Medicine and Innovation: the key role of AI in enhancing healthcare 

In today’s technological landscape, few innovations have been as powerful and transformative as artificial intelligence (AI).

Medicine is undoubtedly one of the fields where AI is making the most significant impact.

The adoption of AI in medicine offers a wide range of applications, each of which can be tailored and adapted to meet the specific needs of various medical disciplines and processes.

AI is redefining how doctors and researchers approach care, diagnosis, and patient assistance, as well as how people access more effective treatments and healthcare pathways.

What practical solutions does AI offer to advance healthcare processes?

Data is at the heart of AI.

The nature of the healthcare industry — where research, analysis, and data interpretation are crucial for delivering optimal results — makes AI is an unparalleled tool.
The ability to process and interpret large volumes of data with precision and speed opens new opportunities for enhancing diagnoses, personalizing treatments, and even predicting future healthcare needs.

Whether it’s about collecting and leveraging data, using it for predictions, or turning it into actionable insights in real-time, data is the foundational element around which innovative products and solutions revolve.

Below, we’ll explore how Almawave manages and maximizes data at every stage: data processing, expansion, evaluation, and transformation.

In healthcare, patient data confidentiality and protection are critically important. At the same time, using that data to support medical analyses and decisions remains essential. Almawave’s solutions aim to balance these aspects with complete security.

• Pseudonymization and Security

Almawave’s data anonymization and pseudonymization solution ensures full compliance with privacy regulations while allowing reliable and comprehensive use of clinical data for research and development. Pseudonymization replaces real names with pseudonyms in a fully automated and robust manner.
For medical reports written in natural language, for example, the solution anonymizes and analyzes these reports, making them available for further processing.
Using a Named Entity Recognition (NER) model based on BERT, the tool identifies patient names and other sensitive information and automatically replaces them with pseudonyms generated by the system. This facilitates the reconciliation of patient journey data and ensures the ongoing applicability of natural language processing (NLP) systems.

• Data Federation and Segregation

Data federation allows for the combination and analysis of information from various sources without the need to centralize the data into a single environment, thus keeping the data separated.
Similarly, federated AI enables algorithms to be trained across multiple decentralized devices or servers that hold local data, without transferring this data to a central server.
This approach is particularly valuable in the medical field, where data may be distributed across different institutions and systems, ensuring that sensitive data remains fully protected.

To enhance the reliability and completeness of the responses provided by artificial intelligence, it is crucial to expand and diversify the data sources used. Below are two innovative solutions that address this challenge.

• Patient Survey

The Patient Survey solution is a tool designed to automate the procedures for surveys and questionnaires related to patient-reported information while structuring this data. This enables:

1. Expanding the raw and structured data pool.
2. Automating the administration and management of Patient-Reported Outcome Measures (PROMs) and Patient-Reported Experience Measures (PREMs) through voice channels.
3. Automated navigation and filtering of the collected data.
4. Creating predictive algorithms and tools that consider aspects such as Quality of Life (QoL) and a patient-centric perspective.

• Data Augmentation from Unstructured Sources

In healthcare, much of the information is inherently unstructured. Medical reports, patient history, family background, and therapy details are based on evaluations where the language is unstructured. Making this information easily accessible is a significant benefit for clinicians and can also support potential further analysis.
Natural Language Processing (NLP) in AI encompasses the techniques and technologies focused on managing natural language. Recently, this field has experienced a significant technological leap.

Large Language Models (LLMs) have introduced what is known as generative AI—a powerful tool for understanding and generating language. However, LLMs have some limitations, especially in critical or sensitive areas. They do not always provide reliable or verified answers because they often cannot integrate information from sources beyond their training data.
Almawave’s “composite” approach combines various technologies and methodologies to maximize results while ensuring response security. One of the tools used to overcome these limitations is Retrieval Augmented Generation (RAG), which allows access to multiple additional data sources, making responses more comprehensive and addressing the constraints of technologies based solely on language models.

Accurate evaluation of clinical data is essential for improving diagnoses and treatments. Almawave’s solutions leverage advanced artificial intelligence techniques to extract, organize, and analyze clinical information, enabling more effective and integrated data management.

• Broadening and Expanding Data Sources

Almawave’s solutions use Natural Language Processing (NLP) techniques to structure and classify data within clinical documents. The goal is to extract useful information from reports by categorizing the relevant details contained within them. This approach facilitates easier consultation and integration of clinical information, thereby enhancing the accuracy and effectiveness of diagnoses and treatments.

For example, the Radiology Report Data Extraction System uses Named Entity Recognition (NER) to extract crucial information from radiology reports, including patient demographics, medical history, clinical observations from instrumental tests, examination results, and clinician notes.
This system automatically integrates this information with standard nomenclature tools (such as ICD, SNOMED, LOINC), facilitating understanding and interoperability both between different tools and among various professionals.

• CDSS and Knowledge Graph for Identifying Correlations and Pathways

The Clinical Decision Support System (CDSS) is an AI solution that uses standard hospital data and patient journey information to create and present predictive algorithms for clinical decision support. It aids in planning care and therapy pathways through data-driven tools powered by AI technology.
The CDSS includes a clinical viewer, a component that allows users to access and explore all available data and information. This includes data from hospital sources such as reports, demographic and clinical data, as well as synthetic data, like previously discussed indicators.
The entire data set is structured as a knowledge graph. Information is represented through nodes and edges, standardized according to the MeSH ontology, ensuring that the data is consistent. The graph facilitates data reading and evaluation by clearly visualizing the relationships and connections between different pieces of information, enabling a more intuitive and immediate analysis of clinical correlations.

• LLM for Patient Summarization

Within the CDSS, Almawave’s NLP technologies, when applied to various data sources, leverage a foundational language model (LLM) to organize information into a usable clinical summary for healthcare providers. This tool is also accessible remotely, helping clinicians save time by supporting decision-making processes, enhancing clinical accuracy, and improving understanding of the patient’s history.
The patient summary is especially valuable in clinical boards, as it provides a clear and concise view of relevant information, facilitating multidisciplinary discussions and care planning

The final and crucial step that AI can enable in the medical field is transforming data into actionable insights that can guide personalized solutions based on patient needs. Almawave’s solutions include real-time KPI monitoring, predictive analytics, and more accurate patient status summaries to drive this process.

• Innovative Metrics for a New Data-Driven Approach to Care Pathways

Artificial intelligence applied to big data enables the development of complex new indicators, such as the Clinical Stability Index, to assist doctors in providing the most appropriate therapies for individual patients in real-time. This approach synthesizes a wide range of information into a predictive tool that is fast, easy to understand, and user-friendly for clinicians.
A clear demonstration of this is Almawave’s RicovAI project. RicovAI is an AI-based system designed for home monitoring of patients affected by SARS-CoV-2. The solution aims to predictively stratify the risk of patient deterioration and care needs through the use of AI-powered tools.

In this context, the KPI’s features are:

1. Prognostic: The stability indicator forecasts future trends.
2. Explainable and Transparent: The indicator details the variables that contribute to a given prediction, allowing clinicians to verify and structure an intervention based on those variables (augmented decision-making).
3. Synthetic: The indicator becomes more comprehensive with more data, given the same resource requirements (one-to-many approach).
4. Stratification: It enables the customization of care pathways and actions specific to each patient (personalized care).

• Explainability

A crucial aspect of adopting artificial intelligence in medicine is explainability—the ability to understand and interpret the results generated by AI algorithms.
Almawave’s solutions aim not only to provide accurate predictions in the medical field but also to ensure transparent decision-making processes. This is demonstrated by knowledge graphs, which visualize the interrelationships between various clinical data, offering a clear and intuitive representation of correlations and diagnostic and therapeutic pathways.
Explainability is achieved by linking a prediction to the underlying variables and sources. This approach allows for both verification and understanding of the recommendation, as well as targeting and personalizing interventions based on specific variables with causally related therapies or actions.

• Extended Holistic Predictions

Training artificial intelligence systems on expanded datasets that include broader patient information allows doctors to access summaries and predictions that go beyond isolated clinical data, enabling a more holistic approach to proposed treatments. The goal is to enhance not only short-term therapies but also long-term quality of life.
For instance, the previously mentioned RicovAI project is a model that could easily be applied to various areas of healthcare in the near future.

The OncologIA project illustrates this approach perfectly. OncologIA represents an advanced diagnostic system based on artificial intelligence technologies and tools such as Digital Twins (Digital Health Identities). This system is designed to serve regional oncology networks and support medical personnel in identifying personalized therapies for patients, predicting potential complications, and determining the optimal follow-up strategy based on available local services.
OncologIA leverages a digital twin of the patient, enabling doctors and caregivers to continuously access a comprehensive and integrated view of the patient’s health status. The main tools and activities within the project include:

• Data Enrichment: Using NLP techniques on unstructured data sources to simplify information access and expand the knowledge base for further processing.
• Natural Language Tools: Collecting patient-reported data and structuring it for analysis through conversational Patient-Reported Outcome Measures (PROMs).
• Clinical Decision Support System (CDSS): Based on diverse data sources, this system assists clinicians by providing synthetic indicators that facilitate comparison and adherence to protocols and guidelines.

Interested in learning more about Almawave’s solutions for the healthcare sector?