Search

In this work, we provide specific clinical examples to demonstrate basic practical techniques involved in image segmentation, computer-aided design, and 3D printing. A step-by-step approach using United States Food and Drug Administration cleared software is provided to enhance surgical intervention in a patient with a complex superior sulcus tumor. Furthermore, patient-specific device creation is demonstrated using dedicated computer-aided design software. Relevant anatomy for these tasks is obtained from CT Digital Imaging and Communications in Medicine images, leading to the generation of 3D printable files and delivery of these files to a 3D printer.

Deep neural networks have been widely adopted in numerous domains due to their high performance and accessibility to developers and application-specific end-users. Fundamental to image-based applications is the development of Convolutional Neural Networks (CNNs), which possess the ability to automatically extract features from data. However, comprehending these complex models and their learned representations, which typically comprise millions of parameters and numerous layers, remains a challenge for both developers and end-users. This challenge arises due to the absence of interpretable and transparent tools to make sense of black-box models. There exists a growing body of Explainable Artificial Intelligence (XAI) literature, including a collection of methods denoted Class Activation Maps (CAMs), that seek to demystify what representations the model learns from the data, how it informs a given prediction, and why it, at times, performs poorly in certain tasks. We propose a novel XAI visualization method denoted CAManim that seeks to simultaneously broaden and focus end-user understanding of CNN predictions by animating the CAM-based network activation maps through all layers, effectively depicting from end-to-end how a model progressively arrives at the final layer activation. Herein, we demonstrate that CAManim works with any CAM-based method and various CNN architectures. Beyond qualitative model assessments, we additionally propose a novel quantitative assessment that expands upon the Remove and Debias (ROAD) metric, pairing the qualitative end-to-end network visual explanations assessment with our novel quantitative "yellow brick ROAD" assessment (ybROAD). This builds upon prior research to address the increasing demand for interpretable, robust, and transparent model assessment methodology, ultimately improving an end-user's trust in a given model's predictions. Examples and source code can be found at: https://omni-ml.github.io/pytorch-grad-cam-anim/. [ABSTRACT FROM AUTHOR]

Background: This prospective study investigated whether the use of 3D-printed model facilitates novice learning of radiology anatomy on multiplanar computed tomography (CT) when compared to traditional 2D-based learning tools. Specifically, whether the use of a 3D printed model improved interpretation of multiplanar CT tracheobronchial anatomy. Methods: Thirty-one medical students (10F, 21 M) from years one to three were recruited, matched for gender and level of training and randomized to 2D or 3D group. Students underwent 20-min self-study session using 2D-printed image or 3D-printed model of the tracheobronchial tree. Immediately after, students answered 10 multiple-choice questions (Test 1) to identify tracheobronchial tree branches on multiplanar CT images. Two weeks later, identical test (Test 2) was used to assess retention of information. Mean scores of 2D and 3D groups were calculated. Student's t test was used to compare mean differences in tests scores and analysis of variance (ANOVA) was used to assess the interaction of gender, CT imaging plane and time on test scores between the two groups. Results: For test 1, 2D group had higher mean score than 3D group although not statistically significant (7.69 and 7.43, p = 0.39). Mean scores for Test 2 were significantly lower than for Test 1 (7 and 7.57, p = 0.03) with mean score decline for 2D group (Test 1 = 7.69, Test 2 = 6.63, p = 0.03), and similar score for 3D group (Test 1 and 2 = 7.43). There was no statistically significant interaction of gender and test score over time. Significant interaction between group and time of test was found for axial CT images but not for coronal images. Conclusions: Use of a 3D-printed model of the tracheobronchial anatomy had no immediate advantage over traditional 2D-printed images for learning CT anatomy. However, use of a 3D model improved students' ability to retain learned information, irrespective of gender. [ABSTRACT FROM AUTHOR]