• Accuracy The total correct predictions (true positives and true negatives) divided by the total set of predictions; see Table 17-3 for mathematical description.

• Adaptive histogram equalization Histogram equalization that occurs at the “local” level, wherein distinct sections of an image are equalized across a histogram of light intensity; it can enhance edges in each region of an image.

• Artificial intelligence A field of study that focuses on the study of algorithms that give machines the ability to reason and perform cognitive functions.

• Aspect ratio Ratio of an image’s or video’s width to height; the traditional aspect ratio is 4:3, whereas the more modern or widescreen aspect ratio is 16:9.

• Automatic speech recognition (ASR) See computerized voice recognition.

• Autonomy With regard to ethical principles in medicine, refers to the respect for the decisions of adults who have decision-making capacity.

• Back-end speech recognition (BESR) A human-hybrid production approach that applies computerized voice recognition to prepare a draft transcript of human-dictated speech, which is then passed to human transcriptionists for final review and editing.

• Backpropagation An algorithm to adjust weights within a neural network from final layer back to input layer based on each weight’s contribution to error during forward propagation.

• Beneficence The moral obligation to act for the benefit of others.

• Biomedical informatics Branch of health informatics that uses data to help clinicians, researchers, and scientists improve human health and provide health care.

• Channel effect An effect wherein a model may learn characteristics of the manner in which data were recorded, in addition to the nature of the data themselves.

• Classification decision trees A hierarchical structure to achieve classification of data in an iterative fashion such that the final structure represents a “tree.” Each tree is composed of a node, which reflects the test or question on which the data is divided; a branch, which is the outcome of a test performed at the preceding node; and leaf nodes, which represent the final node that yields an outcome. The data are split iteratively into partitions until each partition is sufficiently homogenous to yield an acceptable results based on predefined stopping criteria.

• Computer vision Machine understanding of visual data (ie, images and videos).

• Computerized voice recognition An extension of signal processing to include recognition of speech utterances, application of language and speech models, and computational machine learning (assisted by neural networks and cloud computing) to enable machine awareness and understanding of spoken human speech.

• Conditional random field A statistical modeling method that is used for structured predictions and takes context into account. For example, in natural language processing, sequential data about surrounding words can be taken into account, or in computer vision, data about nearby structures can be used to help predictions.

• Contextual bias Bias that may occur if one ignores what may appear to be trivial caveats (eg, whether the operating surgeon is right- or left-handed) or if one presumes that the same level of resources is available across operating rooms and hospitals.

• Convolutional neural network A type of ...

### Pop-up div Successfully Displayed

This div only appears when the trigger link is hovered over. Otherwise it is hidden from view.