Case studies

Applications of artificial intelligence in the healthcare industry

Credit: Bert van Dijk/Getty images.

Powered by

Nabla uses GPT-3 to power its healthcare assistant platform, Copilot  

French digital health startup Nabla launched a digital assistance tool for doctors in March 2023. The tool, named Copilot, uses OpenAI’s GPT-3 to translate conversations between doctors and patients. Copilot is added as a Google Chrome extension and helps to record and repurpose information from video calls. It is compatible with any video call system, including Google Meet, Zoom, and other telehealth products. 

The tool automatically converts doctor-patient conversations into various documents, including prescriptions, confirmation letters, and consultation summaries, as well as updating patient medical records. Copilot does not retain or store any data; all information is deleted when the browser is closed. 

Nabla Copilot helps doctors focus on the patient, allowing for better engagement throughout consultations. Doctors can also ensure that everything is recalled both during the appointment and for follow-up encounters. The platform can reduce the often-overwhelming burden of repetitive tasks for doctors, freeing up more time to attend additional patient appointments. Nabla plans to develop a model tailored to different languages. The startup is also planning to launch an in-person consultation tool soon. 

Harvard Medical School researchers use AI to turn low-field strength MRI scans into high-resolution brain images 

Neuroimaging with magnetic resonance imaging (MRI) is one of the most valuable tools to study the human brain in vivo and can potentially increase understanding of neurological conditions and genetic linkages. Currently, neuroimaging tools need MRI scans with several requirements involving resolution and contrast for accurate 3D analysis.

However, most MRI scans worldwide do not have the required criteria. In clinics, clinicians prefer 2D scans with lower resolution. The data from these clinical scans cannot be used in neuroimaging research, a huge barrier to progression in the field.

Harvard Medical School researchers have developed an ML super-resolution technique named SynthSR, using AI to bridge the gap between clinical and research-grade brain MRI scans. SynthSR takes clinical brain lower-resolution MRI scans and turns them into high-resolution T1 scans, which are then usable by all existing neuroimaging tools.

This tool has the potential to enable research on populations and rare illnesses which are currently underrepresented in neuroimaging research. Enhancing the image quality of portable MRI scanners could also revolutionise their use in critical conditions or medically deprived locations where MRI suites are unavailable or inaccessible (Iglesias et al., 2023). 

King’s College London researchers create AI-model to predict cancer spread  

Researchers at the Breast Cancer Now Unit at King's College London have built an AI model to predict the likelihood of breast cancer spreading in patients with triple-negative breast cancer. Triple-negative breast cancer, a rarer and more aggressive form, accounts for 15% of all breast cancers in the UK, often leading to poorer outcomes (Cancer Research, 2023). 

For the study, published in The Journal of Pathology, researchers tested their AI model on over 5,000 lymph nodes donated by 345 patients to biobanks such as the Breast Cancer Now Tissue Bank and Tianjin Medical University (Verghese et al., 2023). The AI-model, a deep learning (DL) framework named smuLymphNet was used to carry out image analysis of lymph nodes in cancer patients, cross referencing against patient records and whether the cancer had spread. 

The study was based on scientific findings that even in cases where breast cancer cells have not spread to lymph nodes, the lymph node immune responses and changes can predict the likelihood of the triple-negative breast cancer spreading to other organs in the body. The AI model made this prediction by analysing the immune responses in the lymph nodes and finding specific patterns between different patients. 

Following these findings, the model is to be tested further at centres across Europe and in clinical trials. Hopefully, this innovative model will provide affected patients with a more tailored treatment based on the likelihood of the cancer spreading, helping save lives through earlier and more targeted treatment. 

GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.    

GlobalData’s Thematic Intelligence uses proprietary data, research, and analysis to provide a forward-looking perspective on the key themes that will shape the future of the world’s largest industries and the organisations within them.