SummaryThe Optimization‐Driven Multispectral Gamma Correction (ODMGC) algorithm overcomes challenges in gathering subtle information and detecting cancer in dense breast thermograms. This algorithm enhances the accuracy of true positives and true negatives while minimising false negatives and false positives. The ODMGC involves a multi‐step optimisation process that categorises grey‐scale images of breast thermograms based on mean brightness. Then, based on the grey levels of the pixels, we grouped each categorisation into sub‐regions. Followed by each group has undergone individually optimised base enhancement. This process enhances the contrast between cancerous and normal tissues, eliminates over‐ and under‐enhancement, and supports breast tumour diagnosis. The optimised‐based enhancement images serve as a reference point for the histogram specification of the V component of the thermograms in the HSV (Hue, Saturation, and Value) model. Further, we evaluated the proposed model using both qualitative and quantitative measures. Finally, using dimension‐reduced significant Grey‐Level Co‐occurrence Matrix (GLCM) features, we validated the results with a Random Forest (RF) classifier. The algorithm was successfully implemented in MATLAB 2020a, and the classifier was developed in Jupyter Notebook using Python. The subjective comparison confirmed the proposed method's superior resolution in normal and malignant cases. The classifier results showed an accuracy of 96.4%, sensitivity of 98.1%, and specificity of 96.9%.

中文翻译：

---

通过优化驱动的多光谱伽玛校正 (ODMGC) 诊断乳腺癌

摘要优化驱动的多光谱伽玛校正 (ODMGC) 算法克服了在密集乳房热像图中收集细微信息和检测癌症方面的挑战。该算法提高了真阳性和真阴性的准确性，同时最大限度地减少了假阴性和假阳性。 ODMGC 涉及一个多步骤优化过程，该过程根据平均亮度对乳房热像图的灰度图像进行分类。然后，根据像素的灰度级，我们将每个类别分组为子区域。接下来每组都经过了单独优化的基础强化。这个过程增强了癌组织和正常组织之间的对比度，消除了过度增强和增强不足，并支持乳腺肿瘤的诊断。基于优化的增强图像可作为 HSV（色调、饱和度和值）模型中热分析图 V 分量直方图规范的参考点。此外，我们使用定性和定量措施评估了所提出的模型。最后，使用降维显着灰度共生矩阵（GLCM）特征，我们用随机森林（RF）分类器验证了结果。该算法在 MATLAB 2020a 中成功实现，分类器是使用 Python 在 Jupyter Notebook 中开发的。主观比较证实了所提出的方法在正常和恶性病例中具有优异的分辨率。分类器结果显示准确度为 96.4%，敏感性为 98.1%，特异性为 96.9%。