An Innovative Faster R-CNN-Based Framework for Breast Cancer Detection in MRI.

Replacing lung cancer as the most commonly diagnosed cancer globally, breast cancer (BC) today accounts for 1 in 8 cancer diagnoses and a total of 2.3 million new cases in both sexes combined. An estimated 685,000 women died from BC in 2020, corresponding to 16% or 1 in every 6 cancer deaths in women. BC represents a quarter of a total of cancer cases in females and by far the most commonly diagnosed cancer in women in 2020. However, when detected in the early stages of the disease, treatment methods have proven to be very effective in increasing life expectancy and, in many cases, patients fully recover. Several medical imaging modalities, such as X-rays Mammography (MG), Ultrasound (US), Computer Tomography (CT), Magnetic Resonance Imaging (MRI), and Digital Tomosynthesis (DT) have been explored to support radiologists/physicians in clinical decision-making workflows for the detection and diagnosis of BC. In this work, we propose a novel Faster R-CNN-based framework to automate the detection of BC pathological Lesions in MRI. As a main contribution, we have developed and experimentally (statistically) validated an innovative method improving the "breast MRI preprocessing phase" to select the patient's slices (images) and associated bounding boxes representing pathological lesions. In this way, it is possible to create a more robust training (benchmarking) dataset to feed Deep Learning (DL) models, reducing the computation time and the dimension of the dataset, and more importantly, to identify with high accuracy the specific regions (bounding boxes) for each of the patient's images, in which a possible pathological lesion (tumor) has been identified. As a result, in an experimental setting using a fully annotated dataset (released to the public domain) comprising a total of 922 MRI-based BC patient cases, we have achieved, as the most accurate trained model, an accuracy rate of 97.83%, and subsequently, applying a ten-fold cross-validation method, a mean accuracy on the trained models of 94.46% and an associated standard deviation of 2.43%.

Representation learning for mammography mass lesion classification with convolutional neural networks.

BACKGROUND AND OBJECTIVE: The automatic classification of breast imaging lesions is currently an unsolved problem. This paper describes an innovative representation learning framework for breast cancer diagnosis in mammography that integrates deep learning techniques to automatically learn discriminative features avoiding the design of specific hand-crafted image-based feature detectors. METHODS: A new biopsy proven benchmarking dataset was built from 344 breast cancer patients' cases containing a total of 736 film mammography (mediolateral oblique and craniocaudal) views, representative of manually segmented lesions associated with masses: 426 benign lesions and 310 malignant lesions. The developed method comprises two main stages: (i) preprocessing to enhance image details and (ii) supervised training for learning both the features and the breast imaging lesions classifier. In contrast to previous works, we adopt a hybrid approach where convolutional neural networks are used to learn the representation in a supervised way instead of designing particular descriptors to explain the content of mammography images. RESULTS: Experimental results using the developed benchmarking breast cancer dataset demonstrated that our method exhibits significant improved performance when compared to state-of-the-art image descriptors, such as histogram of oriented gradients (HOG) and histogram of the gradient divergence (HGD), increasing the performance from 0.787 to 0.822 in terms of the area under the ROC curve (AUC). Interestingly, this model also outperforms a set of hand-crafted features that take advantage of additional information from segmentation by the radiologist. Finally, the combination of both representations, learned and hand-crafted, resulted in the best descriptor for mass lesion classification, obtaining 0.826 in the AUC score. CONCLUSIONS: A novel deep learning based framework to automatically address classification of breast mass lesions in mammography was developed.

Convolutional neural networks for mammography mass lesion classification.

Feature extraction is a fundamental step when mammography image analysis is addressed using learning based approaches. Traditionally, problem dependent handcrafted features are used to represent the content of images. An alternative approach successfully applied in other domains is the use of neural networks to automatically discover good features. This work presents an evaluation of convolutional neural networks to learn features for mammography mass lesions before feeding them to a classification stage. Experimental results showed that this approach is a suitable strategy outperforming the state-of-the-art representation from 79.9% to 86% in terms of area under the ROC curve.

Impact of ensemble learning in the assessment of skeletal maturity.

The assessment of the bone age, or skeletal maturity, is an important task in pediatrics that measures the degree of maturation of children's bones. Nowadays, there is no standard clinical procedure for assessing bone age and the most widely used approaches are the Greulich and Pyle and the Tanner and Whitehouse methods. Computer methods have been proposed to automatize the process; however, there is a lack of exploration about how to combine the features of the different parts of the hand, and how to take advantage of ensemble techniques for this purpose. This paper presents a study where the use of ensemble techniques for improving bone age assessment is evaluated. A new computer method was developed that extracts descriptors for each joint of each finger, which are then combined using different ensemble schemes for obtaining a final bone age value. Three popular ensemble schemes are explored in this study: bagging, stacking and voting. Best results were achieved by bagging with a rule-based regression (M5P), scoring a mean absolute error of 10.16 months. Results show that ensemble techniques improve the prediction performance of most of the evaluated regression algorithms, always achieving best or comparable to best results. Therefore, the success of the ensemble methods allow us to conclude that their use may improve computer-based bone age assessment, offering a scalable option for utilizing multiple regions of interest and combining their output.

Discovering mammography-based machine learning classifiers for breast cancer diagnosis.

This work explores the design of mammography-based machine learning classifiers (MLC) and proposes a new method to build MLC for breast cancer diagnosis. We massively evaluated MLC configurations to classify features vectors extracted from segmented regions (pathological lesion or normal tissue) on craniocaudal (CC) and/or mediolateral oblique (MLO) mammography image views, providing BI-RADS diagnosis. Previously, appropriate combinations of image processing and normalization techniques were applied to reduce image artifacts and increase mammograms details. The method can be used under different data acquisition circumstances and exploits computer clusters to select well performing MLC configurations. We evaluated 286 cases extracted from the repository owned by HSJ-FMUP, where specialized radiologists segmented regions on CC and/or MLO images (biopsies provided the golden standard). Around 20,000 MLC configurations were evaluated, obtaining classifiers achieving an area under the ROC curve of 0.996 when combining features vectors extracted from CC and MLO views of the same case.

A software framework for building biomedical machine learning classifiers through grid computing resources.

This paper describes the BiomedTK software framework, created to perform massive explorations of machine learning classifiers configurations for biomedical data analysis over distributed Grid computing resources. BiomedTK integrates ROC analysis throughout the complete classifier construction process and enables explorations of large parameter sweeps for training third party classifiers such as artificial neural networks and support vector machines, offering the capability to harness the vast amount of computing power serviced by Grid infrastructures. In addition, it includes classifiers modified by the authors for ROC optimization and functionality to build ensemble classifiers and manipulate datasets (import/export, extract and transform data, etc.). BiomedTK was experimentally validated by training thousands of classifier configurations for representative biomedical UCI datasets reaching in little time classification levels comparable to those reported in existing literature. The comprehensive method herewith presented represents an improvement to biomedical data analysis in both methodology and potential reach of machine learning based experimentation.