Objectives: to assess the ability of high-performing open-weight large language models (LLMs) in extracting key radiological features from prostate MRI reports. Methods: Five LLMs (Llama3.3, DeepSeek-R1-Llama3.3, Phi4, Gemma-2, and Qwen2.5-14B) were used to analyze free-text MRI reports retrieved from clinical practice. Each LLM processed reports three times using specialized prompts to extract (1) dimensions, (2) volume and PSA density, and (3) lesion characteristics. An experienced radiologist manually annotated the dataset, defining entities (Exam) and sub-entities (Lesion, Dimension). Feature- and physician-level performance were then assessed. Results: 250 MRI exams reported by 7 radiologists were analyzed by the LLMs. Feature-level performances showed that DeepSeek-R1-Llama3.3 exhibited the highest average score (98.6% ± 2.1%), followed by Phi4 (98.1% ± 2.2%), Llama3.3 (98.0% ± 3.0%), Qwen2.5 (97.5% ± 3.9%), and Gemma2 (96.0% ± 3.4%). All models excelled in extracting PSA density (100%) and volume (≥98.4%), while lesions’ extraction showed greater variability (88.4–94.0%). LLMs’ performance varied among radiologists: Physician B’s reports yielded the highest mean score (99.9% ± 0.2%), while Physician C’s resulted in the lowest (94.4% ± 2.3%). Conclusions: LLMs showed promising results in automated feature-extraction from radiology reports, with DeepSeek-R1-Llama3.3 achieving the highest overall score. These models can improve clinical workflows by structuring unstructured medical text. However, a preliminary analysis of reporting styles is necessary to identify potential challenges and optimize prompt design to better align with individual physician reporting styles. This approach can further enhance the robustness and adaptability of LLM-driven clinical data extraction.

Prostate cancer (PCa) is one of the most common tumors diagnosed in men worldwide, with approximately 1.7 million new cases expected by 2030. Most cancerous lesions in PCa are located in the peripheral zone (PZ); therefore, accurate identification of the location of the lesion is essential for effective diagnosis and treatment. Zonal segmentation in magnetic resonance imaging (MRI) scans is critical and plays a key role in pinpointing cancerous regions and treatment strategies. In this work, we report on the development of three advanced neural network-based models: one based on ensemble learning, one on Meta-Net, and one on YOLO-V8. They were tailored for the segmentation of the central gland (CG) and PZ using a small dataset of 90 MRI scans for training, 25 MRIs for validation, and 24 scans for testing. The ensemble learning method, combining U-Net-based models (Attention-Res-U-Net, Vanilla-Net, and V-Net), achieved an IoU of 79.3% and DSC of 88.4% for CG and an IoU of 54.5% and DSC of 70.5% for PZ on the test set. Meta-Net, used for the first time in segmentation, demonstrated an IoU of 78% and DSC of 88% for CG, while YOLO-V8 outperformed both models with an IoU of 80% and DSC of 89% for CG and an IoU of 58% and DSC of 73% for PZ.

Objectives: Whole-body MRI is increasingly used for preventive health screening; however, the prevalence and distribution of incidental oncologically relevant findings in asymptomatic individuals remain incompletely characterized. The primary objective of this multicentre study was to describe the frequency and anatomical distribution of clinically relevant findings detected by whole-body MRI in an asymptomatic adult population. As a secondary objective, we evaluated the feasibility of using a large language model (LLM) for retrospective extraction and structuring of radiology report data for research purposes. Methods: Radiology reports from 327 asymptomatic adults who underwent MRI as part of a preventive screening program were retrospectively collected from four diagnostic centers. All MRI examinations were interpreted by subspecialist radiologists (neuroradiology, body imaging, and musculoskeletal imaging), and clinical findings were classified using the Oncologically Relevant Findings Reporting and Data System (ONCO-RADS). A senior radiologist analyzed the reports to extract only findings with ONCO-RADS values ≥3, creating a reference standard. In addition, an LLM (DeepSeek-R1-Llama3.3, Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., Hangzhou, Zhejiang, China) was subsequently applied to automatically extract all findings from free-text reports with associated ONCO-RADS categories (including also ONCO-RADS 1 and 2) and anatomical locations. LLM outputs were compared with the radiologist-defined ONCO-RADS to assess the performance. Results: Among the 327 individuals (213 males, 114 females; median age 52 years), a total of 237 findings (5%) classified as ONCO-RADS ≥ 3 were extracted by the radiologist, affecting 138 individuals (42.2%). The majority of ONCO-RADS ≥ 3 findings were ONCO-RADS 3 (232, 97.9%). Three findings (1.3%) were ONCO-RADS 4, and two (0.8%) were ONCO-RADS 5. Three malignant lesions were confirmed (prostate cancer, renal cell carcinoma, and appendiceal carcinoma), corresponding to a cancer prevalence of 0.9% in the screened population. Regarding the performance of the LLM in extracting and structuring radiologist-reported findings, 207 (87.3%) ONCO-RADS were correctly extracted, 17 (7.2%) were missed, and 13 (5.5%) were incorrectly localized. Conclusion: Whole-body MRI detected a small number of confirmed occult malignancies and a higher prevalence of oncologically relevant findings (ONCO-RADS ≥ 3). Regarding the LLM, results showed that it can support efficient and standardized retrospective extraction of structured information from heterogeneous radiology reports.

Primary prevention is essential to reduce disease burden before clinical onset, yet it remains less systematically integrated into care than diagnosis and treatment. Although artificial intelligence (AI) is increasingly used in medicine, most applications have focused on secondary and tertiary prevention, including diagnosis, prognostic stratification, and disease management, while its role in primary prevention remains less defined. This narrative review examines current AI applications across four modifiable lifestyle domains relevant to prevention and healthspan promotion: nutrition, physical activity, sleep, and mental health. We synthesize evidence on machine-learning models, wearable-derived algorithms, computer-vision tools, just-in-time adaptive interventions, and conversational agents used in consumer, community, and hybrid clinical–digital settings. AI applications support postprandial glycemic prediction, automated dietary assessment, meal-planning adherence, sedentary-pattern detection, personalized exercise recommendations, adaptive behavioral nudges, sleep monitoring, circadian-aware recommendations, psychoeducation, stress-management support, and early identification of psychological vulnerability. Collectively, these tools may extend prevention beyond episodic clinical encounters toward continuous, personalized, and context-aware support. However, evidence remains limited by short follow-up, reliance on surrogate or engagement outcomes, digitally literate populations, and insufficient validation in real-world preventive-care pathways. AI is therefore a promising enabling technology for proactive, healthspan-oriented medicine, provided future studies demonstrate long-term effectiveness, equity, safety, and responsible implementation.