Real-time fidelity assessment of fluorescence molecular imaging without reference images

Fluorescence molecular imaging (FMI) has been approved for surgical guidance in several procedures. However, its interpretation still requires substantial clinician expertise due to variable signal levels. For FMI to gain wider clinical acceptance, image quality should be assessed automatically, thereby prompting clinicians to retake low-quality images when necessary. Yet, no user-friendly fidelity assessment tools currently address this need.

Herein, we introduce a real-time, frame-by-frame FMI fidelity assessment method, termed no-reference fidelity assessment of fluorescence molecular imaging (NRFA-FMI).

NRFA-FMI quantifies image fidelity using a combination of histogram-based and textural metrics. A central advantage of NRFA-FMI is that it operates without reference images, making it particularly suitable for clinical FMI workflows.

We show that NRFA-FMI outperforms two state-of-the-art no-reference image fidelity assessment methods, the blind/referenceless image spatial quality evaluator (BRISQUE) and the natural image quality evaluator (NIQE).

The introduction of NRFA-FMI enables real-time FMI fidelity assessment, an essential step toward reducing data misinterpretation and preventing incorrect diagnosis. This approach provides clinicians with an intuitive, automated indicator of image reliability, supporting more consistent and accurate intraoperative decision-making.