Synthetic Dataset of Structured and Free-Text Primary Staging Rectal Cancer MRI Reports

This synthetic dataset in CSV format contains 1,000 paired synthetic primary-staging rectal MRI reports. Each row includes a JSON-formatted structured_report and its corresponding free_text_report. The reports describe primary rectal cancer staging features, including tumour morphology and location, T and N stage, anterior peritoneal reflection relationship, sphincter and adjacent-structure involvement, mesorectal fascia status, lymph nodes, tumour deposits, and extramural vascular invasion Structured cases were generated using reproducible rule-based sampling following anatomically plasible rules. Free-text narratives were produced using large language models and subsequently assessed for correspondence with the structured data. A radiologist reviewed the anatomical generation rules and a sample of 20 structured/free-text pairs, providing feedback on clinical plausibility, terminology, staging, and reporting style. This review does not constitute validation of every record. The dataset is intended for research in radiology NLP, report generation, information extraction, and structured-to-free-text alignment. All cases are entirely synthetic and contain no patient data. The dataset is not intended for clinical decision-making. Example: structured_report free_text_report {"local_tumor_status": {"morphology": "semi-annular","mucinous_component": "no","circumferential_tumor_involvement_start": 4,"circumferential_tumor_involvement_end": 10,"rectal_tumor_location_in_rectum": "mid","distance_from_anorectal_junction_to_distal_tumor_border_cm": 5.0,"relation_to_anterior_peritoneal_reflection": "below","tumor_length_cm": 2.9,"t_stage": "T3b","invaded_structures": "","sphincter_complex_invasion": "no","lowest_part_of_sphincter_invasion": "not applicable"},"mesorectal_fascia_involement": {"margin": "clear(> 1 mm)","circumferential_location_of_shortest_distance": null,"involved_by": "not applicable"},"lymph_nodes_and_tumor_deposits": {"regional_total_suspicious_mesorectal_nodes": null,"nodes_short_axis_ge_9mm": null,"nodes_from_5_to_8mm_with_2_morphologic_criteria": null,"nodes_5mm_with_all_3_morphologic_criteria": null,"mucinous_nodes_any_size": null,"regional_total_suspicious_extramesorectal_nodes": 1,"n_stage": "N1","tumor_deposits_within_mesorectum": null},"emvi": {"presence_of_emvi": "yes","emvi_clock_position": 8}} MRI of the Rectum Technique: Multiplanar T2-weighted images of the rectum were obtained, supplemented with T1-weighted and diffusion-weighted sequences. Findings:There is a semi-annular tumor of the mid rectum, extending from 4 to 10 o’clock, with a craniocaudal length of approximately 2.9 cm. The lower margin of the lesion is situated 5.0 cm above the anorectal junction, and the tumor lies below the anterior peritoneal reflection. No evidence of mucinous component is seen. Imaging demonstrates transmural invasion with spiculated extension into the mesorectal fat, consistent with T3b staging. The mesorectal fascia is clear, with a minimum distance of 3.4 mm from the tumor. Extramural vascular invasion is present, identified at the 8 o’clock position. There is one suspicious extramesorectal lymph node in the internal iliac region. Impression:Mid-rectal semi-annular tumor (T3b) with extramural vascular invasion (EMVI+), clear mesorectal fascia (MRF-), and one suspicious internal iliac lymph node (N1). Findings should be interpreted in conjunction with the patient’s clinical picture.