Parts | Days | Selection | Search | Updates | Downloads | Help

Q: Fachverband Quantenoptik und Photonik

Q 40: Poster – Photonics

Q 40.18: Poster

Wednesday, March 4, 2026, 17:00–19:00, Philo 1. OG

Machine learning driven texture analysis of laser speckle imaging for early breast cancer detection — •Doaa Youssef — Department of Engineering Applications of Lasers, National Institute of Laser Enhanced Sciences (NILES), Cairo University, Egypt

Breast cancer remains one of the most prevalent malignancies, and early detection is crucial for improving patient outcomes. We present an image-guided breast cancer detection method based on laser speckle imaging and machine learning. The approach exploits the distinct absorption and scattering properties of healthy and malignant breast tissue. A dual-wavelength optical system using low-power 532 nm and 632 nm lasers was developed to record speckle patterns from ex vivo breast samples. Diagnostic information is extracted using a strategy that combines multi-neighborhood local entropy with a Gabor filter bank to generate texture maps that reveal subtle structural changes. These maps are fused to form a single informative image, from which texture features are obtained using two histogram-based methods and refined through data reduction techniques. The discriminative capability of the extracted features was assessed using three supervised classification models: support vector machine (SVM), ensemble k-nearest neighbor (E-kNN), and extreme gradient boosting (XGB). Combining information from both wavelengths improved overall diagnostic performance, yielding an accuracy of 98.48% and a weighted F1 score of 98.54%. These findings demonstrate the potential of laser speckle-based optical diagnostics integrated with AI for affordable, non-destructive early breast lesion detection.

Keywords: Breast Cancer; Laser Speckle Imaging; Local Entropy; Gabor Filter Bank; Machine learning