Case Study

IEO innovates mammographic screening with agent-based systems powered by artificial intelligence

Laife Reply collaborates with the European Institute of Oncology to enhance mammography screening through AI agents that support faster, more accurate, and patient-oriented diagnoses.

#ArtificialIntelligence
#DeepLearning
#GenerativeAI
#AIAgents
#DigitalHealth

The Challenge

Optimise breast cancer screening to ensure timely and accurate diagnoses, reducing wait times and clinical uncertainty for patients.

SCENARIO

Technology and clinical accuracy at the service of early diagnosis

Breast cancer screening is a crucial tool for early diagnosis and timely initiation of treatments, with a direct impact on patient survival rates. However, the current model presents several critical issues: prolonged waiting times, significant workloads for radiologists, and delays in communicating results that create bottlenecks in diagnostic pathways and delays in clinical interventions.

The European Institute of Oncology (IEO) is an international benchmark in the field of oncological healthcare, where innovation and research play a central role. To effectively address the challenges related to mammographic screening, institutions need advanced solutions capable of optimizing workflows, increasing diagnostic accuracy, and ensuring timely follow-up. Fundamental objectives to improve both clinical outcomes and system efficiency.

THE SOLUTION

X-RAIS, the AI-based agent screening system

In a path of continuous evolution and search for efficiency in tools for the early diagnosis of breast cancer, IEO and Laife Reply consolidate their collaboration contributing to the evolution of X-RAIS, the artificial intelligence platform developed by Laife Reply. Based on neural networks and radiomics techniques, X-RAIS integrates an agentic system designed to support the analysis of mammograms in real-time, accurately identifying suspicious lesions and classifying them as benign or malignant. Fully integrated into the workflow, the system provides immediate support to radiologists, optimising time and resources during screening: a first AI agent, dedicated to generating reports, uses these outputs, possibly enriched by the patient's clinical history through tool calling, to create preliminary structured clinical reports, intended for the radiologist's review. In parallel, a second agent handles prioritisation, assigning scores to cases based on urgency and clinical context, to build an optimised review pipeline for reports.
The central element of the model is the “human-in-the-loop” approach, which keeps radiologists at the center of the decision-making process. Every output generated by the artificial intelligence is subjected to clinical validation, ensuring high standards of accuracy and safety.

HOW WE DID IT

Multi-agent orchestration with the integration of generative AI

Laife Reply has developed a modular agent architecture built in Python, based on advanced generative artificial intelligence models, including GPT‑4.1, and on retrieval augmented generation (RAG) techniques. The architecture is designed to support two main activities in the radiological field:

Automatic generation of preliminary reports

Intelligent case prioritisation

A second agent leverages generative AI models to assign an urgency score to clinical cases and generate an optimised priority list, assisting radiologists in the immediate identification of the most critical cases.
At the heart of the solution is the X-RAIS system, which utilises convolutional neural networks (CNN) to analyse mammographic images, producing bounding boxes for lesions, classifications (benign or malignant), and corresponding confidence scores.
The entire system is integrable via clinical APIs compliant with interoperability standards (FHIR), ensuring full compatibility with existing healthcare infrastructures.

Efficiency and precision thanks to artificial intelligence

The integration of X-RAIS into the clinical workflow of IEO demonstrates how agent-based artificial intelligence can translate into tangible benefits, improving diagnostic accuracy, optimizing resources, and contributing to a more responsive healthcare system.

Increase in capacity
for screening

Radiologists can manage up to double the exams per shift, increasing from 5,000 to 10,000 annual mammograms per specialist.

Significant reduction
in waiting times

The average recall time in critical cases can be reduced from about 28 days to just 6 hours, accelerating clinical intervention and improving the effectiveness of early actions.

Greater attention
to complex cases

AI supports the preliminary generation of reports and prioritisation, allowing radiologists to focus on the most critical and clinically uncertain cases, with a positive impact on diagnostic accuracy and safety.

IEO - European Institute of Oncology

The IEO European Institute of Oncology is a Scientific Institute for Hospitalization and Care (IRCCS) dedicated to oncology. Within it, there is a complete integration of the various areas of the fight against cancer: laboratory research, clinical research, prevention, diagnosis, treatment, and training. The Institute has implemented a new principle in cancer care by shifting the focus from the disease to the patient. Thanks to the immediate transfer of new results from research to clinical practice, it has abandoned traditional therapies based on the maximum intervention that the patient can tolerate, to apply innovative methods that, while maintaining effectiveness, ensure minimal harm to the person.

Laife Reply

Laife Reply promotes digital innovation in the Healthcare sector by leveraging its high expertise in Artificial Intelligence and Big Data Analytics technologies and a deep know-how in the medical field. Laife Reply specializes in the application of Artificial Intelligence algorithms to clinical images and data, aiming to add real value to healthcare processes and contribute to the reduction of clinical risk through support for diagnosis and treatment activities.