A.I. in fight with COVID-19 (n-COV19) Open Source

5 min readDec 19, 2020

A.I. result of one image, when looking for non-covid-19 and covid-19 pneumonia, using x-ray

The idea was to use current MODERN technology to help fight the new virus 🦠 covid-19.

As China has provided an online tool to scan CT images for the virus (you can access it here http://121.40.75.149/znyx-ncov/index#/app/online_try), around the world, it is cheaper, faster and maybe the only solution to test for pneumonia. Some hospitals have none and some have only one CT. To be able to disinfect that after screening patients would take way to much and it might be needed for an emergency. An x-ray is fast and easy to to and nearly every hospital has at least one device.

I LOVE turn-key solutions, the hassle-free to start compiling, building, coding etc. The next thing I did was to search for an A.I. solution that was open source, already built, a turn-key solution. Didn’t quite find anything. Not even sources to train a model for COVID-19.

Then I found a project that Jordan has built. I’ve downloaded it and tried it on my mac, then moved to a Ubuntu in a VirtualBox. Long story short: it had some issues and there was some improvement needed on a more functional and practical way. Worked with Jordan on that.

Now, you can simply download it from this link and install in any hospital you need! It is open source so if other would like to connect and improve, please reach out!

The goal is to help save lives!

Installation instructions:

Download and install VirtualBox according to your Operating System https://www.virtualbox.org/wiki/Downloads
Download the OVA file (Ubuntu image with the A.I. Detector pre-installed and configured) https://drive.google.com/open?id=1Z-4SKbEuG5n8riqNXhDqw0VJtv83AXdc.
Import the downloaded OVA file.
The virtual machine is set to 1GB RAM for computers that lack too much RAM memory. If possible, give it more.
Create a shared folder or map a network drive where you will be saving the x-ray images.
Username: “user”; Password: “user”.
Run the “A.I. Detector” icon on Desktop and wait for it to load.
Simply activate the detection of non-covid19 or covid19 OR both simultaneously, then drag and drop the image.
Wait for the result (instantly up to a few seconds). Repeat from step 8 for a new image.

MORE images would help us make it better!

Stats for nerds:

#######TRAINED NON-COVID19 PNEUMONIA VS NORMAL LUNG TEST REPORT [LOADED MODEL/WEIGHTS]
Set: train, normal images: 1341, illness-positive images: 1335
Set: val, normal images: 8, illness-positive images: 8
Set: test, normal images: 234, illness-positive images: 390
Found 13108 images belonging to 4 classes.
Found 624 images belonging to 2 classes.
2020–03–22 18:06:56.495726: W tensorflow/core/framework/allocator.cc:107] Allocation of 8640000 exceeds 10% of system memory.
2020–03–22 18:06:56.508791: W tensorflow/core/framework/allocator.cc:107] Allocation of 46080000 exceeds 10% of system memory.CONFUSION MATRIX FORMAT — — — — — — — — —[true positives false positives]
[false negatives true negatives]CONFUSION MATRIX — — — — — — — — — 
[[207 27]
 [ 41 349]]TEST METRICS — — — — — — — — — — — 
Accuracy: 89.1025641025641%
Precision: 92.81914893617021%
Recall/Sensitivity: 89.48717948717949%
Specificity 88.46153846153845%
F1-score: 91.12271540469973#######TRAINED COVID19 PNEUMONIA VS NORMAL LUNG TEST REPORT [LOADED MODEL/WEIGHTS]
Set: train, normal images: 74, illness-positive images: 74
Set: test, normal images: 20, illness-positive images: 20
Found 148 images belonging to 2 classes.
Found 40 images belonging to 2 classes.CONFUSION MATRIX FORMAT — — — — — — — — —[true positives false positives]
[false negatives true negatives]CONFUSION MATRIX — — — — — — — — — 
[[14 6]
 [ 3 17]]TEST METRICS — — — — — — — — — — — 
Accuracy: 77.5%
Precision: 73.91304347826086%
Recall/Sensitivity: 85.0%
Specificity 70.0%
F1-score: 79.06976744186045

For assistance or contribution to the AI join us on Slack https://join.slack.com/t/aicovid19pneu-o7g9651/shared_invite/zt-ctp3ulle-tzzagHT61BjTtfhusk6b6w

Show ❤️ & support other projects too by

BTC: 3FJBr6WhgCnbToGBWQewMTHt17v4ZxLAZ4

ETH: 0xB09894a834cefCf1a2Eec05C192Ff088955B63eF

A.I. in fight with COVID-19 (n-COV19) Open Source

Written by Komodevid