Staff

Abstract:
A new detailed dataset of breast ultrasound scans (BrEaST) containing images of benign and malignant lesions as well as normal tissue examples, is presented. The dataset consists of 256 breast scans collected from 256 patients. Each scan was manually annotated and labeled by a radiologist experienced in breast ultrasound examination. In particular, each tumor was identified in the image using a freehand annotation and labeled according to BIRADS features and lexicon. The histopathological classification of the tumor was also provided for patients who underwent a biopsy.
The BrEaST dataset is the first breast ultrasound dataset containing patient-level labels, image-level annotations, and tumor-level labels with all cases confirmed by follow-up care or core needle biopsy result. To enable research into breast disease detection, tumor segmentation and classification, the BrEaST dataset is made publicly available with the CC-BY 4.0 license.

Abstract:
Objective: Neoadjuvant chemotherapy (NAC) in breast cancer requires non-invasive methods of monitoring its effects after each dose of drug therapy. The aim is to isolate responding and non-responding tumors prior to surgery in order to increase patient safety and select the optimal medical follow-up. Methods: A new method of monitoring NAC therapy has been proposed. The method is based on image quality indexes (IQI) calculated from ultrasound data obtained from breast tumors and surrounding tissue. Four different tissue regions from the preliminary set of 38 tumors and three data pre-processing techniques are considered. Postoperative histopathology results were used as a benchmark in evaluating the effectiveness of the IQI classification. Results: Out of ten parameters considered, the best results were obtained for the Gray Relational Coefficient. Responding and non-responding tumors were predicted after the first dose of NAC with an area under the receiver operating characteristics curve of 0.88 and 0.75, respectively. When considering subsequent doses of NAC, other IQI parameters also proved usefulness in evaluating NAC therapy. Conclusions: The image quality parameters derived from the ultrasound data are well suited for assessing the effects of NAC therapy, in particular on breast tumors.

Keywords:
Quantitative ultrasound; Image quality; Neoadjuvant chemotherapy; Breast cancer; Treatment response

Abstract:
In an interesting article previously published in Data in Brief [Dataset of breast ultrasound images" by W. Al-Dhabyani, M. Gomaa, H. Khaled & A. Fahmy, Data in Brief, 2020, 28, 104863], the authors presented a dataset of breast ultrasound images containing lesions. As of April 22, 2023, this study has garnered significant attention from researchers, as evident by its 298 citations in Scopus data. This is unsurprising considering that the study presents one of the few publicly available datasets on breast ultrasound images, as well as binary masks highlighting the lesions. When implementing various aspects of explainable AI, we verify the correctness of the input data at every stage, especially when using various data sources. In an attempt to use this dataset for research, we did some exploration and identified some inconsistencies that could have a significant impact on the results of the studies utilizing them. As the role of tumor detection is indisputable we feel obliged to point attention to some aspects that need to be kept in mind while using this database in order to receive reliable and good quality results.

Abstract:
In breast cancer diagnosis, early detection of tumors and accurate differentiation of malignant and benign breast lesions are key demands. Tumor size, as a measure of tumor progression, is related to recurrence rate and patient survival. This study aims to determine which sonographic features al- low the differentiation of small breast lesions into benign and malignant. Inclusion criteria for the analysis were tumors with the longest diameter of less than or equal to 10 mm and tumors with confirmed classification by follow-up care or core needle biopsy result. Following the criteria, 1515 cases were analyzed, including 365 carcinomas and 1150 benign lesions. To quantitatively evaluate the images, 383 ultrasound parameters
(BI-RADS features, morphological features, fractal features, his- togram features, and texture parameters) were used. Univariate and multiple logistic regression analyses were used to assess the significance of various diagnostic features and their combinations. The combined classifier (based on 19 quantitative features) yields an area under the ROC curve of 0.91.

Keywords:
Ultrasound imaging, Breast cancer, Small lesion, Feature extraction

Abstract:
Manual image segmentations are naturally subject to inaccuracies related to systematic errors (due to the tools used, eye-hand coordination, etc.). This was noted earlier when a simplified accuracy scale was proposed [1]. This scale arbitrarily divides a given range of values of the Kappa measurement parameter into classes: almost perfect (>0.80), substantial (0.61 - 0.80), moderate (0.41 - 0.60), fair (0.21 - 0.40), slight (0.00 - 0.21) and poor (< 0.00). However, the determination of threshold values between classes is not entirely clear and seems to be application-dependent. This is particularly important for images in which the tumor-normal tissue boundary can be very indistinct, as is observed in ultrasound imaging of the most common cancer in women - breast cancer [2]. In machine learning, there is an ongoing contest over the values of performance indicators obtained from new neural network architecture without accounting for any ground truth bias. This raises the question of what relevance, from a segmentation quality point of view, a gain at the level of single percentages has [3] if the references have much greater uncertainty. So far, research on this topic has been limited. The relationship between the segmentations of breast tumors on ultrasound images provided by three radiologists and those obtained using deep learning model has been studied in [4]. Unfortunately, the indicated segmentation contour sometimes varied widely in all three cases. A cursory analysis by multiple physicians, which focused only on the Kappa coefficient in the context of physicians’ BI-RADS category assignments, was conducted in the [5]. In this article, we present a preliminary analysis of the accuracy of experts’ manually prepared binary breast cancer masks on ultrasound images and their impact on performance metrics commonly used in machine learning. In addition, we examined how tumor type or BI-RADS category [6] affects the accuracy of tumor contouring.

Abstract:
Understanding the relationship between acoustic properties of breast lesions and resulting ultrasound images may contribute to an earlier and more accurate diagnosis of the most common cancer in women. In addition to in vitro studies, in silico tumor models can provide a lot of crucial information due to the possibility of precise determination of the influence of changes in tissue structure on the resulting ultrasound echoes. The purpose was to develop the numerical phantom of the breast with the tumor for a reliable simulation of ultrasound images. In modeling the tissue structures of the breast, the VICTRE phantom, developed by the FDA for the simulation of X-ray mammography, was used. The numerical ultrasound model of breast cancer allows the simulation of ultrasound signals and images. It could be used to interpret, validate and develop new ultrasound methods for cancer diagnosis

Keywords:
breast tumor, numerical phantom, ultrasound imaging

Abstract:
Standard classification methods based on hand-crafted morphological and texture features have achieved good performance in breast mass differentiation in ultrasound (US).
In comparison to deep neural networks, commonly perceived as ‘black-box’ models, classical techniques are based on features that have well-understood medical and physical interpretation. However, classifiers based on morphological features commonly
underperform in the presence of the shadowing artifact and ill-defined mass borders, while texture based classifiers may fail when the US image is too noisy. Therefore, in practice it would be beneficial to select the classification method based on the appearance of the particular US image. In this work, we develop a deep meta-network that can automatically process input breast mass US images and recommend whether to apply the shape or
texture based classifier for the breast mass differentiation. Our preliminary results demonstrate that meta-learning techniques can be used to improve the performance of the standard classifiers based on handcrafted features. With the proposed meta-learning based approach, we achieved the area under the receiver operating characteristic curve of 0.95 and accuracy of 0.91.

Keywords:
breast mass classification, deep learning, meta-learning, morphological features, texture features