Yes, we are using Explainable AI, using rule-based and case-based reasoning, because we know that doctors and patients need to stay in control of the decision.

To be able to take the decision, patient, clinicians, and scientists need to know why and how AI produced the results. Vitae Evidence web-based application provides quality control checkpoints for each step of the process (information sources updates, search strategies from case summary, references inclusion/exclusion rules, evaluation of evidence and built insights), and the Coremine Vitae reports explain the results using Clinical Decision Support and Patient Decision Aids best practices.

We bring complementary, highly specialised knowledge to the healthcare team, presenting all available treatment options with their rationale, risks, expected outcomes, and practical information, to aid the clinical decision.

Vitae Evidence combines human expertise and machine intelligence to find personalised medicine evidence at scale.

Bias may arise if the AI does not reflect the diversity of the patients, clinicians, or scientists it serves. A lack of diversity in the data used to train the AI may create blind spots. Vitae Evidence assigns more weight to the expert rules system supporting Evidence Base Medicine processes rather than on the data themselves to prevent this risk from occurring. Usability engineering of our AI suggestions indicates their evidence basis and reliability level to end-users with little time at hand.

Vitae Evidence keeps learning and therefore changing after it is released. AI algorithms can evolve as they encounter new data. That is why our built-in feedback loop is vital to capture end-users curation of insights. Coremine Vitae Analysts uses the software every day. We have a tight feedback loop between our Engineering team and Coremine Vitae scientists.

Did you notice how anything a computer does those days is now considered Artificial Intelligence (AI)? It seems like AI is frequently confused with Algorithms. Then for some others, Machine Learning (ML) is the only kind of AI there is.

Algorithms are, by definition, step by step instructions automatically executed. Algorithms become Artificial Intelligence (AI) when they interpret input information to decide on the best action to take to achieve a given goal.

Strengths of Vitae Evidence Intelligence

1) It is not limited to a given diagnosis;
2) Its knowledge-base stays current;
3) Experts stay in control and teach the AI as they use it.

Vitae Evidence analyses population data and user preferences to decide on the Search Strategy and Rules for answering well focused clinical questions. Because it collects all analysts' actions, it is good at finding which rules they are likely to apply to include or exclude references based on their history. It also groups similar users and similar cases to suggest possible treatment options.

Algorithms are continuously fetching new knowledge from reputable information sources worldwide. They can annotate any text or data for any biomedical topic on the fly. That is how Vitae Evidence can map the search strategy to the relevant pieces of knowledge from the aggregated references and stay current. It is also not limited to a specific diagnosis.

All insights are quality controlled by our community of Subject Matter Experts. So the more it's used, the better it gets, for everyone. Our algorithms run continuously. Once a case is configured on Vitae Evidence, Subject Matter Experts get smart notifications when new personalized knowledge is discovered.

Rule-based reasoning
Vitae Evidence enables subject matter experts to define rules like: 'IF x happens, THEN do Y', and organises them in complex decision trees, to appraise evidence according to Evidence-Based Medicine methodology and Real-World Research princples.

Case-based reasoning and Similarity Maps
Using data points from the patient summary, Vitae Evidence helps determine search strategies and rules for a "case like this". It also groups the previous reports from similar cases to find patterns and suggests possible treatment options.

Machine learning and natural language processing
Natural language processing (NLP) annotators chain extract relevant elements from large text corpus and datasets, leveraging gold-standard Machine Learning libraries. Curated patterns, controlled vocabularies and ontologies are used to improve the results confidence.

Feedback loop
The artificial intelligence of Vitae Evidence learns from direct feedback from users, when they revise the search terms, curate the references annotated by AI, create new rules to filter references, and confirm insights.
It will also learn from patient-reported outcomes (ePRO) and the recorded clinician treatment decisions.

Vitae Evidence cannot use "common sense", adapt on the fly, or understand cause and effects. Like all AI, it also cannot reason ethically. That is why human expertise is always going to be necessary.

Cleaning the input data, making sure that it is fit for purpose is one important step that requires human experts and is necessary to ensure the quality of the output.

Recurring benchmark of the algorithm is also an important manual effort that must be performed by human experts to continuously monitor the quality of the output, and initiate AI modifications if necessary.

The more Vitae Evidence is used, the more rules and knowledge it will contain for future reasoning, so the amount of manual configuration will decrease.