Medicine and health

Developing models to detect heart disease and cancer

Background

Medical images that are captured from inside the body using various scanning and imaging techniques have traditionally been difficult and time consuming to analyse by trained experts.Well-performing deep learning models in the medical domain have the potential to assist doctors to increase certainty and to streamline the analysis of such medical images.

In around 2015 the first papers on applications of deep learning on medical images were published. The largest interest at that time was for modalities like MRI, CT and microscopy, but less for PET and hybrid modalities and for the more low-cost modalities like X-ray and ultrasound. The first papers on deep learning in radiology and cancer screening appeared in 2016/2017. Current screening cases are often related to detection, classification and segmentation of tumors, for instance in lungs. In cardiac ultrasound, the first attempts using deep learning were published in 2016.

Main objective

The innovations in the field of medical imaging aim at obtaining more efficient tools for diagnosis support and decision support for diseases such as heart disease and cancer through the use of deep learning technologies.

Challenges

One of the major obstacles for this is the availability of training data within this area. Another very important aspect, when moving into diagnosis, is estimation of confidence and uncertainty in the predictions as wells as explanation of predictions. Furthermore, increased robustness to shifts and changes in shifts in sensors, parameters and cohorts is important as well as real-time inference and the combination of images and data sources such as time series.

Visual Intelligence is advancing deep learning in medicine and health to solve these challenges.

Highlighted publications

New Visual Intelligence paper accepted to NeurIPS
September 23, 2022
ProtoVAE explainability paper by Srishti Gautam and co-authors is published to NeurIPS 2022.
Multi-modal land cover mapping of remote sensing images using pyramid attention and gated fusion networks
August 1, 2022
We present a novel pyramid attention and gated fusion method (MultiModNet) for multi-modality land cover mapping in remote sensing.
Using Machine Learning to Quantify Tumor Infiltrating Lymphocytes in Whole Slide Images
June 21, 2022
Developing artificial intelligence methods to help pathologists in analysis of whole slide images for cancer treatment and detection.