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]

Tissues and organs consist of cells organized in specified patterns that support their function, as exemplified by tissues such as skin, muscle, and cornea. It is, therefore, important to understand how external cues, such as engineered surfaces or chemical contaminants, can influence the organization and morphology of cells. In this work, we studied the impact of indium sulfate on human dermal fibroblast (GM5565) viability, production of reactive oxygen species (ROS), morphology, and alignment behavior on tantalum/silicon oxide parallel line/trench surface structures. The viability of cells was measured using the alamarBlue™ Cell Viability Reagent probe, while the ROS levels in cells were quantified using cell-permeant 2′,7′-dichlorodihydrofluorescein diacetate. Cell morphology and orientation on the engineered surfaces were characterized using fluorescence confocal and scanning electron microscopy. When cells were cultured in media containing indium (III) sulfate, the average cell viability decreased by as much as ~32% and the concentration of cellular ROS increased. Cell geometry became more circular and compact in the presence of indium sulfate. Even though actin microfilaments continue to preferentially adhere to tantalum-coated trenches in the presence of indium sulfate, the cells are less able to orient along the line axes of the chips. Interestingly, the indium sulfate-induced changes in cell alignment behavior are pattern dependent—a larger proportion of adherent cells on structures with line/trench widths in the range of 1 μm and 10 μm lose the ability to orient themselves, compared to those grown on structures with line widths smaller than 0.5 μm. Our results show that indium sulfate impacts the response of human fibroblasts to the surface structure to which they adhere and underscores the importance of evaluating cell behaviors on textured surfaces, especially in the presence of potential chemical contaminants. [ABSTRACT FROM AUTHOR]

Cell adhesion is an essential biological function for division, migration, signaling and tissue development. While it has been demonstrated that this cell function can be modified by using nanometer-scale surface topographic structures, it remains unknown how contaminants such as indium (III) ion might influence this specific cell behavior. Herein, the influence of indium chloride on human dermal fibroblast (GM5565) adhesion characteristics was investigated, given the frequent contact of contaminants with skin. The morphology of the adherent cells and their mitochondrial reticulum was characterized on cell culture dishes and nanopatterned surfaces by using fluorescence confocal microscopy and scanning electron microscopy. Results showed a significant proportion of cells lost their ability to align preferentially along the line axes of the nanopattern upon exposure to 3.2 mM indium chloride, with cells aligned within 10° of the pattern line axes reduced by as much as ~70%. Concurrent with the cell adhesion behaviors, the mitochondria in cells exposed to indium chloride exhibit a punctate staining that contrasts with the normal network of elongated tubular geometry seen in control cells. Our results demonstrate that exposure to indium chloride has detrimental effects on the behavior of human fibroblasts and adversely impacts their mitochondrial morphology. This shows the importance of evaluating the biological impacts of indium compounds. [ABSTRACT FROM AUTHOR]