CHARACTERIZATION OF BREAST TISSUES LABELED WITH QUANTUM DOTS THROUGH IMAGE DESCRIPTOR ANALYSIS

Quantum dots (QDs) are versatile fluorescent semiconductor nanocrystals widely used in various scientific fields, particularly in biology, due to their exceptional resistance to photodegradation and chemically active surfaces that allow conjugation with molecules such as lectins and antibodies. QD-based nanoplatforms have been applied for studying biological changes in cells and tissues, as well as for diagnostic purposes. In addition, while flow cytometry is suitable for quantitative assays with cells, quantitative approaches for analyzing tissues still present challenges. There remains a need to develop reproducible and automated methodologies for measuring fluorescence levels in tissue images and distinguishing between control and test samples. Therefore, in this study, we employed computer vision to analyze and compare normal, fibroadenoma, and invasive ductal carcinoma (IDC) labeled with the QDs-Cramoll nanoprobe. This conjugate was applied to study glucose and mannose patterns in the tissues, as Cramoll lectin has an affinity to these carbohydrates. Our proposal aimed to differentiate these tissues through the extraction and analysis of grayscale and color features using texture and shape descriptors. In addition, we developed an automated method for quantifying fluorescence per pixel through image processing. Using ORB (Oriented FAST and Rotated BRIEF), we extracted descriptors from 18 real and 72 artificial instances. The classification using decision trees achieved 100% accuracy, a sensitivity of 1.0, a specificity of 1.0, and an Area Under the ROC Curve of 1.0 in all color spaces except for grayscale.Measuring pixel-by-pixel fluorescence levels, varied intensities patterns were observed: ORB keypoints were concentrated in high and uniform fluorescence areas. In both normal and fibroadenoma tissue images, ductal cells showed high labeling (~70%) by the conjugate, while stromal regions exhibited predominantly autofluorescence. IDC tissue images displayed higher labeling (~75%) in both ductal and stromal areas (~70%) by the nanoprobe. This study highlights the use of QD-based conjugates and the role of artificial intelligence in distinguishing labeling patterns, thereby improving tissue differentiation and fluorescence quantification, opening perspectives for advancing fluorescence microscopy methodologies.This work was supported by CNPq, CAPES, and FACEPE. This work is also linked to theNational Institute of Photonics (INCT-INFo).