The team of CoLumbo.
SHARE
SHARE
Published by Valeri Petrov, Roberta Tihomirova, and Nikola Yanev.
A few years ago, Nedelcho Georgiev, the mastermind behind CoLumbo, suffered a back injury. After a consultation with a specialist and a detailed examination, it turned out that he needed to have an MRI scan. In a conversation with the radiologists during the scan, he learned that there are various software solutions for reading MRI scans of different parts of the body like the heart and the brain, but no such software for the spine – which puts a lot of pressure on radiologists and leads to mistreatments, stress, and valuable time wasted.
This is how the idea for CoLumbo was born and the teams’ proven track record of developing numerous projects in healthcare related to data extraction, AI-based image processing, and document classification, came to play.
CoLumbo is the first AI tool
that helps radiologists analyze lower spine MRI images. The algorithm recognizes common soft tissue and MSK (Musculoskeletal) pathologies of the lumbar region and drafts a report, which the doctor can review and edit before making the final diagnosis.
The lumbar is the most vulnerable part of the spine
which according to studies, has the highest number of diseases and is the third most common reason to visit the doctor’s office. Lumbar spine scans are the second most prominent medical scan in the world (following brain scans). At the same time, scan readings are still 100% dependent on human interpretations and lumbar spine imaging marks a pain point for patients and radiologists alike.
0
%
of adults experience back pain at some point in their lifetime.
36
/2
radiologists have 1/2 of the time required for a thorough analysis.
0
%
of the radiologists disagree in 30% of the reviewed cases.
Typically, one radiologist does around 30 diagnoses per day under a lot of time pressure. On average, radiologists spend 13-15 minutes per diagnosis, but they need about two times more time for a quality analysis. As a result, 30% of the radiologists disagree on a given diagnosis, leading to stress, mistreatments for patients, and liability risks for the clinics. Also, there’s a knowledge gap as 70% of radiologists are generalists, which means they are not specializing in a particular body part. Globally, there is an increasing shortage of radiologists, and their salaries are significant, making it difficult to combat this issue by simply increasing the number of radiologists working in a given hospital.
At the same time, image diagnosis and AI in imaging gain traction,
and the market is expected to grow with 75% per annum in the upcoming years. There are various software products focusing on imaging diagnosis for different body parts, however, one technology cannot be easily transferred to serve other purposes. The reason is multifaceted. On the one hand, different pathologies require different types of scans in order to be detected. And on the other hand, each pathology and type of scan needs separate training of the algorithms with new datasets for the particular purpose. The annotation of the scans must be done manually before the algorithm is fully trained to recognize different pathologies, which makes it difficult (and very expensive) to introduce new pathologies or expand the scope of the software. This is a key advantage of Columbo, which utilizes rich university data sets and radiologists in Bulgaria to annotate the scans.
Columbo is the first multi-pathology MRI scans diagnostics software
focusing on the lumbar spine. It is to be used as an assisting tool, helping radiologists spend less time per diagnosis, but does not remove the responsibility they carry for each diagnosis. CoLumbo helps radiologists work efficiently with MRI and achieve consistent, high-quality results:
- 25% time-saving in its first version;
- 90%+ accuracy;
- Automatic report generation;
- Consistency (eliminating bias) and standardization;
- Peace of mind for radiologists, clinics and patients.
The software offers a full description of the pathologies, colored with the different segments, as well as partially pre-populated reports. Currently, this results in a 25% saving of the time radiologists need per analysis, expecting to save up to 80% in later versions, as well as up to 15% reduced errors on average in all pathologies. The accuracy of the algorithm has been validated by an academic team from the University Hospital Bonn in Germany and the product is used by several hospitals in Germany, France and Bulgaria.
So what’s next after the investment of Eleven and BrightCap Ventures? Currently, CoLumbo’s team is further developing the software, with their next step being to create a version, which helps spine surgeons plan their operations more accurately.
If you are curious to learn more about CoLumbo or request a demo visit their website – https://columbo.me.
Check out other #Healthcare companies which we’ve backed – BeMe.ai, Healee, Biopix, AYO, mBrainTrain.
