Artificial Intelligence in Healthcare

A Brief Overview of AI’s Role in Modern Healthcare

From a Distant Promise to Clinical Reality

 In just a decade, artificial intelligence (AI) has gone from being a niche topic among technologists to becoming a daily support tool for healthcare professionals. This shift has been made possible thanks to the abundance of clinical data, advances in computing power, and the maturity of deep learning models.
According to the European Commission, medical AI “has become a tangible reality.” Healthcare systems are increasingly integrating AI-based tools to optimize resources, improve diagnostic accuracy, and streamline administrative processes in hospitals. This helps reduce consultation overload and allows more time for meaningful doctor-patient interactions.

Beyond its use in radiology, dermatology, and personalized medicine, AI is transforming how we detect diseases like cancer, diabetes, and Alzheimer’s. Algorithms trained on millions of medical images are assisting clinicians in identifying diseases at earlier stages.
The adoption of AI-powered virtual assistants is also alleviating administrative burdens by automating medical record management, appointment scheduling, and chronic patient monitoring. In emergency care, AI systems enable fast and efficient triage, prioritizing patients based on the severity of their conditions.
Despite this progress, key challenges remain—such as data privacy, algorithm transparency (explainability), and the need for clear regulatory frameworks. However, with proper collaboration between medical, technological, and legislative institutions, AI is poised to continue evolving as a critical tool for more precise, efficient, and patient-centered care.

Medical Fields Where AI Is Making a Difference

Artificial intelligence is already leaving a significant mark across various medical specialties, with real-world applications delivering tangible benefits:
Radiology and medical imaging: AI systems are improving diagnostics in mammography and chest CT scans. These tools enable earlier and more accurate detection of pathologies while also reducing patient exposure to radiation. Progress in this field has been showcased in forums like [RSNA] and publications such as [IEEE].

Ophthalmology: Tools like [IDx-DR] and the [Tartaglia Project] have proven highly effective in detecting diabetic retinopathy, allowing for large-scale screenings in primary care. Early detection is crucial to prevent irreversible blindness, especially among poorly controlled diabetic populations.
Digestive endoscopy: AI-powered systems have enhanced polyp detection during gastroscopies. As reported in journals like [ACTA], this technology significantly contributes to colorectal cancer prevention by increasing the rate of clinically relevant findings during screenings.

Mental health: Innovations such as Therabot—a chatbot-assisted robot—interact with patients to help alleviate symptoms of anxiety and depression. While not a replacement for therapists, tools like this offer 24/7 complementary and accessible support.

Orthodontics: According to studies indexed in [PubMed], AI-based automation of cephalometric analysis has accelerated treatment planning, reducing the time needed for this process by up to 90%.
Neonatal intensive care: An AI algorithm has been developed to optimize parenteral nutrition in premature newborns. As highlighted by Nature, this advancement not only improves neonatal health and safety but also enhances clinical resource management and reduces hospital costs.

The Regulatory Framework for Medical AI in Europe

The implementation of the AI Act (effective August 2024) establishes a regulatory framework for medical AI in Europe, classifying it as “high-risk AI.”
To be approved in the EU, a medical AI solution must meet the following requirements:

• High-quality datasets with demographic representation and fairness considerations
• Medical data protected under data privacy laws
• Risk management protocols
• Traceability of processes and outcomes
• Human oversight in critical decisions
• Continuous retraining with local data to avoid temporal drift
• External evaluations and audits

Spain has led the way by launching the Spanish Agency for AI Supervision (AESIA), active since February 2025, which oversees compliance with the AI Act at the national level.

Medical AI in Spain

 Below are some notable projects developed by Spanish hospitals and institutions, within a broader national landscape full of diverse AI initiatives:

• Hospital Universitario de Navarra: Implemented NaIA-RD, an AI tool for diabetic retinopathy screening, which has improved diagnostic sensitivity and reduced professionals’ workloads. [arXiv].
• Hospital Vall d’Hebron (Barcelona): Developed the predictive tool ‘Promise Score’ to estimate 90-day mortality in cancer patients admitted to emergency care, based on clinical data and lab results.
• Computational Medicine Platform – Fundación Progreso y Salud, Andalusian Public Health System: Uses biomedical data to develop personalized treatments, analyze genomes for hereditary disease prevention, and enhance epidemiological surveillance.
•  Hospital Universitario Ramón y Cajal (Madrid): Developing a tool to identify hereditary cancer risks through genetic predisposition analysis based on family data.
• Emily (Barcelona): A solution by Aether Tech for automatic regulation of oxygen flow in patients with respiratory diseases.
  The Blue Box (Reus): Developed by biomedical engineer Judit Giró, this device detects early-stage breast cancer using urine samples.
•  QP-Prostate (Valencia): A computer-aided diagnosis system for prostate MRI developed by Quibim, aimed at identifying and diagnosing prostate cancer lesions.

Our Contribution: Tartaglia

have collaborated closely with institutions like Fundación Rioja Salud, Instituto de Investigación Sanitaria La Fe, and the Galician Health Service (Sergas).
Tartaglia’s goal is to accelerate clinical research for the early detection of conditions such as diabetic retinopathy, prostate cancer, Alzheimer’s, and cardiovascular disease. One of its main innovations is the creation of a federated data network, which allows AI models to be trained locally without transferring medical images between institutions—thus ensuring patient data privacy.
At Pixelabs, we’ve been responsible for designing and training federated algorithms using distributed data to automate the screening of diabetic retinopathy. By analyzing retinal images with computer vision, the system distinguishes between those showing signs of disease and those that do not. It then aggregates relevant features for diagnosis or further processing. You can read all the details here.

References:

1. European Commission. Artificial Intelligence in Healthcare (2024).
2. van Kolfschooten H. The EU Artificial Intelligence Act: Implications for Healthcare. Health Policy 149:105152 (2024).
3. Retina Specialist. AI for DR Screening: Where are we in 2025? (2025).
4. Park A. Google’s AI Will Now Be Used in Mammograms. TIME (2022).
5. Phongpreecha T. AI-guided precision parenteral nutrition for NICU. Nature Medicine 31:824-836 (2025).
6. Ortiz O et al. AI-assisted colonoscopy in Lynch syndrome (TIMELY). Lancet Gastroenterol Hepatol 9:802-810 (2024).
7. Jacobson N et al. Randomized Trial of a Generative AI Chatbot for Mental Health Treatment. NEJM AI (2025).
8. Almeida C et al. AI in Orthodontics: Revolutionizing Diagnostics and Treatment. J Clin Med 13:344 (2024).
9. Comisión Europea & Parlamento Europeo. Reglamento (UE) 2024/1685 “AI Act”.
10. AESIA. Garantizando una IA ética y responsable (2025)
11. Kohane I et al. Bias in Medical AI: Implications for Clinical Decision Making.J Am Med Inform Assoc 32:e155-e163 (2024).