A metaheuristics-equipped post-processing model for coronary angiograms

Contrast-limited adaptive histogram equalization (CLAHE) is a post-processing algorithm used for improving the quality of coronary angiograms. Proper tuning of the clip limit (CL) is crucial to fetch high-quality and distortion-free outputs from the CLAHE.

We propose a parameter optimization model to automate the selection of CL in the CLAHE deployed for post-processing the coronary angiograms. The grey wolf optimization (GWO) and patch-based contrast quality index (PCQI) fitness are embedded in the model. We compared the optimum CLAHE (OCLAHE) with state-of-the-art schemes for improving the quality of the angiograms, namely histogram equalization (HE), unsharp masking (USM), morphological filtering (MF), and Frangi filter (FF), in terms of subjective visual appeal of the enhanced angiograms and objective quality measured via image quality assessment (IQA) indices. We used four IQA indices, namely PCQI, quality-aware relative contrast measure (QRCM), entropy, and blind/referenceless image spatial quality evaluator (BRISQUE), to objectively compare the quality of OCLAHE outputs with that of the state-of-the-art schemes.

The OCLAHE produces outputs with higher PCQI, QRCM, and entropy and lower BRISQUE compared to HE, USM, MF, and FF.

Higher QRCM value indicates OCLAHE outputs with preserved gradient profiles that are free of textural distortions. High entropy signifies OCLAHE outputs with high information content and greater contrast. Low BRISQUE reflects the natural appearance of OCLAHE outputs. High PCQI shows OCLAHE outputs with higher local contrast that are free from over-enhancement/overshoot.

The OCLAHE amplifies the intensity differences among vessel regions and backgrounds without exaggerating the overall brightness of the angiograms. Thus, improved visual quality of the coronary angiograms facilitated by the OCLAHE will ease automated and subjective detection of stenosis.