Today’s healthcare industry strongly relies on precise diagnostics provided by medical imaging. In this article, we’ll describe this vast landscape of computer vision for medical imaging, and try to cover both well established and new medical imaging techniques and approaches. Let’s start with some abbreviations which we’ll use along the article: CV – computer vision, IP – image processing, MI – medical imaging, ML – machine learning, HC – healthcare, DL – deep learning.

Broadly, how and where are computer vision and machine learning applied in healthcare industry?

Medical Imaging

  • biggest and most established area
  • analysis of medical images for computer-aided diagnostics

Predictive Analytics & Therapy

based on novel algorithms in:
  • computer vision
  • machine learning
  • artificial intelligence

Infrastructure for Healthcare

  • mass-analysing, storing, mining, sharing and tracking data
  • documentation platforms, etc.

Mobile Healthcare

  • mobile devices as biometric sensors
  • portable and personal capture devices
  • combination of hardware and software (CV/IP/ML/big data)

Imaging Materials or E-Health

  • image sharing within HC systems
  • automated and AI-aided personalized therapy planning and care assistance for better decision-making

Education and Research

  • healthcare professionals training
  • patient education
  • data visualization for research institutions

Medical imaging works with data obtained by different diagnostic technologies

Having years of image processing expertise, we can help you with your medical imaging needs for all basic and advanced modalities.

Basic modalities

  • Optical microscopy – thin blood images, bone marrow, other tissues
  • General X-ray, fluoroscopy (real-time X-ray)
  • Mammographic X-ray
  • Ultrasound – cart-based, portable

Advanced modalities

  • Computed tomography (CT or CAT)
  • Magnetic resonance (MRI)
  • Nuclear/molecular imaging (use of biomarkers for in vivo imaging)
  • Single-photon emission computed tomography (SPECT), positron emission tomography (PET)

Other data/image sources with typical uses:

  • infrared imaging – for temperature monitoring
  • near-infrared spectroscopy – NIRS, OCT… – huge use in neuro-imagery, ophthalmology
  • photoacoustic imaging
  • fluorescence guided imaging (used in real-time – surgery)
  • echocardiography (heart ultrasound basically)
  • endoscopy (visual light modality)
  • consumer photography – smartphones, webcams, portable personal devices
SHOW MORE

Medical imaging process: general steps

The general workflow for computer vision for medical imaging tasks looks as follows: medical imaging process

At Abto Software we have gathered immense experience in the image processing domain.

We can help you solve all kinds of complex, challenging and interesting problems where strong computer vision expertise in needed.

Market analysis

Problems and opportunities

  • Rising demand (increasing number of senior citizens, new healthcare markets)
  • A strong trend towards less radiation exposure
  • Costly equipment and services – especially important for the public sector; private sector adoption also grows
  • Not enough radiologists – high need to automate analysis currently done manually
  • A massive increase in volume, fidelity, and complexity of imaging date – strong need for data compression, storage and lookup/access streamlining
  • Sharing data, expertise, results is hard – currently data is mainly locked in PACS (picture archiving and communication system for DICOM-standard data (Digital Imaging and Communications in Medicine))
SHOW MORE

Challenges in the medical imaging field

  • High development costs for serious research
  • Data gathering for testing/training may be hard/expensive; (hardware tends to be very expensive)
  • Clinical validation of developed techniques is mandatory, long and expensive
  • Adherence to varying local and international policies is required
  • Probably rather inert market – integration of new services with existing solutions may be hard and costly for customers
SHOW MORE

Medical imaging split by use of specific algorithms

Old-school medical imaging

Geometrical image processing (we are very good at this)

Newer CV / ML / AI Techniques

Deep learning, big data (we are starting to add novel approaches to our toolkit)

Global market

$30 bn

Global market revenue in 2017

4-5%

Annual growth rate since 2010

Hardware manufacturers

hardware manufacturers

Global market split

global market split

Computer Vision for medical imaging: current and future usage by domain

Clinical domains

  • Oncology of all sorts, especially: breast cancer, lung cancer, leukemia, prostate cancer – looking for metastases in the tissue; wide use of SPECT and PET
  • Cardiology, atherosclerosis, cardiovascular diseases: vascular imaging, artery highlighting
  • Neuroscience: brain lesions detection and tracking
  • Ophthalmology
  • Endocrinology (esp. diabetes)
  • Pharmacokinetics and pharmaceutical clinical trials: growing usage of imaging biomarkers
  • Lab tests automation: blood counting, tissue cells analysis, changes tracking
  • Other established domains: surgery assistance, planning and automation, healing tracking, better image visualization for healthcare professionals, recommendation systems and aided diagnosis, etc.
  • Novel domains: stats-base diseases prediction, specific diseases tracking (like malaria parasite detection), self-assessment suits.
SHOW MORE

Trending areas

  • AI enabled pose estimation and analysis – accurate assessment and analysis for patients that undergo physical therapy or rehabilitation
  • Breast imaging: for cancer prevention in early stages – the growth of awareness, a lot of campaigns lately
  • chest CT: for chest pain assessment in emergency departments (ED), cardiac angiography (radiography of blood or lymph vessels, carried out after the introduction of a radiopaque substance), lung cancer screening
  • Point-of-care ultrasound – for hand-carried ultrasound devices
  • Emergency medicine: more and more imaging equipment in EDs
  • Neuro-molecular imaging: novel radiotracers to aid in early diagnosis of neurodegenerative conditions (Parkinson’s)
  • Functional neuroimaging: for cognitive neuroscience (PET, fMRI, NIRS, EEG/MEG, etc.)
  • Robotic-assisted and/or controlled surgeries – hugely based on pre-operational and inter-operational images
  • Ese of deep learning for EVERYTHING – the more data, the better
  • 3d visualization, VR/AR applications – for assisted interventions, training, clinical workflow aid, etc
SHOW MORE

Examples of computer vision applications in healthcare

Microscopic thin blood images

  • Automatic differential blood counting, classification and analysis.
  • K-means cluster algorithm based on color image enhancement for cell segmentation.
  • Artery Highlighting.
  • Blood vessel counting; quantifying arteriole formation.
  • Color blood cell image segmentation and recognition.
  • Microscopic thin blood images.
  • Red blood cells classification using image processing. – 81% accuracy.
SHOW MORE

Ophthalmology and diabetes

  • Deep image mining for diabetic retinopathy screening.
  • Deep Learning for Detection of Diabetic Eye Disease – DL-based (developed by Google).

Cancer

  • Cancer metastases detection in biopsy images (e.g. breast cancer in biopsies from lymph nodes) – DL-based; by Google, others.
  • Automatic feature learning using multi-channel ROI based on deeply structured algorithms for computerized lung cancer diagnosis – DL-based.

AML/ALL – Leukemia

  • Classification of acute leukemia using CD markers – SVM-based, 93.89 % accuracy!
  • Chronic lymphocytic leukemia cell segmentation from microscopic blood images using the watershed algorithm and optimal thresholding.
  • Automated screening system for acute myelogenous leukemia (AML) detection in blood microscopic Images.
  • Microscopic image classification for the detection of acute lymphoblastic leukemia (ALL) – DCT-based.
  • Detection of leukemia based on morphological contour segmentation.
  • Automated leukemia detection in blood microscopic images using statistical texture analysis.
  • Fuzzy C Means Detection of Leukemia Based on Morphological Contour Segmentation.
SHOW MORE

Other miscellaneous examples and applications

  • AI supported pose detection and analysis for remote physical therapy and telerehabilitation with no additional sensors and markers
  • Batch-invariant color segmentation of histological cancer images.
  • Automatic shadow enhancement in intra-vascular ultrasound (IVUS) images.
  • Remote noninvasive temperature monitoring system based on infrared imaging.
  • 3d visualization services for microscopy imaging and cell biology (da-cons.de).
  • Multiple sclerosis: automated lesion changes tracking (MRI based, example: icometrix.com).
  • Mole growth tracking, structure and color changes detection (by consumers themselves using specialized app).
  • Recording/broadcasting clinical procedures (surgeries) – multiple angles and feeds.
  • Gesture-recognition based surgery assistance – for hands-free manipulation of patient scans and other information during surgical procedures (adora-med.com).
  • Automated malaria parasite and their stage detection in microscopic blood images.
  • Red Blood based disease screening using marker controlled watershed segmentation and post-processing.
  • A new detection algorithm based on fuzzy cellular neural networks for white blood cell detection.
  • Morphology-based segmentation of bone marrow cell images.
  • 3D reconstruction for assisted navigation (bronchoscopy; solution to guide the endoscopist to target peripheral lesions for biopsy and histological analysis).
  • Automatic segmentation of kidneys using deep learning for total kidney volume quantification. – DL-based.
  • Measuring peripheral vascular reactivity with diffusive optical imaging.
  • Counting contacts between health-care workers and patients within hospital rooms.
  • Tracking patient movement (esp. in ICU).
  • Measuring HC workers protective equipment adherence within hospital rooms.
  • Forensics: MRI-based age estimation (based on bone structure).
SHOW MORE
Abto Software has more than ten years of experience in the domain of computer vision applications development, with dozens of completed projects under our belt – including those developed for the healthcare industry. We will be happy to help you with your next project – please feel free to contact us.

Contact Us

To find out more about Abto Software expertise, request a quote or get a demo of your custom solution.

  • Clicking this button, I agree to the processing of my personal data.

Contact us

Tell your idea, request a quote or ask us a question