Background: We directly compared the mammography image processing results obtained with the help of the YOLOv4 convolutional neural network (CNN) model versus those obtained with the help of the NCA-based nested contours algorithm model. Method: We used 1080 images to train the YOLOv4, plus 100 images with proven breast cancer (BC) and 100 images with proven absence of BC to test both models. Results: the rates of true-positive, false-positive and false-negative outcomes were 60, 10 and 40, respectively, for YOLOv4, and 93, 63 and 7, respectively, for NCA. The sensitivities for the YOLOv4 and the NCA were comparable to each other for star-like lesions, masses with unclear borders, round- or oval-shaped masses with clear borders and partly visualized masses. On the contrary, the NCA was superior to the YOLOv4 in the case of asymmetric density and of changes invisible on the dense parenchyma background. Radiologists changed their earlier decisions in six cases per 100 for NCA. YOLOv4 outputs did not influence the radiologists’ decisions. Conclusions: in our set, NCA clinically significantly surpasses YOLOv4.

We developed the nested contour algorithm (NCA) - a segmentation method for X-ray mammography images and tested it on a set of 1532 images of 356 women with morphologically proven breast cancer of various characteristics located on different density background. As a result NCA correctly marked 48 of 52 (92.31 %) star-like lesions, 12 of 14 (85.71 %) architectural distortions, 51 of 58 (87.93 %) lesions with irregular shape and unclear margin, all 18 lobular and round lesions, 17 of 18 (94.4 %) partially visualized lesions, 13 of 18 (72.2 %) asymmetric areas and 7 of 16 (43.8 %) unclearly visible or invisible lesions. Overall sensitivity of NCA in our set was 90.73 % (323 of 356 cases). The mean rate of false-positive marks was 1.3 per image - for ACR A-B mammograms and 1.8 - for ACR C-D mammograms.

This letter considers a new method for sample detection of ionospheric propagation modes at the receiver output of a chirp ionosonde. The method is based on the use of the hazard function. Within the framework of the two-component mixture model, it is shown that the hazard function can be used to detect samples of the second component in the mixture when the corresponding fraction of samples in the total number is small. The implementation of the method is carried out using training sets. An average normalized hazard function is constructed for the training set not containing the signal. Normalization is carried out in such a way that the hazard function can be considered as a probability density function of a certain random variable. We used Pearson's statistic to test the agreement between the average normalized hazard function of samples that do not contain the signal and an arbitrary normalized hazard function. The critical value of the statistic was determined using the Neyman-Pearson criterion. The effectiveness of the proposed method has been tested for the determination of the lowest and maximum observed frequencies of radio paths from oblique ionospheric sounding ionograms obtained in the Russian network of circumpolar radio paths. The method suggested in this letter can be used not only when processing ionograms but also for the detection of samples of signals of arbitrary nature against a background noise, provided that statistical signal-to-noise ratios are similar to those considered here.