See all the episodes of the Engineering Leaders Podcasts

Ming-Ming Cheng was building Artificial Intelligence to recognize human diseases from CT scans. When COVID hit, Ming-Ming was able to utilize his existing networks to train his AI to assist doctors in recognizing the disease from CT scans.


We know that science, education and engineering… those are most effective ways for people to change their life and to change the environment we are living in.


About Ming-Ming Cheng

Ming-Ming Cheng is a professor with the College of Computer Science, Nankai University, leading the Media Computing Lab. He received his Ph.D. degree from Tsinghua University in 2012 and then worked with Prof. Philip Torr in Oxford for 2 years.

His research interests include computer vision and computer graphics. He has published over 60 papers in leading journals and conferences, such as IEEE TPAMI, ACM TOG, IEEE CVPR, etc. Many of his algorithms have become quite popular in the community, receiving more than 19,000 paper citations.

He received several research awards including ACM China Rising Star Award, the IBM Global SUR award, etc. He is a senior member of the IEEE and on the editorial boards of the IEEE TIP.

You can view Ming-Ming’s online demos for computer vision algorithms: . Many companies play with these demos and come to Ming-Ming for collaboration. Making a lots of exciting software tools for: medical image diagnosis, defect detection, Industrial Safety, insect pests detection, etc.

Ming-Ming’s homepage contains a further introduction to Ming-Ming and his research team, as well as some latest research they are doing. They make the algorithms they develop open source so that these algorithms could help people from many different areas.

About Ramaley Media

Melanie & Dominic De Gioia, from Ramaley Media, are also the hosts of the podcast Engineering Leaders.

Dominic is a mechanical engineer and the Director of a multi-disciplinary engineering firm in Sydney, Australia.

Melanie is the Director of Ramaley Media, a specialised media company promoting STEM. She is the Producer of Engineers Australia’s podcast, Engineering Heroes and is the project manager at The Warren Centre for Advanced Engineering, within the University of Sydney.


This is a “close” copy of the words that were spoken during Ming-Ming’s interview

It is not 100% accurate.

(00:00:00) Ming-Ming: When we were a child and we only have very limited way of working. So my parents are working, everything themselves. Like growing food. And even you want clothes, you need to grow the cotton first.

(00:00:14) And doing everything, you know, very low, efficient way. And we are suffering from many difficulties, but however, when I was small, we know that we don’t have too much opportunities. like in many other countries you can have many variety of working styles. When I was a child the situation is very difficult.

(00:00:35) We know that science education and engineering, those are most effective ways for people to change their life and the change, the environment we are living. So that’s why I’m choosing the engineering science as my major and the, because I’m studying the computer science.

(00:00:55) It’s a very effective way of engineering, the tools to how people working. So that’s why I’m becoming an engineer.

(00:01:03) Mel or Dom: So when you went into engineering, was the main focus so that you could help people back um, in your hometown or where you grew up.

(00:01:11) Ming-Ming: Yes. Uh, partially, uh, for my country. So it’s a very broad area. So, you know, not only my own regions is all over the China, we are working all over China, because computer science, many of the, we develop different softwares, different tools, computer software. We use different tools, uh, to work with many different areas.

(00:01:33) Those tools cannot only help people from my region, but all other regions from China or even other countries. So that’s why I think computer science is very important, it can help people.

(00:01:47) Mel or Dom: So what are some of the key engineering issues that you have experienced or you know, of in your region?

(00:01:53) Ming-Ming: So actually, there are many issues that were in my region and the, which I think engineering’s can play an important role. Again, uh, with an act to use myself as an example. Uh, so when I was very young, so I was living in the Western part of China. So where the economic condition is very poor.

(00:02:17) When I was in the my middle school every month, I only have something like 100RMB to live. 100RMB equals to less then 20 euros for a whole month. You only have that amount of money for you to rent some, bed to sleep and all the food and the other education or everything that covers, that 20 Euros, everything.

(00:02:45) Oh, well, I’m all too, the covers everything. So at that time, so there are many challenges or difficulties, we were suffering. So just to name a few I feel, for example, so if people don’t have enough money to support they are having a good living condition. You know, if you have only 20 Euros for a month, you know, you’re definitely not living, you know, very good, condition.

(00:03:10) So that’s one problem. So economic is to grow and a second, so we don’t have too much food. When I was in my middle school, actually for every meal I only had 1RMB that is so, you know, six RMP equals 2 Euro. So I only have one RMB for Winedale, but not, uh, for many of my meals at we don’t have enough food.

(00:03:35) And also also quality of education. So, you know, locations in big cities, they are much better than in the countryside or in the poor areas. So we were having many, many issues. However, I think, many of these issues, they can be changed, uh, because even when I was in the middle school, I believe that many of these situations could be changed through Engineering, because engineering’s provide us a professional tools to improve the efficiency, how we were working. So while we were very poor, or while we don’t have enough money to live. It is because when we were working, for example, when we want to have more food to eat, we need to grow the food at first.

(00:04:25) And if we want to have a clothes. Uh, beautiful clothes. We needed to grow the cotton first and then make the material for building, uh, the clothes. So every year we have very limited clothes to change. So that’s, that’s why, I think, we need to develop our engineering ability to work more efficient.

(00:04:45) Once we work more efficient, we can deal with all these problems very easily. For example, when you are working very efficient, you’re developing something to, to have many people, and then naturally you’re gotten more income. And it change your economic condition. Also, if you are having the very good engineering tool, it helps to grow food or it helps to create food more efficiently than people have more food to live now in, in China.

(00:05:15) So that, that situation I was describing was something like 10 or 20 years ago, more than 10 years ago. And the many, many areas of China’s people don’t have enough food to eat. But, uh, nowadays, uh, this situation nearly gone. So because China’s economy grows very fast now with nearly don’t have those kinds of hunger.

(00:05:37) People can suffer from that. And also other educations and the many other areas have very good green. And I said, many of these things is because of engineers.

(00:05:49) Mel or Dom: I’d just like to ask that obviously issues such as food security and poverty and, they’re such critical issues. What do you say the priorities for engineers in regards to addressing those issues?

(00:06:02) Ming-Ming: Yes. The priority for engineers. So the most important issue is that we should build engineering tools. So how people are help us to work more efficiently and the once people work more efficiently, the whole economics become better, and then people have more money.

(00:06:23) People have more food, people have better resources to change their life.

(00:06:29) Mel or Dom: I’m just wondering what are you, what are you doing about the issues that you’re seeing around you in your region?

(00:06:37) Ming-Ming: So I’m working in computers, I’m working as a computer scientist. Actually, I am a professor at Nankai University my major research area is computer vision. What is computer vision is a tool that, so for example, uh, we have computers, cameras, we see many things and this capability, if it only allows people to do it, then it can, people need to do everything themselves.

(00:07:04) If we want computers to do these things all, for example, if we have computers to recognize people, to recognize different packets, Then we can allow the machines to do many things themselves. So this takes engineering too, to have machines, to do the tedious or boring jobs and you start over humans and it can do it more efficiently.

(00:07:28) I just gave you a few examples for you to better understand that what I’m doing.

(00:07:34) Mel or Dom: Yeah that would be great

(00:07:35) Ming-Ming: The CT images, for example, we have doctors to recognize different CT images. If people want to go to a hospital and then people have, some people have some CT scans where usually, the doctors can get actual, like, 300 high resolution images of the CT scans.

(00:07:57) So among all of these images, there might be some areas that might indicate problem or health issues. However if you ask a doctor to go through all these images carefully, a doctor would need something like a half an hour or one hour to check every position. If computer science do it, then if we have many, many data and historical data or people or doctors saying, okay, this patient has this certain kind of disease, or that patient has another kind of disease.

(00:08:32) Once we have many data like this, we can learn from this data so that the computers can read the CT images itself. And then we can allow the computers to a help doctors to check whether there are problems for different patients.

(00:08:50) So because, I’m working in computer vision more specifically we build the computer software to allow machines to understand the images.

(00:08:59) Once this images is medical images, then we help doctors. There are also other examples. For example, if you’re using cell phones now, nowadays people like to use cell phones to take pictures. Actually, we have cellphone companies to build the computer software that makes you to have better pictures.

(00:09:20) For example, when we were talking, we can detect where is the human region or many other regions, and then we build this capability so that you can change the background or you remove the noise or make people looks more beautiful or this can helping people take better pictures? This is how we work with a smartphone companies.

(00:09:44) We also work with other areas like, uh, wind energy companies. So there is a very big wind energy devices company in China because that’s a device is very big. The hightest pointer to the ground is like 200 meters. It’s tackling in some areas where we are, the wind is very big. Then in the winter, there might be some snow or ice covering the device.

(00:10:14) Then if that device keeps running, then it can cause many problems. Later one, we developed some computer software that’s a actually is a camera looking at the device, just the 24 hours a day. Once it finds some problems, it will alert the backgrounds and also try to stop the system or to stop causing problems.

(00:10:38) It’s kind of a capability that allows computers to recognize images so that people used to do it themselves are many boring work. You look at a similar images, like I thought of the everyday or every hour, so you’ll know hundreds of them.

(00:10:57) And that it’s very boring. Now, we change, we switch this working in modality to computers, then computers try to recognize all of these images and have them replace those tedious work. Because if we allow these things to be done by humans, if you’re doing one images, it’s good. Everywhere is good at it.

(00:11:20) At there professional areas, but if you are working with 1000 images, Even if you are very professional, you’ll match you’ll be doing a very bad job just because there are so many repitition. There are many, many applications. And we are working with many different areas.

(00:11:36) Mel or Dom: And a lot of efficiencies are coming through. And I just want to take a little bit back. So you were doing this aI camera imaging initially on tuberculosis. Uh, so from that medical example that you were talking about earlier, was that your initial area of study

(00:11:56) Ming-Ming: yes. That’s the one area of my research

(00:12:00) The way we are doing this is actually, there are some other company who are really interested in solving this problem and they think that we have pretty good, uh, computer vision techniques. And then we worked together and then we discussed this issue. And then after discussion, we found that it’s really a big issue and, uh, whilst we have it already, it could have be able to help many people.

(00:12:23) Then also they have many data collected from historical data, from many hospitals. And, uh, we have the computer tools or engineering tools, different softwares to work with those data. And then we work together to develop your computer software like to recognize different disease. So the tuberculosis is only one disease.

(00:12:47) Mel or Dom: So with all these applications that you’ve been able to utilize the scans within the hospitals how have you applied that to COVID-19.

(00:12:55) Ming-Ming: So for this, medical images, we are not working with a single disease. We are just actually different disease, just a different data for us. We are simply working with images and notations and all the, these things.

(00:13:09) So, in February last year and, uh, and when the COVID19 is it become really a big issue in China . So when COVID-19 is really a big issue in China, and because that company, they have a system running in different hospitals, collecting the CT images that scanning for other diseases, and then they have many data they can get from the hospitals. because we have previous collaboration, Recognizing disease from our CT images for other disease.

(00:13:40) Now we are discussing. So can we apply the similar techniques to deal with COVID-19 as well.

(00:13:47) Mel or Dom: So you’re, you’re able to apply those applications to COVID, to diagnose, COVID 19 in patients, the scans.

(00:13:55) Ming-Ming: Yes CT scans. Yes.

(00:13:58) Mel or Dom: That’s a, it’s an amazing that you’ve been able to pivot your technology to respond to the, uh, the needs that had come about in February, 2020, how has that been?

(00:14:09) What obstacles have you had to overcome for that?

(00:14:13) Ming-Ming: So actually the techniques. So the computer vision software, or the computer algorithms, they are ready. So the major challenge is data. So once you have enough data from patients and have enough annotations from doctors, finding later on issues is not too, too big issues because we have really many experience.

(00:14:36) Working with CT images and the different other diseases. COVID-19 is just a new pattern from a image recognition point of view. It’s just a new group of data from our point of view. So, because we work with a company, that company worked with more than 300 hospitals all over China, most of them are in very big hospitals.

(00:15:00) And, uh, because that company has working with many hospitals in China, they have the system running and the system collected many, many CT images and after the COVID-19 outbreak. So, the company partnered with, uh, hospitals In Wuhan. And then they say they are asking if this data could be used for emergency and to deal with this COVID19 issue. And then, those hospitals allowed this data to be provided to researchers and also they organized doctors from other places, because in Wuhan the doctors are too busy. They organize doctors from other areas of China, just online send the data to these doctors and then the doctor had to annotate many images and then after we collected so many data, then we can learn from this data.

(00:15:55) So it was a major challenge. Is that in that situation, so you’ll know in February and that’s the year and the situation in China was really a serious. And most people were not allowed to go out and the people stay in their home, but I didn’t go out to my home for three or four months.

(00:16:12) I just didn’t go out and we work everything online. So it’s the biggest challenges that how to organize so many doctors, help us annotate them. And to get to the data ready. Once the data is ready, then we can have everything working online. The computer software and all these things are what we are familiar with.

(00:16:36) Mel or Dom: With the, the way that it’s spread and the, the speed in which it did, did these sorts of technologies help speed up diagnosis as well?

(00:16:46) Ming-Ming: Yes. Especially in the early stage, this type of technology help the doctors, even in Wuhan and many other areas very well. So basically if we want to detect whether there is a COVID-19 or, or not, We have this, testing. Even they use the traditional other testing techniques, seeing data.

(00:17:07) So these people have no problem. This CT images, give us another perspective. So if it gave us another tool to have a second check whether the CT image, you can find a problems. So this is one issue. The other issue is that as those CT image is a very efficient.

(00:17:27) After you’re doing the scan you can now upload the data to the computer and then computer can help to identify the disease region just in a few seconds. So this helps the doctors you’ll know that’s a situation. The doctors are crazy busy.

(00:17:44) They don’t have so many time to carefully check . So the system can detect a 200 or 300 imagesin only a few seconds or 10 seconds.

(00:17:55) So if you’re simply just to have more computers, you can make it, even less time to detect all the problematic regions that it can alert the doctors. If it detects some patients with COVID-19, then it helps the doctor to work more effeciently and save their valuable time.

(00:18:17) Mel or Dom: That’s amazing. So you get, what is it? 200 to three. How many did you say? 200 to 300.

(00:18:22) Ming-Ming: Yes.

(00:18:24) Mel or Dom: Wow.

(00:18:25) Ming-Ming: People can do it as well. However, the advantage for the computer software is that it can do something much faster if you have more computers, or if you have more graphics tab on your desktop, on your computer, you simply just do it 100, the 200 or even a seven times faster than a normal human could do. Also, it can work 24 hours a day.

(00:18:50) If one is not enough, you buy another or you buy 1000, always make thing more efficient.

(00:18:58) Mel or Dom: Yeah. Increasing that efficiency. This is a podcast for World Federation of Engineering Organizations, and we’re coming into our World Engineering Day for 2021.

(00:19:08) All about engineering for a healthy planet. Why do you feel World Engineering Day is it an important day?

(00:19:16) Ming-Ming: I think Engineering Day is a very important day because the engineering day gives the message to ordinary people, to the more general population that what engineers are doing. Engineers is a fascinating, it’s very interesting because.

(00:19:36) If you’ll have good engineering tools it really help people to work more efficiently. Sometimes the efficiency is really amazing. So the capability or what we can do is really beyond the imagination of many people. For engineering day we know we have many examples, we remind the people to know more about engineering and to be more familiar with engineers. Especially the children, who will find engineering really interesting and fascinating.

(00:20:09) So that’s why I think Engineering Day is so important because it allows the general public, especially the children to understand more about engineering.

(00:20:18) Mel or Dom: Yeah, the hope is we can get future generations interested in engineering and hopefully have a wide range of engineers solving the problems of the future.

(00:20:28) Ming-Ming: Yes.

(00:20:28) Mel or Dom: Thank you so much for joining us tonight. It’s been really great hearing about the work that you’re doing.

(00:20:33) It has been. Thank you so much for joining us.

(00:20:35) Ming-Ming: Yes. Thank you so much.