With breast cancer being one of the most widespread causes of death for women, there is an unmet need for its early detection. For this purpose, we propose a non-invasive approach based on X-ray scattering. We measured samples from 107 unique patients provided by the Breast Cancer Now Tissue Biobank, with the total dataset containing 2958 entries. Two different sample-to-detector distances, 2 and 16 cm, were used to access various structural biomarkers at distinct ranges of momentum transfer values. The biomarkers related to lipid metabolism are consistent with those of previous studies. Machine learning analysis based on the Random Forest Classifier demonstrates excellent performance metrics for cancer/non-cancer binary decisions. The best sensitivity and specificity values are 80% and 92%, respectively, for the sample-to-detector distance of 2 cm and 86% and 83% for the sample-to-detector distance of 16 cm.

We report the results of X-ray diffraction (XRD) measurements of the dogs’ claws and show the feasibility of using this approach for early, non-invasive cancer detection. The obtained two-dimensional XRD patterns can be described by Fourier coefficients, which were calculated for the radial and circular (angular) directions. We analyzed these coefficients using the supervised learning algorithm, which implies optimization of the random forest classifier by using samples from the training group and following the calculation of mean cancer probability per patient for the blind dataset. The proposed algorithm achieved a balanced accuracy of 85% and ROC-AUC of 0.91 for a blind group of 68 dogs. The transition from samples to patients additionally improved the ROC-AUC by 10%. The best specificity and sensitivity values for 68 patients were 97.4% and 72.4%, respectively. We also found that the structural parameter (biomarker) most important for the diagnostics is the intermolecular distance.

With the growing number of cancer cases and deaths around the world, fast, non-invasive, and inexpensive screening is paramount. We examine the feasibility of such cancer detection using the X-ray scattering properties of nails in the canine model. A total of 945 samples taken from 266 dogs were measured, with 84 animals diagnosed with cancer. To analyze the obtained X-ray diffraction patterns of keratin, we propose a method based on the two-dimensional Fourier transformation of the images. We compare 745 combinations of data preprocessing steps and machine learning classifiers and determine the corresponding performance metrics. Excellent classification results are demonstrated, with sensitivity or specificity achieving 100% and the best value for balanced accuracy being 87.5%. We believe that our approach can be extended to human samples to develop a non-invasive, convenient, and cheap method for early cancer detection.

Background/Objectives: With the number of detected breast cancer cases growing every year, there is a need to augment histopathological analysis with fast preliminary screening. We examine the feasibility of using X-ray diffraction measurements for this purpose. Methods: In this work, we obtained more than 6000 diffraction patterns from 211 patients and examined both standard and custom-developed methods, including Fourier coefficient analysis, for their interpretation. Various preprocessing steps and machine learning classifiers were compared to determine the optimal combination. Results: We demonstrated that benign and cancerous clusters are well separated, with specificity and sensitivity exceeding 0.9. For wide-angle scattering, the two-dimensional Fourier method is superior, while for small angles, the conventional analysis based on azimuthal integration of the images provides similar metrics. Conclusions: X-ray diffraction of biopsy tissues, supported by machine learning approaches to data analytics, can be an essential tool for pathological services. The method is rapid and inexpensive, providing excellent metrics for benign/cancer classification.

As the number of cancer cases and deaths growing around the world, fast, non-invasive, and inexpensive screening is paramount. We examine the feasibility of such cancer detection using the X-ray scattering properties of nails in the canine model. 945 samples taken from 266 dogs were measured, with 84 animals diagnosed with cancer. To analyze the obtained X-ray diffraction patterns of keratin, we propose a method based on the two-dimensional Fourier transformation of the images. We compare 745 combinations of data preprocessing steps and machine learning classifiers and determine corresponding performance metrics. Excellent classification results are demonstrated, with sensitivity or specificity achieving 100% and the best value for balanced accuracy being 87.5%. We believe that our approach can be extended to human samples to develop a non-invasive, convenient, and cheap method for early cancer detection.

Background/Objectives: Machine learning approaches are widely used in modern medical diagnostics, including cancer detection. The results can be significantly improved by aggregating individual measurements, and appropriate aggregation methods should be established. Methods: We applied various measurement aggregation strategies both before and after machine learning modeling to two datasets of X-ray diffraction images: human breast biopsy samples and canine claw samples. Two classifiers, Random Forest and Logistic Regression, were used to determine classification metrics: the area under the receiver operating characteristic curve (ROC-AUC) and balanced accuracy. Results: We found that all aggregation types improve classification metrics, with aggregation after modeling yielding better performance. Depending on the dataset and approach, either classifier can produce better results. For human breast samples, Random Forest with the logit aggregation strategy provides an ROC-AUC exceeding 0.9. For the canine dataset, both Random Forest with the logit aggregation strategy and Logistic Regression with the median of cancer probabilities achieve an ROC-AUC of about 0.85. Conclusions: We examined several simple, straightforward aggregation methods for patient diagnosis based on multiple measurements per patient and achieved significant improvements in classification metrics.