"Language model" refers to a computational model that can be configured to process, generate, predict, transform, classify, summarize, or otherwise analyze linguistic or symbolic information. A language model may be implemented using neural network architectures, transformer-based architectures, machine learning systems, or other computational frameworks capable of modeling relationships among tokens, symbols, or representational units, or the like. A language model may operate in generative, discriminative, inferential, or analytical modes. Synonyms include generative model, neural language processor, sequence model, generative AI model, or the like. … more "Language model" refers to a computational model that can be configured to process, generate, predict, transform, classify, summarize, or otherwise analyze linguistic or symbolic information. A language model may be implemented using neural network architectures, transformer-based architectures, machine learning systems, or other computational frameworks capable of modeling relationships among tokens, symbols, or representational units, or the like. A language model may operate in generative, discriminative, inferential, or analytical modes. Synonyms include generative model, neural language processor, sequence model, generative AI model, or the like.

“Annotated Image” refers to any digital image, radiographic image, fluoroscopic image, video frame, video sequence, three-dimensional rendering, or the like, that includes one or more annotations, markings, overlays, augmentations, or the like. Annotations may include, without limitation, landmarks, lines, curves, indicators, measurements, text, predictive markings, graphical objects, or the like, and may be static, dynamic, interactive, or generated in response to user input. An annotated image may depict any projection or view, including anterior–posterior (AP), lateral (LAT), oblique, axial, weight-bearing, non-weight-bearing, or the like. In one embodiment, an “annotated AP image” is an annotated image depicting an AP projection or view (including dorsoplantar (DP) equivalents). In one embodiment, an “annotated LAT image” is an annotated image depicting a lateral projection or view. The annotated image may be structured in various ways, including as an original image with an overlay, as an original image with metadata defining the placement of annotations and/or measurements, or as a combined data file that integrates both image data and annotation data, or the like. The annotated image may be generated in whole or in part by computational techniques, including software modules, artificial intelligence systems, machine learning models, deep learning models, large language models (LLMs), or the like. The annotated image may represent anatomic data, predictive data, or simulation data, and may be displayed, stored, transmitted, or further processed in various permutations, including single images, video sequences, interactive user interfaces, or the like. In one embodiment, annotated AP images and annotated LAT images are generated from radiographs, fluoroscopic frames, reconstructed projections or views, or digital photographs of displayed radiographic images, and the like. … more “Annotated Image” refers to any digital image, radiographic image, fluoroscopic image, video frame, video sequence, three-dimensional rendering, or the like, that includes one or more annotations, markings, overlays, augmentations, or the like. Annotations may include, without limitation, landmarks, lines, curves, indicators, measurements, text, predictive markings, graphical objects, or the like, and may be static, dynamic, interactive, or generated in response to user input. An annotated image may depict any projection or view, including anterior–posterior (AP), lateral (LAT), oblique, axial, weight-bearing, non-weight-bearing, or the like. In one embodiment, an “annotated AP image” is an annotated image depicting an AP projection or view (including dorsoplantar (DP) equivalents). In one embodiment, an “annotated LAT image” is an annotated image depicting a lateral projection or view. The annotated image may be structured in various ways, including as an original image with an overlay, as an original image with metadata defining the placement of annotations and/or measurements, or as a combined data file that integrates both image data and annotation data, or the like. The annotated image may be generated in whole or in part by computational techniques, including software modules, artificial intelligence systems, machine learning models, deep learning models, large language models (LLMs), or the like. The annotated image may represent anatomic data, predictive data, or simulation data, and may be displayed, stored, transmitted, or further processed in various permutations, including single images, video sequences, interactive user interfaces, or the like. In one embodiment, annotated AP images and annotated LAT images are generated from radiographs, fluoroscopic frames, reconstructed projections or views, or digital photographs of displayed radiographic images, and the like.

“Viewfinder Overlay” refers to a graphical element, display aid, or interface presented on a display of a device, such as a mobile device, camera, or imaging system, that assists a user in capturing an image. A viewfinder overlay may provide alignment cues, boundary indicators, targeting reticles, crosshairs, grids, framing boxes, shading, highlighting, or other visual markers, or the like, to guide the positioning of a subject within a field of view. In certain embodiments, a viewfinder overlay may be used during capture of a radiographic image displayed on a screen to ensure that the radiograph is properly aligned, centered, and scaled relative to the device’s camera. A viewfinder overlay may be static or dynamic, may change appearance in response to detected motion or alignment, and may provide visual, auditory, or haptic feedback to the user. Examples include overlays that change color when alignment is achieved, gridlines that indicate perspective or scale, or bounding boxes that indicate when an image of interest is positioned within a capture region, or the like. … more “Viewfinder Overlay” refers to a graphical element, display aid, or interface presented on a display of a device, such as a mobile device, camera, or imaging system, that assists a user in capturing an image. A viewfinder overlay may provide alignment cues, boundary indicators, targeting reticles, crosshairs, grids, framing boxes, shading, highlighting, or other visual markers, or the like, to guide the positioning of a subject within a field of view. In certain embodiments, a viewfinder overlay may be used during capture of a radiographic image displayed on a screen to ensure that the radiograph is properly aligned, centered, and scaled relative to the device’s camera. A viewfinder overlay may be static or dynamic, may change appearance in response to detected motion or alignment, and may provide visual, auditory, or haptic feedback to the user. Examples include overlays that change color when alignment is achieved, gridlines that indicate perspective or scale, or bounding boxes that indicate when an image of interest is positioned within a capture region, or the like.

“Correction Procedure” refers to any method, technique, workflow, treatment, operation, intervention, or the like that is intended to produce a correction. Examples include, without limitation, surgical procedures (e.g., osteotomy, arthrodesis, soft-tissue balancing), minimally invasive or percutaneous techniques, intraoperative adjustments, image-guided manipulations, external fixation adjustments, orthotic or prosthetic fitting, bracing, casting, physical therapy protocols, rehabilitation regimens, pharmacologic or biologic interventions, staged or combined approaches, and the like. A correction procedure may be performed on one or more parts of a foot and/or ankle and may include invasive surgical procedures and/or minimally invasive surgical procedures. Foot and ankle examples of correction procedures include, without limitation, Lapidus procedures (first tarsometatarsal arthrodesis), metatarsus adductus corrections (including metatarsal base osteotomies or soft-tissue balancing), Akin osteotomy, Chevron (Austin) distal metatarsal osteotomy, Scarf osteotomy, proximal metatarsal osteotomy, Weil osteotomy, Cotton osteotomy (medial cuneiform opening wedge), Evans lateral column lengthening, medial displacement calcaneal osteotomy (MDCO), Dwyer or other calcaneal closing-wedge osteotomies, first metatarsophalangeal arthrodesis, cheilectomy, bunionette correction, hammertoe correction (PIP arthrodesis or arthroplasty), gastrocnemius recession, Achilles tendon lengthening, plantar fascia release, spring-ligament repair, posterior tibial tendon procedures (including Kidner), Lisfranc arthrodesis, tarsal coalition resection, lateral ligament reconstruction (Broström), subtalar arthrodesis, arthrodesis, and the like. Minimally invasive surgery (MIS) examples of correction procedures include, without limitation, percutaneous or arthroscopically assisted techniques such as MICA/PECA (minimally invasive/percutaneous Chevron-Akin bunion correction), DMMO (distal minimally invasive metatarsal osteotomy) for lesser metatarsals, percutaneous Akin osteotomy, percutaneous distal or proximal metatarsal osteotomies, percutaneous bunionette osteotomy, minimally invasive or percutaneous Lapidus arthrodesis, minimally invasive cheilectomy, endoscopic or percutaneous gastrocnemius recession, percutaneous Achilles tendon lengthening, endoscopic plantar fasciotomy, arthroscopic Broström lateral ligament repair, endoscopic calcaneoplasty for Haglund deformity, percutaneous MDCO or Dwyer calcaneal osteotomy, minimally invasive subtalar arthrodesis, arthroscopically assisted Lisfranc reduction/arthrodesis, percutaneous hammertoe correction (including PIP arthrodesis with intramedullary fixation), percutaneous screw fixation of fractures or osteotomies, and the like. A correction procedure may be planned, simulated, rehearsed, or executed in one or multiple stages, may use instruments, devices, implants, or software, and may be performed manually, robotically, or via computer-assisted systems, and the like. A correction procedure may include procedures that comport to conventional standard of care, as well as standard of care procedures that have certain measurements, corrections, a planned change in measurements, a correction intended to produce different measurements, adjustments, edits, or changes applied. These modifications may be based on surgeon discretion, user preferences, patient preferences, institutional protocols, system-generated recommendations, software-generated recommendations, colleague input, or other factors. In certain embodiments, a correction procedure may be tailored to optimize clinical outcomes, patient satisfaction, workflow efficiency, or procedural safety, or the like. … more “Correction Procedure” refers to any method, technique, workflow, treatment, operation, intervention, or the like that is intended to produce a correction. Examples include, without limitation, surgical procedures (e.g., osteotomy, arthrodesis, soft-tissue balancing), minimally invasive or percutaneous techniques, intraoperative adjustments, image-guided manipulations, external fixation adjustments, orthotic or prosthetic fitting, bracing, casting, physical therapy protocols, rehabilitation regimens, pharmacologic or biologic interventions, staged or combined approaches, and the like. A correction procedure may be performed on one or more parts of a foot and/or ankle and may include invasive surgical procedures and/or minimally invasive surgical procedures. Foot and ankle examples of correction procedures include, without limitation, Lapidus procedures (first tarsometatarsal arthrodesis), metatarsus adductus corrections (including metatarsal base osteotomies or soft-tissue balancing), Akin osteotomy, Chevron (Austin) distal metatarsal osteotomy, Scarf osteotomy, proximal metatarsal osteotomy, Weil osteotomy, Cotton osteotomy (medial cuneiform opening wedge), Evans lateral column lengthening, medial displacement calcaneal osteotomy (MDCO), Dwyer or other calcaneal closing-wedge osteotomies, first metatarsophalangeal arthrodesis, cheilectomy, bunionette correction, hammertoe correction (PIP arthrodesis or arthroplasty), gastrocnemius recession, Achilles tendon lengthening, plantar fascia release, spring-ligament repair, posterior tibial tendon procedures (including Kidner), Lisfranc arthrodesis, tarsal coalition resection, lateral ligament reconstruction (Broström), subtalar arthrodesis, arthrodesis, and the like. Minimally invasive surgery (MIS) examples of correction procedures include, without limitation, percutaneous or arthroscopically assisted techniques such as MICA/PECA (minimally invasive/percutaneous Chevron-Akin bunion correction), DMMO (distal minimally invasive metatarsal osteotomy) for lesser metatarsals, percutaneous Akin osteotomy, percutaneous distal or proximal metatarsal osteotomies, percutaneous bunionette osteotomy, minimally invasive or percutaneous Lapidus arthrodesis, minimally invasive cheilectomy, endoscopic or percutaneous gastrocnemius recession, percutaneous Achilles tendon lengthening, endoscopic plantar fasciotomy, arthroscopic Broström lateral ligament repair, endoscopic calcaneoplasty for Haglund deformity, percutaneous MDCO or Dwyer calcaneal osteotomy, minimally invasive subtalar arthrodesis, arthroscopically assisted Lisfranc reduction/arthrodesis, percutaneous hammertoe correction (including PIP arthrodesis with intramedullary fixation), percutaneous screw fixation of fractures or osteotomies, and the like. A correction procedure may be planned, simulated, rehearsed, or executed in one or multiple stages, may use instruments, devices, implants, or software, and may be performed manually, robotically, or via computer-assisted systems, and the like. A correction procedure may include procedures that comport to conventional standard of care, as well as standard of care procedures that have certain measurements, corrections, a planned change in measurements, a correction intended to produce different measurements, adjustments, edits, or changes applied. These modifications may be based on surgeon discretion, user preferences, patient preferences, institutional protocols, system-generated recommendations, software-generated recommendations, colleague input, or other factors. In certain embodiments, a correction procedure may be tailored to optimize clinical outcomes, patient satisfaction, workflow efficiency, or procedural safety, or the like.

“Generative Network” refers to a model configured to synthesize data samples (e.g., images) from random inputs and/or conditioning signals. Examples include GANs, variational autoencoders (VAEs), diffusion models, autoregressive models, or hybrids. A generative network may be conditioned on pixel-wise representations, masks/label maps, edges, landmarks, text prompts, or other anatomic data. … more “Generative Network” refers to a model configured to synthesize data samples (e.g., images) from random inputs and/or conditioning signals. Examples include GANs, variational autoencoders (VAEs), diffusion models, autoregressive models, or hybrids. A generative network may be conditioned on pixel-wise representations, masks/label maps, edges, landmarks, text prompts, or other anatomic data.

“Convolutional Neural Network (CNN)” refers to a neural network employing convolutional operations (optionally with pooling and up/down-sampling) suited to images or grid-structured data. Examples include U-Net and U-Net-derived encoder–decoder architectures with skip connections. CNNs may be used for segmentation, detection, classification, denoising, super-resolution, and the like. … more “Convolutional Neural Network (CNN)” refers to a neural network employing convolutional operations (optionally with pooling and up/down-sampling) suited to images or grid-structured data. Examples include U-Net and U-Net-derived encoder–decoder architectures with skip connections. CNNs may be used for segmentation, detection, classification, denoising, super-resolution, and the like.

“Deep Learning (DL)” refers to the use of neural networks with multiple layers (deep neural networks) that learn hierarchical feature representations. Deep learning architectures include, without limitation, convolutional neural networks (CNNs), recurrent networks, and transformers, and may be trained in supervised or unsupervised/self-supervised regimes. … more “Deep Learning (DL)” refers to the use of neural networks with multiple layers (deep neural networks) that learn hierarchical feature representations. Deep learning architectures include, without limitation, convolutional neural networks (CNNs), recurrent networks, and transformers, and may be trained in supervised or unsupervised/self-supervised regimes.

“Machine Learning (ML)” refers to techniques by which models improve performance from data. ML may be supervised, semi-supervised, self-supervised, unsupervised, or reinforcement learning, and may address tasks such as classification, regression, clustering, dimensionality reduction, and segmentation. Representative models include, without limitation, neural networks, transformers, support-vector machines, decision trees/ensembles, Bayesian models, and clustering algorithms. Models may be trained by minimizing a loss function and evaluated using metrics such as accuracy, precision/recall/F1, ROC/AUC, or the like. … more “Machine Learning (ML)” refers to techniques by which models improve performance from data. ML may be supervised, semi-supervised, self-supervised, unsupervised, or reinforcement learning, and may address tasks such as classification, regression, clustering, dimensionality reduction, and segmentation. Representative models include, without limitation, neural networks, transformers, support-vector machines, decision trees/ensembles, Bayesian models, and clustering algorithms. Models may be trained by minimizing a loss function and evaluated using metrics such as accuracy, precision/recall/F1, ROC/AUC, or the like.

“Artificial Neural Network (ANN)” refers to a parameterized computational architecture comprising layers of interconnected units (neurons) that apply learned weights and nonlinear activation functions to input data. ANNs may be trained by backpropagation (optionally with stochastic gradient methods) and may include components such as convolutional layers, pooling, normalization, attention, and skip connections. ANN families include feedforward, convolutional, recurrent, and transformer-based networks. … more “Artificial Neural Network (ANN)” refers to a parameterized computational architecture comprising layers of interconnected units (neurons) that apply learned weights and nonlinear activation functions to input data. ANNs may be trained by backpropagation (optionally with stochastic gradient methods) and may include components such as convolutional layers, pooling, normalization, attention, and skip connections. ANN families include feedforward, convolutional, recurrent, and transformer-based networks.

“Anatomy Plane” refers to a geometric plane in three-dimensional space associated with an anatomical region of interest. Examples include, without limitation, a plane approximating the level of selected anatomical structures (e.g., metatarsal heads), a plantar reference plane, a sagittal or coronal plane, or the like. An anatomy plane may be defined explicitly from landmarks, fitted to points or contours, or inferred from acquisition geometry including the positions and/or orientations of imaging emitters and sensors (e.g., X-ray source and detector, fluoroscope C-arm, camera or mobile-device camera), and may be modeled as coincident with, parallel to, or at a known or estimated offset from a detector plane or display plane. An anatomy plane may also be estimated computationally. As used herein, “at an anatomy plane” denotes that a measurement or calibration corresponds to the physical scale at that anatomical level; where the anatomy is not perfectly planar, the term includes a representative plane within an acceptable tolerance. In one embodiment, the anatomy plane is used for patient-plane calibration to compute a pixel-to-length conversion in millimeters per pixel. … more “Anatomy Plane” refers to a geometric plane in three-dimensional space associated with an anatomical region of interest. Examples include, without limitation, a plane approximating the level of selected anatomical structures (e.g., metatarsal heads), a plantar reference plane, a sagittal or coronal plane, or the like. An anatomy plane may be defined explicitly from landmarks, fitted to points or contours, or inferred from acquisition geometry including the positions and/or orientations of imaging emitters and sensors (e.g., X-ray source and detector, fluoroscope C-arm, camera or mobile-device camera), and may be modeled as coincident with, parallel to, or at a known or estimated offset from a detector plane or display plane. An anatomy plane may also be estimated computationally. As used herein, “at an anatomy plane” denotes that a measurement or calibration corresponds to the physical scale at that anatomical level; where the anatomy is not perfectly planar, the term includes a representative plane within an acceptable tolerance. In one embodiment, the anatomy plane is used for patient-plane calibration to compute a pixel-to-length conversion in millimeters per pixel.

As used herein, "Plumbline (PL)" refers to a radiographic assessment reference construct and/or anatomic data representing forefoot alignment, used to evaluate whether space exists to align the first ray relative to a longitudinal foot axis without requiring lesser-ray realignment. In one embodiment, a PL is a constructed reference. In an AP projection or view (including dorsoplantar (DP) equivalents), the PL is constructed from landmarks associated with the medial cuneiform and the first tarsometatarsal (TMT) joint and extended distally toward the lesser metatarsals. A ‘positive PL’ indicates that the PL intersects the second metatarsal head (or shaft), suggesting metatarsus adductus (MTA) of a degree that may impede complete first-ray correction; a ‘negative PL’ indicates that the PL is tangential to, or does not intersect, the second metatarsal, suggesting sufficient clearance for first-ray correction. Minor variations in beam angle, source-to-detector distance, patient positioning, camera viewpoint, or reconstruction that do not materially alter the assessment are included. In some embodiments, a proximity tolerance ε is used to classify borderline cases as intersecting or tangential. McAleer, J. P., DeCarbo, W. T., D’Antonio, P. C., & Hatch, D. J. (2024). "A Simplified Preoperative Radiographic Assessment for Metatarsus Adductus Associated With Hallux Valgus." Foot & Ankle Orthopaedics, 9(2). https://doi.org/10.1177/24730114241246830[] … more As used herein, "Plumbline (PL)" refers to a radiographic assessment reference construct and/or anatomic data representing forefoot alignment, used to evaluate whether space exists to align the first ray relative to a longitudinal foot axis without requiring lesser-ray realignment. In one embodiment, a PL is a constructed reference. In an AP projection or view (including dorsoplantar (DP) equivalents), the PL is constructed from landmarks associated with the medial cuneiform and the first tarsometatarsal (TMT) joint and extended distally toward the lesser metatarsals. A ‘positive PL’ indicates that the PL intersects the second metatarsal head (or shaft), suggesting metatarsus adductus (MTA) of a degree that may impede complete first-ray correction; a ‘negative PL’ indicates that the PL is tangential to, or does not intersect, the second metatarsal, suggesting sufficient clearance for first-ray correction. Minor variations in beam angle, source-to-detector distance, patient positioning, camera viewpoint, or reconstruction that do not materially alter the assessment are included. In some embodiments, a proximity tolerance ε is used to classify borderline cases as intersecting or tangential. McAleer, J. P., DeCarbo, W. T., D’Antonio, P. C., & Hatch, D. J. (2024). "A Simplified Preoperative Radiographic Assessment for Metatarsus Adductus Associated With Hallux Valgus." Foot & Ankle Orthopaedics, 9(2). https://doi.org/10.1177/24730114241246830[]

"Constructed Reference” refers to a derived geometric construct computed from landmarks or anatomic data, including points, centroids, axes, lines, rays, curves, planes, regions, tolerance bands, or the like, used for measurement, registration, planning, guidance, or assessment. Constructed references may be rendered as annotations and/or stored as data. … more "Constructed Reference” refers to a derived geometric construct computed from landmarks or anatomic data, including points, centroids, axes, lines, rays, curves, planes, regions, tolerance bands, or the like, used for measurement, registration, planning, guidance, or assessment. Constructed references may be rendered as annotations and/or stored as data.

“Geometric Rectification” refers to processing that transforms a captured image to compensate for perspective and/or lens distortion and to isolate a depiction of an image of interest, including, without limitation, homography or projective warping, undistortion, dewarping, rotation, scaling, cropping, resampling, glare/reflection suppression, and the like. … more “Geometric Rectification” refers to processing that transforms a captured image to compensate for perspective and/or lens distortion and to isolate a depiction of an image of interest, including, without limitation, homography or projective warping, undistortion, dewarping, rotation, scaling, cropping, resampling, glare/reflection suppression, and the like.

“Hallux Valgus Angle (HVA)” refers to an angular measurement formed between a long axis of the first metatarsal and a long axis of the proximal phalanx of the hallux, typically measured on an anterior–posterior view. HVA may be computed from anatomic data and/or constructed references and may be used to classify deformity severity. … more “Hallux Valgus Angle (HVA)” refers to an angular measurement formed between a long axis of the first metatarsal and a long axis of the proximal phalanx of the hallux, typically measured on an anterior–posterior view. HVA may be computed from anatomic data and/or constructed references and may be used to classify deformity severity.

“Intermetatarsal Angle (IMA)” refers to an angular measurement formed between long axes of the first and second metatarsals, typically measured on an anterior–posterior view. IMA may be computed from anatomic data and/or constructed references and may be used to select and/or evaluate correction procedures. … more “Intermetatarsal Angle (IMA)” refers to an angular measurement formed between long axes of the first and second metatarsals, typically measured on an anterior–posterior view. IMA may be computed from anatomic data and/or constructed references and may be used to select and/or evaluate correction procedures.

“Lesser-Ray Realignment” refers to any correction of alignment of one or more lesser rays (metatarsals two through five), including, without limitation, osteotomies, arthrodesis, soft-tissue procedures, image-guided manipulations, or the like, performed alone or in combination with other corrections. … more “Lesser-Ray Realignment” refers to any correction of alignment of one or more lesser rays (metatarsals two through five), including, without limitation, osteotomies, arthrodesis, soft-tissue procedures, image-guided manipulations, or the like, performed alone or in combination with other corrections.

“Masks” refers to pixel-wise representations of regions or classes in an image, including, without limitation, binary masks, label maps, instance masks, probability maps, heatmaps, run-length–encoded masks, vector contours rasterized to a grid, or the like, at native or resampled resolution. … more “Masks” refers to pixel-wise representations of regions or classes in an image, including, without limitation, binary masks, label maps, instance masks, probability maps, heatmaps, run-length–encoded masks, vector contours rasterized to a grid, or the like, at native or resampled resolution.

“Metatarsus Adductus (MTA)” refers to medial deviation of the forefoot relative to the midfoot and/or hindfoot in the transverse plane, characterized by adduction of one or more metatarsals toward the foot midline. MTA may be flexible or rigid and may be assessed using angular, axis-based, or categorical measures. … more “Metatarsus Adductus (MTA)” refers to medial deviation of the forefoot relative to the midfoot and/or hindfoot in the transverse plane, characterized by adduction of one or more metatarsals toward the foot midline. MTA may be flexible or rigid and may be assessed using angular, axis-based, or categorical measures.

“Metatarsus Adductus Metric” refers to any quantitative or categorical representation of MTA, including, without limitation, angles, distances, ratios, scores, thresholds, or a categorical indicator derived from one or more constructed references, and the like. … more “Metatarsus Adductus Metric” refers to any quantitative or categorical representation of MTA, including, without limitation, angles, distances, ratios, scores, thresholds, or a categorical indicator derived from one or more constructed references, and the like.

“Patient-Plane Calibration” refers to determining a scale factor that converts pixel measurements to physical dimensions at the anatomical plane of interest. In one embodiment, calibration is obtained using a calibration object of known size positioned at the anatomy plane; in another embodiment, calibration is computed from imaging geometry and/or metadata. … more “Patient-Plane Calibration” refers to determining a scale factor that converts pixel measurements to physical dimensions at the anatomical plane of interest. In one embodiment, calibration is obtained using a calibration object of known size positioned at the anatomy plane; in another embodiment, calibration is computed from imaging geometry and/or metadata.