AI Integration for Sectra Breast Imaging

AI integration for Sectra Breast Imaging connects at three primary workflow surfaces: the reading worklist, the diagnostic viewer, and the structured reporting module. The integration uses Sectra's APIs and DICOMweb services to inject AI-derived observations—such as lesion probability scores, BI-RADS density assessments, and prior comparison flags—directly into the radiologist's native environment. This allows for AI-triggered worklist prioritization, where suspicious cases from screening tomosynthesis are flagged for expedited review, and context-aware decision support during the diagnostic review of MRI and ultrasound studies.

Implementation typically involves a secure, containerized inference service that listens to the Sectra PACS broker for new DICOM studies tagged with breast imaging modalities. Upon ingestion, the service processes the images, runs configured AI algorithms (e.g., for mass detection, calcification clustering, or parenchymal texture analysis), and returns structured results as DICOM SR or via a REST API. These results are then mapped to Sectra's hanging protocols to display AI markers as overlays and populate relevant fields in the Sectra Mammography Reporting template, reducing manual data entry and standardizing report language.

Rollout requires careful governance, starting with a silent mode where AI inferences are logged but not displayed, allowing for validation against historical outcomes. A phased activation follows, often beginning with density assessment AI to automate a repetitive task, then progressing to lesion tracking AI for MRI follow-ups. Critical to success is configuring the system's audit trail to log every AI interaction and building a feedback loop where radiologists can confirm, reject, or modify AI suggestions, continuously improving model performance and clinical trust. For a broader view of integrating AI across the entire imaging enterprise, see our guide on AI Integration for Sectra Enterprise Imaging.

Integration begins at the modality or PACS router, where new breast imaging studies (Digital Breast Tomosynthesis, MRI, Ultrasound) are ingested. Using Sectra's DICOM Service Class Provider (SCP) and HL7 interfaces, studies meeting pre-defined criteria—such as a screening mammogram or a diagnostic MRI for high-risk patients—are automatically routed to a secure, HIPAA-compliant AI inference queue. This decoupled architecture ensures the core PACS performance is unaffected. The AI service, hosted within your health system's private cloud or a compliant Inference Systems environment, pulls anonymized studies via DICOMweb WADO-RS, processes them through specialized models (e.g., for lesion detection, density assessment, or prior comparison), and returns results as a DICOM Structured Report (SR) or a JSON payload via a RESTful API.

The returned AI findings are then injected back into the Sectra workflow at multiple, clinically relevant points. For the radiologist, key integrations include:

• Worklist Prioritization: The SR is parsed, and studies with high-probability findings or significant changes from prior exams are flagged and elevated in the Sectra Breast Imaging PACS worklist.
• Hanging Protocol & Viewer Overlay: Quantitative results and lesion markers are rendered as graphical overlays directly on the images within the Sectra viewer, following its hanging protocols for multimodal review (e.g., correlating a DBT finding with an ultrasound).
• Reporting Support: Key measurements, BI-RADS assessments, and differential suggestions are pre-populated into the Sectra Reporting module or sent via HL7 to your speech recognition system, creating a draft report for radiologist verification and editing.

Governance and rollout require a phased, monitored approach. We recommend starting with a silent mode where AI runs in the background, and results are logged but not displayed, to validate performance and build radiologist trust. A human-in-the-loop approval step is configured within Sectra's workflow manager before any AI finding becomes part of the permanent record. All data flows, AI inferences, and user interactions are logged to a dedicated audit database for model performance monitoring, drift detection, and compliance reporting. This controlled integration pattern, managed via Inference Systems' orchestration layer, allows for incremental rollout—from a single screening mammography AI to a full suite of multimodal breast imaging algorithms—without disrupting existing clinical operations.

Integrating AI into a diagnostic breast imaging workflow requires a governance-first approach. This starts with a secure data pipeline where DICOM studies from Sectra's breast PACS (tomosynthesis, MRI, ultrasound) are de-identified for AI inference, with results returned as DICOM Structured Reports (SR) or via a dedicated API. All data flows must be encrypted in transit and at rest, with strict access controls tied to Sectra's existing RBAC for radiologists and technologists. AI-generated findings should be stored as non-destructive overlays or linked annotations within the Sectra study, creating a full audit trail of the AI's input, model version, and the radiologist's final interpretation for compliance and quality assurance.

A phased rollout is critical for clinical adoption and risk management. Phase 1 (Silent Mode) runs AI algorithms in the background on a subset of studies, comparing AI outputs to final reports without displaying results to radiologists, establishing baseline performance and building trust. Phase 2 (Assistive Mode) introduces AI findings as a secondary, clearly labeled finding list or overlay in the Sectra viewer for non-priority cases, allowing radiologists to accept, modify, or reject suggestions. Phase 3 (Integrated Workflow) expands to priority workflows like screening recalls or biopsy planning, using AI to auto-populate structured report templates in Sectra Reporting and potentially triage urgent cases to the top of the worklist. Each phase requires defined metrics for accuracy, turnaround time, and user feedback before proceeding.

Ongoing governance involves a multidisciplinary oversight committee (radiologists, IT, compliance, AI engineers) to review model drift, clinical feedback, and adverse event reports. AI model updates should follow a validated change control process, similar to medical device software. Furthermore, integration with existing quality and peer review workflows in Sectra is essential, allowing AI performance to be monitored as part of the department's continuous quality improvement program. For a deeper technical dive on connecting AI to the core PACS workflow, see our guide on AI Integration for Sectra PACS.