Invoice OCR - extracting data from scanned invocies


CoLumbo reads, segments, measures and classifies medical images in order to provide a preliminary diagnosis. The project received European Commission Seal of Excellence and there is USPTO Patent Pending US 62/573,201.

  • Automatically segments MRI images
  • Detects discs and herniated parts, vertebral bodies, etc
  • Provides graded and measured radiology findings
  • Integrates questionnaire data
  • Provides a preliminary diagnosis

CoLumbo Principles

CoLumbo is a product for disc herniation diagnosis based on neural networks. CoLumbo is not just a program which can read MRI images and classify them. The key novelty of CoLumbo lies in the unique machine learning algorithm, based on convolutional neural networks which are capable of reading MRI images and combining this information with the feedback received from patients’ questionnaires in order to provide a complete diagnosis. First, the MRI images are received and segmented, then they are typified and measured, and at the third stage the image data is combined with the questionnaire data in order to provide a final result. Segmentation is made on a pixel level and the use of fully convolutional networks allows quick segmentation maps to be generated for images of any size.

The Team

The team involved in development of CoLumbo is made up of successful award winning developers, 2 neurosurgeons and 2 radiologists as consultants. The team built a neural network architecture that selects and performs training, uses pre-trained models, integrate questionnaires, determines condition severity and much more.

Health Conditions

Currently there are situations when patients may need to wait a very long time for diagnosis and pay a very high price. CoLumbo aims in offering a reliable and accurate diagnosis of back pain problems for a fraction of the current costs. Furthermore, the platform offers completely unbiased diagnosis with no ulterior motives. Blind study indicates up to 30% disagreement between radiologists while CoLumbo’s herniated discs accuracy confirmed by a radiologist is 90%, precision is 83%, recall is 99% and herniation height average error is 0.9 mm. There are 25 types of tissues and over 30 pathologies detected by CoLumbo.

Social Impact

Herniated discs are a major cause for lower back pain and disability and the latter being a significant socio-economic problem all over the world. Annually there are 10 million MRI spine exams in US alone with radiologist wages of USD 140 per hour and second opinion price USD 50 per exam. The main benefits of CoLumbo are lower cost, ease of use, optimal cost-result ratio. CoLumbo holds the potential to alleviate the economic burden imposed on the business as the low cost of the service will render it available to a bigger part of the population, thus helping with the implementation of preventative measures which will lower the number of chronically disabled employees.


Images used

CoLumbo uses standard DICOM images so there is no need to convert images received from MRI into other format.

Automatic processing

CoLumbo automatically detects and displays the slices with most prominent herniation, indicates which discs are herniated, if the herniation is central, lateral or far lateral, the size of the herniation, etc.

Highly confident and user friendly

CoLumbo offers an efficient and user-friendly intuitive interface. It does not require extensive special training. Results confirmed by a radiologist indicate 90% accuracy, 83% precision.

Pathologies detected

CoLumbo can detect the following pathologies Disc Herniation; Spinal stenosis; Juxtafacet cyst; Osteomyelitis / discit / infection; Epidural, para spinal subdural abscess or haematoma; Abdominal aortic aneurysm; Nerve root impingement; Scoliosis; Achondroplasia; Postoperative complications; Spondylolisthesis; Spondylolysis; Osteoporotic fracture; Ankylosing spondylitis; Fibrous dysplasia spine; Vertebral column tumors; Cauda equina and conus medullaris tumors; Spinal vascular malformations; Spinal lipomatosis; Schmorl nodes; Anatomic variations; Modic changes - type I, II and III; Bulging disc; Arachnoid cyst / tarlov cyst; Scheuermann’s disease; Spondyloarthrosis; Other detectable abnormal tissue, Other abnormality without detectable abnormal tissue

image description

Knowledge Base

Didn't find what you were looking for? Please let us know.

image description


Didn't find what you were looking for? Please let us know.

image description

Video Tutorials

Didn't find what you were looking for? Please let us know.