast cancer and brain tumor' is not a professional but an amateur programmer, buy dozens of high-performance graphics cards on his own

In 2018, a programmer named ' coolwulf ' released a free website to detect breast cancer with about 90% accuracy from uploaded X-ray images. In China's online developer forum 'V2EX', various questions were asked to developers, but coolwulf remained silent without answering personal questions. And in 2022, coolwulf released a platform called ' NeuralRad ' that detects the exact location of brain tumors with AI. Toutiao, a Chinese news site, is interviewing coolwulf, who has reappeared on the front stage.

This name is 'Introduction', 50 张 1080Ti Opposite Cancer Disease-Today's Toutiao
https://www.toutiao.com/article/7094940100450107935/?wid=1653278073761

This 'amateur' programmer fought cancer with 50 Nvidia Geforce 1080Ti
https://howardchen.substack.com/p/this-amateur-programmer-fought-cancer

The programmer who works under the handle coolwulf is Hao Jiang, who lives in the Midwestern United States. Jiang is a person with a degree and a PhD from the Department of Physics at Nanjing University and the Department of Radiology, Faculty of Nuclear Engineering, University of Michigan, and did not major in programming as an academic discipline. 'My main career is medical imaging, but in my spare time I'm an'amateur'programmer working on open source projects,' Jiang describes himself.

Jiang has been interested in programming since he was a child, and he posted side projects on SourceForge.net and his website even before the development platform GitHub appeared. Around 2001, he participated in the Mozilla Foundation 's open source project, and also contributed code to the K-Meleon web browser that uses the HMTL rendering engine Gecko and the Firefox predecessor web browser 'Phoenix'. In 2009, he also launched a website to help international students in North America book hotels at low prices.

Jiang, who has been involved in various projects as a programmer, did not choose a programmer as a profession, and after learning about medical imaging at the University of Michigan, he directed research and development of medical imaging equipment at Bruker and Siemens . bottom. Since then, he and Dr. Lu Weiguo , a lifelong professor at the University of Texas Southwest Medical Center, have set up two software companies targeting radiation therapy to work on cancer radiation therapy and AI development.

The man wearing the cap on the right side of the photo is Mr. Jiang.

Jiang seemed to continue his career as a science entrepreneur, but at the age of 34 he faced an incident in which a classmate at Nanjing University suffered from breast cancer and died leaving a 4-year-old child. And that. Knowing that illnesses can destroy families and that there are many cases where breast cancer detection is delayed and treatment is not in time, Jiang said, 'AI detects breast cancer from X-ray images,' which was not common at the time. I started working on the idea.

Although it was difficult to develop an AI to accurately detect tumors, Jiang downloaded a large X-ray image dataset from the University of Florida website and wrote a dedicated program to use all the information. It was converted to a format. He also energetically gathered training data and read related research literature, such as contacting the University of Barcelona to obtain permission to use the breast cancer dataset, which is a private resource.

In addition, building an AI system requires not only data and code, but also hardware with high computing performance. So Jiang bought 50 Nvidia GTX 1080 Ti with his pocket money to build a high-performance machine locally. It wasn't easy to buy as many as 50 graphics cards because crypto mining was booming at the time, but I asked many friends to check the online site frequently and managed to get 50. Was completed.

This is a machine built by Jiang with 50 graphics cards.

And in 2018, Jiang took three months off from work to complete a free AI breast cancer detection website. It is said that this website does not store data to protect the privacy of patients, so it is unknown how many people actually used it, but under Mr. Jiang, there are many mainly in China. I received a thank-you email from my patient.

This website, where AI detects tumors, was especially useful for people living in remote areas with limited medical resources. At the time of writing the article, many systems for detecting cancer using AI have been developed, but as of 2018, Jiang's website was advanced and attracted a lot of attention from the industry. .. Many medical institutions in China and abroad, such as Fudan University Hospital, expressed their gratitude by e-mail and offered financial and technical support.

Jiang explained why he didn't monetize the website he created by taking out his pocket money: 'Cancer patients and their families are having a hard time and everyone wants to help them. I think I happened to have that power. '

Jiang, who released the AI breast cancer detection website, had a brain tumor in 2021 and was treated with whole-brain irradiation therapy (WBRT) , but the tumor relapsed a few months later and waited for death. There was an event that it became only. WBRT, which irradiates the entire brain, damages not only cancer cells but also normal brain cells, and can be said to be a kind of 'indiscriminate attack'. Therefore, considering the tolerance of radiation dose to important structures of the brain, it seems to be a treatment method that can be used only once in a lifetime.

On the other hand, stereotactic radiation therapy (SRS) , which irradiates a specific site with pinpoint radiation, can treat the tumor without damaging the normal part of the brain, so it can be applied multiple times to one patient. is. However, the disadvantage of performing SRS is that the location of the tumor must be accurately identified in advance, and it takes longer to save one patient. Therefore, it is customary to perform WBRT, which irradiates the entire brain, for patients with 5 or more lesions in the brain.

So Jiang began a project to use AI to pinpoint the location of brain tumors, reduce the burden on doctors, and make more patients eligible for SRS. This problem was much more difficult than breast cancer detection and could no longer be developed by one person alone, so Jiang turned to the University of Texas Western Medical Center and Stanford University for help.

Through a collaborative project, the research team succeeded in developing a 'model that appropriately outlines and labels brain tumors,' 'a model that rapidly reduces false positives,' and 'a model that classifies multiple lesions into appropriate treatment courses.' .. It was announced at the 2022 Spring Clinical Conference of the American Medical Physics Society and the annual general meeting of the American Medical Physics Society, and is widely recognized in the industry at the time of writing the article.

In this interview, Jiang repeatedly emphasized that it wasn't just one person who was able to achieve this result, but many collaborators who were fighting cancer together.