Search

In recent years, deep convolutional neural networks (DCNNs) have shown promising performance in medical image analysis, including breast lesion classification in 2D ultrasound (US) images. Despite the outstanding performance of DCNN solutions, explaining their decisions remains an open investigation. Yet, the explainability of DCNN models has become essential for healthcare systems to accept and trust the models. This paper presents a novel framework for explaining DCNN classification decisions of lesions in ultrasound images using the saliency maps linking the DCNN decisions to known cancer characteristics in the medical domain. The proposed framework consists of three main phases. First, DCNN models for classification in ultrasound images are built. Next, selected methods for visualization are applied to obtain saliency maps on the input images of the DCNN models. In the final phase, the visualization outputs and domain-known cancer characteristics are mapped. The paper then demonstrates the use of the framework for breast lesion classification from ultrasound images. We first follow the transfer learning approach and build two DCNN models. We then analyze the visualization outputs of the trained DCNN models using the EGrad-CAM and Ablation-CAM methods. We map the DCNN model decisions of benign and malignant lesions through the visualization outputs to the characteristics such as echogenicity, calcification, shape, and margin. A retrospective dataset of 1298 US images collected from different hospitals is used to evaluate the effectiveness of the framework. The test results show that these characteristics contribute differently to the benign and malignant lesions' decisions. Our study provides the foundation for other researchers to explain the DCNN classification decisions of other cancer types. [ABSTRACT FROM AUTHOR]

Automatic segmentation and quantification of skeletal structures has a variety of applications for biological research. Although solutions for good quality X-ray images of human skeletal structures are in existence in recent years, automatic solutions working on poor quality X-ray images of mice are rare. This paper proposes a fully automatic solution for spine segmentation and curvature quantification from X-ray images of mice. The proposed solution consists of three stages, namely preparation of the region of interest, spine segmentation, and spine curvature quantification, aiming to overcome technical difficulties in processing the X-ray images. We examined six different automatic measurements for quantifying the spine curvature through tests on a sample data set of 100 images. The experimental results show that some of the automatic measures are very close to and consistent with the best manual measurement results by annotators. The test results also demonstrate the effectiveness of the curvature quantification produced by the proposed solution in distinguishing abnormally shaped spines from the normal ones with accuracy up to 98.6%. [ABSTRACT FROM AUTHOR]