There is currently no AI-based solution for brain MRI detection, which has been CE approved and gained any broad and unanimous use in the general clinic. Most of the technologies focus on specific diagnoses (e.g. bleeds) and support for pharma trials and R&D delivering narrow outcomes. Specific vendors cater to certain disease/pathology areas in the field of brain MRIs: Specifically, within dementia, Quantib/NeuroQuant offer solutions that allow segmentation of MRIs for both volumetry and white matter lesion segmentation, while Icometrix offer white matter lesion segmentation related to multiple sclerosis. None of the 100+ AI image analysis providers interviewed by Signify Research14 enter the actual patient examination step, since they do not deliver real-time image analysis besides focusing either on a very specific brain condition or in a general perspective leading to low accuracy rates. This is precisely one of Apollo´s USPs.
Each of the Apollo product features has distinct uniqueness to existing solutions as described hereafter, and combined they provide a unique value proposition with significant productivity and quality improvements in the general radiological clinic (e.g. improved clinical efficiency and scanning time, less time spent by radiologists, increase on patient care experience, higher safety through reduced use of contrast chemicals, support for non-radiologist reading, etc). Unlike competitors focusing on diagnostics during reporting (or needing to generate more data during scanning), we offer workflow efficiency gains by bringing new insights, ensuring clinically relevant accuracy above 95% and letting the radiologists focus on their core expertise. Apollo will be a breakthrough on the AI market for medical imaging as current products do not focus on improving the overall workflow with true real-time decision support. Generally, existing algorithms are too slow and compute power-intensive, as opposed to deep learning-based algorithms, and they focus on the quantification of a few biomarkers, which is useful in a narrow often research-based context, but less effective in the general clinic. This makes Apollo unique and well ahead compared to the available solutions. The key value streams benchmarking is described below:
1) Smart Protocol. Today, there exists no solution to adapt the protocol during scanning and there are no available commercial products that handle automatic determination of brain MRI protocols. Current state-of-the-art includes different protocols for different pathologies (e.g. epilepsy protocol, granuloma protocol etc.) which are manually selected. Solutions like GE MAGiC/Siemens GOBrain that acquire several fixed sequences are in validation phase. However, special sequences like MRI spectroscopy, or time-of-flight need to be manually set. Currently there are no real-time protocol modification tools available. Most of the modifications are done manually by the radiographers together with a consulting radiologist. In the absence of radiologists, the most elaborate protocols are usually followed. Solutions like ones offered by CorTechLabs/Icometrix may be used to determine the underlying pathology but they do not provide any real-time info, and only offer identification of one specific biomarker, i.e., white matter hyper intensities.
2) Image Reporter. There are many existing image viewer solutions and some quantification solutions that can quantify segments and biomarkers in the brain (e.g. neuroQuant, which provides measurements of total brain volumetry, mostly regarding volume/atrophy of brain structures). Existing solutions are, however, too slow to be relevant during scanning but also too specific (focusing on quantification of brain segments) making it necessary to have several independent solutions. While most of the brain MRI solutions like Icometrix, CorTechLabs etc. offer image reporting, they only rely on appending DICOMs (Digital Imaging and Communications in Medicine) with PDF reports, slowing down the process.
The Image Reporter is a real-time image viewer guiding radiologists to focus on findings whereas it prioritizes the most acute ones. This is mainly applied to help the clinicians to immediately see findings in current scans and to convey the recommended sequence, if any, for the radiographer to perform (see Smart Protocol). Apollo offers additional flexibility on navigating through the findings of segmented MRI images generated by deep learning, on the way to identify a vast range of brain conditions in real-time scan. This information is also stored in PACS as overlays or secondary captures supporting radiologist reporting.
3) Triage Advisor. A few systems focus on triage (e.g. AIdoc) for acute cases, but mainly within the CT modality. Commercially, there are no brain MRI based triage reports being used. The closest AI-based triage solution available for brain is by AIdoc for the emergency condition intercranial haemorrhage (focus on emergency/trauma conditions) based on head CT scans. Enlitic provides a triage solution for lung diseases that is based on deep learning in lung imaging such as chest x-ray and chest CT combined with unstructured information such as reports.
