Systems and methods for determining and / or predicting acute-post operative airway complications associated with anterior cervical spine surgery

Abstract

Embodiments pertain to systems and methods for detecting and / or predicting Post-Operative Airway (POA) Complications, the system comprising: at least one processor; and at least one memory configured to store data and software code portions executable by the at least one processor to cause to perform: receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway, providing an Airway Analysis Engine with the received data; and determining for the received data, by the Airway Analysis Engine, a probability that the mammalian is expected to experience a POA-related complication. Embodiments and methods may also be configured to provide an output relating to the determined probability. In some examples, the Airway Analysis Engine may comprise a trained Machine- Learning (ML) module for predicting probabilities relating to POA complications.

Description

SYSTEMS AND METHODS FOR DETERMINING AND / OR PREDICTING ACUTE-POST OPERATIVE AIRWAY COMPLICATIONS ASSOCIATED WITH ANTERIOR CERVICAL SPINE SURGERYCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This patent application claims domestic and international priority from Israel Patent Application 310199 filed January 16, 2024, titled "Systems and Methods for Determining and / or Predicting Acute-Post Operative Airway Complications Associated with Anterior Cervical Spine Surgery", which is incorporated herein by reference in its entirety.BACKGROUND

[0002] Back pain and pathologies of the spine such as adult spinal deformities are one of the costliest health problems in society, and may be classified based on severity, etiology, radiographic distinctions, neutral upright spinal alignment, or based on any number of pelvic sagittal or other parameters for measurement of the spine.

[0003] Spine pathologies can result in nerve conduction disorders associated with herniation of the intervertebral disc, in which a small amount of tissue protrudes from the sides of the disc into the foramen to compress the spinal cord. Another condition involves the development of small bone spurs, termed osteophytes, along the posterior surface of the vertebral body, impinging on the spinal cord.

[0004] Upon identification of these abnormalities, Anterior cervical spine (ACS) surgery may be required to correct the problem.

[0005] The description above is presented as a general overview of related art in this field and should not be construed as an admission that any of the information it contains constitutes prior art against the present patent application.BRIEF DESCRIPTION OF THE FIGURES

[0006] The figures illustrate generally, by way of example, but not by way of limitation, various embodiments discussed in the present document. For simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity of presentation. Furthermore, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.References to previously presented elements are implied without necessarily further citing the drawing or description in which they appear. The figures are listed below.

[0007] Figure 1A shows a schematic block diagram illustration of a Post-Operative Airway (POA) Complications Detection and / or Prediction System, according to some embodiments.

[0008] Figure IB shows an example Graphical User Interface (GUI) for using the system, according to some embodiments.

[0009] Figure 2 depicts a flowchart of a study design, according to some embodiments.

[0010] Figures 3A and 3B show screenshots of three-dimensional visualization of the automated segmentation outputs, according to some embodiments.[Oil] Figure 4 depicts an Al Model's Training Workflow, according to some embodiments.

[0012] Figure 5 depicts the trained Al model confusion matrix, according to some embodiments.

[0013] Figure 6 depicts a three-dimensional representation of the decision boundary of a CatBoost model, according to some embodiments.

[0014] Figure 7 depicts a flowchart of a method for detecting and / or predicting Post-Operative Airway (POA) Complications Detection, according to some embodiments.DETAILED DESCRIPTION

[0015] ASC surgery may be associated with the rare yet life-threatening complication of acute postoperative airway (POA) obstruction due to edema or hematoma. Acute POA obstruction, if not recognized and treated promptly can have devastating outcomes if not recognized and treated promptly.

[0016] Aspects of embodiments pertain to systems and methods configured to determine, in association with ASC surgery, a probability of and / or predict complications to occur in a patient and / or provide an estimate and / or score and / or any other output that can be indicative of a patient to develop and / or experience acute POA-related complications, such as acute POA-related obstruction. The output may be a numerical output (e.g., score, probability), a categorical output (e.g., list of patients ranking according to risk assessment from highest list to lowest list, e.g., for a selected list of patients), a nominal output (e.g., "high risk", "no risk" patient).

[0017] In some examples, the systems and methods disclosed consider thickness of the patient's PVST for distinguishing between acute and non-acute POA-related obstructions. Example PVST thickness thresholds are disclosed in (19).

[0018] Merely to simplify the discussion that follows, without be construed in a limiting manner, embodiments and / or examples may herein refer to "determining the probability of a patient to experience complications in relation to ASC surgery".

[0019] In some embodiments, the probability and / or estimate may be associated with a time period, and the systems and methods may output a probability-time tuple, where the probability may be associated with a time period within which the complication may be expected to occur.

[0020] In some embodiments, the system and methods may output information about an anatomical location likely being the source for the expected complication such as, e.g., location of the edema and / or hematoma, for instance, relative to at least one vertebra. Accordingly, the systems and methods may output a probability-location tuple relating to a complication that may be associated with ASC surgery.

[0021] In some embodiments, the system and methods may output information about a time period within the complication is likely to occur. Accordingly, the systems and methods may output a probabilitytime tuple relating to a complication that may be associated with ASC surgery, or a probability-time- anatomical location tuple.

[0022] In some embodiments, the system may be configured to identify an anomaly in the patient for determining, based on the detected anomaly, a probability of a patient to experience complications in association with ASC surgeries. In some implementations, the anomaly may be identified pre-operatively, intra-operatively, and / or post-operatively. In some implementations, the probability of a patient to experience complications in association with ASC surgery may be determined pre-operatively, intra- operatively, and / or post-operatively. In some examples, a probability of a patient to experience complications in association with ASC surgery in association with a time period and / or an anatomical location of interest may be determined or identified pre-operatively, intra-operatively, and / or post- operatively.

[0023] In some embodiments, the system may be configured to distinguish between a first anomaly associated with ASC surgery, but which is not related to acute-post operative airway complications (e.g., non-threatening post-operative imaging changes), and a second anomaly associated with ASC surgery and related to acute-post operative airway complications. The system and methods may be configured to determine, with respect to the second anomaly, a probability for a patient to experience acute postoperative airway complications.

[0024] In some embodiments, a system according to some embodiments may include one or more classifiers for identifying, based on an image dataset, acute or critical POA narrowing and / or determining a probability to develop acute POA narrowing. In some examples, the classifier may employ or include on or more ML models trained based on a training image dataset associated with labels relating to "critical"and "non-critical post-operative airway narrowing" for determining and / or predicting a probability that a patient experiences acute POA obstruction.

[0025] In some examples, the image dataset may be segmented for training the ML learning based on datasets of segmented images. In some examples, the segmentation of the images may be performed, for example, on osseous vertebral structures, implanted hardware, retropharyngeal space, and / or the airway. In some examples, the image datasets may be anonymized image datasets. In some examples, the image datasets may be based on one or more imaging modalities including, for example, X-ray images, and / or CT images. In some examples, labels may additionally pertain to additional medical, physiological, and / or socio-economic and / or behavioral patient characteristics such as, for example, gender, age, race, height, BMI, smoking habits, drinking habits, medical history, and / or the like.

[0026] An ML model may be trained using supervised and / or unsupervised learning. In some examples, the classifier may be a regression-based classifier, based on artificial neural networks (ANNs), and / or based on Gradient Boosting model. The Gradient Boosting algorithm produces a prediction model that is based on an ensemble of weak prediction models (e.g., decision trees). The model is designed to solve an optimization problem that tries to minimize the difference between the model predictions on a test dataset, and the real labels on a dataset of labelled data.

[0027] In some examples, the machine learning model may be adapted by evaluating labels produced by a test dataset. The validation measures may include, for example, accuracy, recall and / or precision, with respect to real labels on a dataset of labeled data.

[0028] In some embodiments, the apparatus may be configured to perform image dataset analysis using heuristics models. Further, in some instances, the machine learning and heuristics models may be combined into a hybrid model for analyzing the image dataset.

[0029] In some embodiments, datasets may be excluded as training datasets, based on one more exclusion criteria. In some examples, the system may be configured to automatically include and / or exclude training datasets provided for training the classifier engine. Exclusion criteria include, for instance, cervical spinal cord injuries graded ASIA A-C; respiratory complications necessitating mechanical ventilation unrelated to airway obstruction; and / or absence of post-operative imaging acquired during admission.

[0030] As used herein the term "machine learning" refers to a procedure embodied as a computer program configured to induce patterns, regularities, and / or rules from previously collected data to develop an appropriate response to future data or describe the data in some meaningful way.

[0031] Examples of machine learning procedures suitable for the present embodiments, include, without limitation, clustering, association rule algorithms, feature evaluation algorithms, subset selectionalgorithms, support vector machines, classification rules, cost-sensitive classifiers, vote algorithms, stacking algorithms, Bayesian networks, decision trees, neural networks, instance-based algorithms, linear modeling algorithms, k-nearest neighbors (KNN) analysis, ensemble learning algorithms, probabilistic models, graphical models, logistic regression methods (including multinomial logistic regression methods), gradient ascent methods, singular value decomposition methods and principle component analysis.

[0032] The machine learning procedure used according to some embodiments of the present invention is a trained machine learning procedure, which provides output that is related non-linearly to the parameters with which it is fed.

[0033] In some embodiments, a machine learning procedure can be trained according to some embodiments of the present invention by feeding a machine learning training program with parameters that characterizes each of a cohort of subjects that have been diagnosed as either experiencing or not experiencing acute POA obstruction. Once the data is fed, the machine learning training program generates a trained machine learning procedure or forms a part of a ML module. In some examples, the trained ML module can be used without the need to re-train it. In some other examples, the trained ML module may be further trained and tested.

[0034] In some embodiments, the system includes at least one processor; and

[0035] at least one memory configured to store data and software code portions executable by the at least one processor to cause to perform the following:

[0036] receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway,

[0037] providing a trained ML module with the received data; and

[0038] determining for the data, by the trained ML module, an output relating to or indicative of a comparatively high probability that the mammalian is expected to experience a severe POA-related complication. Optionally, the mammalian is a human subject.

[0039] In some embodiments, the system may provide a physician, based on the performed analysis, with one or more intervention recommendations relating to detected acute POA-related complications. These interventions can include, for example, subjecting patient to ICU observation, intubation, and / or wound revision.

[0040] In some embodiments, the determining may include distinguishing between:

[0041] a first anomaly relating to a non-acute POA obstruction; and

[0042] a second anomaly relating to an acute POA obstruction.

[0043] In some embodiments, the system may produce at least one first output indicative of a probability expected to develop non-acute POA obstruction; and

[0044] at least one second output indicative of a probability expected to develop acute POA obstruction.

[0045] Referring now to Figure 1A, a POA Complications and / or Prediction system 1000 may comprise an I / O device 1100, a processor 1200 and a memory 1300.

[0046] In some example implementations, the system may provide a user thereof with outputs via I / O device 1100 comprising one or more output devices. The one or more output devices may include, for example, devices that are configured to convert electrical signals into outputs that can be sensed as output by a human, such as sound, light, and / or touch. Output devices can include display screens, and / or audio output device(s) such as, for example, speaker(s) and / or earphones.

[0047] I / O device 1100 may further include one or more input devices which are configured to receive any type of data and / or information by converting, for example, or machine-generated signals and / or human-generated signals such as physical movement, physical touch or pressure, and / or the like, into electrical signals as input data into the computing system. Examples of such input devices include touch screens, microphones, hand gesture tracking devices, hand-held pointing devices (e.g., computer mouse, stylus) and / or the like.

[0048] I / O device 1100 may be employed to access data and / or information generated by the system 1000 and / or to provide inputs including, for instance, control commands, operating parameters, queries, and / or the like. For example, I / O device 1100 may allow a user of a system to receive or access medical airway images of a patient and / or other patient-related information. In some examples, I / O Device 1100 may interface with a Graphical User Interface (GUI) 1110, e.g., shown in Figure IB. GUI 1110 may for example be employed for initiating the uploading of medical information (e.g., postoperative CT and / or XR images), and initiating the processing of uploaded medical patient information to output a prediction, (e.g., probability) if the patient will experience acute POA-related complications, if unattended.

[0049] System 1000 may further include a processor 1200 and a memory 1300 which is configured to store data 1310 (e.g., patient data) and algorithm code and / or a machine learning (ML) model 1320. Processor 1200 may be configured to execute algorithm code and / or apply machine learning (ML) model 1320 for the processing of data 1310 resulting in the implementation of an Airway Analysis engine 1400. Engine 1400 may be configured to provide an output relating to a probability that the patient will experience acute POA complications, if unattended.

[0050] The term "processor", as used herein, may additionally or alternatively refer to a controller. Processor 1200 may be implemented by various types of processor devices and / or processor architectures including, for example, embedded processors, communication processors, graphics processing unit(GPU)-accelerated computing, soft-core processors, quantum-based processor and / or general-purpose processors.

[0051] Memory 1300 may be implemented by various types of memories, including transactional memory and / or long-term storage memory facilities and may function as file storage, document storage, program storage, or as a working memory. The latter may for example be in the form of a static randomaccess memory (SRAM), dynamic random-access memory (DRAM), read-only memory (ROM), cache and / or flash memory. As working memory, memory 1300 may, for example, include, e.g., temporally based and / or non-temporally based instructions. As long-term memory, memory 1300 may for example include a volatile or non-volatile computer storage medium, a hard disk drive, a solid-state drive, a magnetic storage medium, a flash memory and / or other storage facility. A hardware memory facility may, for example, store a fixed information set (e.g., software code) including, but not limited to, a file, program, application, source code, object code, data, and / or the like.

[0052] System 1000 may further comprise at least one communication module 1500 configured to enable wired and / or wireless communication between the various components and / or modules of the apparatus and which may communicate with each other over one or more communication buses (not shown), signal lines (not shown) and / or a network infrastructure. Communication module 1500 may be configured for enabling communication using one or more communication formats, protocols, and / or technologies such as, for example, to internet communication, optical or RF communication, telephonybased communication technologies and / or the like. In some examples, communication module 1500 may include I / O device drivers (not shown) and network interface drivers (not shown) for enabling the transmission and / or reception of data over a network. A device driver may, for example, interface with a keypad or to a USB port. A network interface driver may for example execute protocols for the Internet, or an Intranet, Wide Area Network (WAN), Local Area Network (LAN) employing, e.g., Wireless Local Area Network (WLAN)), Metropolitan Area Network (MAN), Personal Area Network (PAN), extranet, 2G, 3G, 3.5G, 4G, 5G, 6G mobile networks, 3GPP, LTE, LTE advanced, Bluetooth® (e.g., Bluetooth smart), ZigBee™, near-field communication (NFC) and / or any other current or future communication network, standard, and / or system.

[0053] System 1000 may further include a power module 1600 for powering the various components and / or modules and / or subsystems of the apparatus. Power module 1600 may comprise an internal power supply (e.g., a rechargeable battery) and / or an interface for allowing connection to an external power supply.

[0054] It will be appreciated that separate hardware components such as processors and / or memories may be allocated to each component and / or module of system 1000. However, for simplicity and without be construed in a limiting manner, the description and claims may refer to a single module and / orcomponent. For example, although processor 1200 may be implemented by several processors, the following description will refer to processor 1200 as the component that conducts all the necessary processing functions of system 1000.

[0055] Functionalities of system 1000 may be implemented fully or partially by a multifunction mobile communication device also known as "smartphone", a mobile or portable device, a non-mobile or nonportable device, a digital video camera, a personal computer, a laptop computer, a tablet computer, a server (which may relate to one or more servers or storage systems and / or services associated with a business or corporate entity, including for example, a file hosting service, cloud storage service, online file storage provider, peer-to-peer file storage or hosting service and / or a cyberlocker), personal digital assistant, a workstation, a wearable device, a handheld computer, a notebook computer, a vehicular device, a non-vehicular device and / or a stationary device. For example, some of engine 1400 functionalities may be implemented on-premises (e.g., in a hospital or other clinical facility), and some by devices, apparatuses and / or system which are located off-premises (e.g., the "cloud"). Alternative configurations may also be conceived.

[0056] EXAMPLES

[0057] Reference is now made to the following examples, which together with the above descriptions illustrate some embodiments in a non-limiting fashion.

[0058] Acute post-operative airway obstruction following anterior cervical spine surgery due to edema or hematoma is a rare but potentially life-threatening complication that may result in devastating outcomes if not recognized and treated promptly. The aim of this study was to develop novel artificial intelligence (Al) software for predicting and identifying post-operative airway obstruction at an early stage based on routine post-operative imaging.

[0059] To achieve this goal, the inventors developed novel artificial intelligence (Al) software for predicting and identifying post-operative airway obstruction at an early stage based on routine postoperative imaging.

[0060] Methods:

[0061] All adult patients operated for anterior cervical fusion between 2012-2019 at a single tertiary care medical center were retrospectively identified. These patients were used for the development of an Al model that autonomously discerned critical airway narrowing from non-threatening post-operative swelling.

[0062] Exclusion criteria included:

[0063] 1. cervical spinal cord injuries graded ASIA A-C,

[0064] 2. respiratory complications necessitating mechanical ventilation unrelated to airway obstruction, and

[0065] 3. absence of post-operative imaging acquired during admission. Various image processing augmentation techniques were applied to establish a segmentation model that incorporated 3D reconstructions for CT scans and pixel analysis for plain radiographs. Multiple training and validation sets were generated and enhanced by repetitive randomization. The CatBoost algorithm was harnessed to refine decision trees and formulate precise predictions. Standard performance metrics were used to assess the model's accuracy.

[0066] Results:

[0067] Overall, 815 patients were identified. The rate of respiratory distress due to airway compromise was 1.4%. 420 patients comprised the dataset for the algorithm training. The model autonomously segmented the airway from Cl downwards, retropharyngeal space, implanted hardware, and osseous anatomy, and predicted post-operative airway compromise with a positive predictive value of 0.98, negative predictive value of 0.9, sensitivity of 0.91, and specificity of 0.99.

[0068] Conclusions:

[0069] The Al model designed in this study showed promising potential in predicting airway compromise following anterior cervical spine surgery, as well as discerning normal postoperative changes from rapidly deteriorating complications like edema or retropharyngeal hematoma. The model demonstrated adaptability to varying radiographic environments, encompassing both radiographs and CT scans. While this tool may assist in the early detection of these life-threatening complications, further studies are necessary to validate our initial findings and integrate this modality into real-life clinical practice.

[0070] Introduction

[0071] Acute post-operative airway obstruction following anterior cervical spine surgery is a rare but potentially life-threatening complication. If not promptly recognized and treated, it can result in devastating outcomes, including prolonged hypoxemia, anoxic brain damage, and the need for crash intubation. The estimated incidence rates of this complication vary between studies, ranging between 0.7% and 2.4% of anterior spinal fusion procedures 6,16. Possible etiologies for post-operative airway obstruction include retropharyngeal hematoma with or without pharyngeal edema, that may be exacerbated by vocal cord paralysis due to recurrent laryngeal nerve injury (4,10,17,20). To date, there remains a limited understanding of the established risk factors contributing to the development of this complication. Previously described risk factors include increased number of surgical levels involved, administration of anticoagulants, extended surgery duration, as well as the presence preexisting conditions such as of diffuse idiopathic skeletal hyperostosis (DISH) and ossification of the posteriorlongitudinal ligament (OPLL) (21 16). Early identification of initial radiographic signs is paramount in the management of this condition and the prevention of permanent damage. This enables timely implementation of urgent wound opening, hematoma removal (if present), and emergent intubation while the airway is still accessible. In cases where these options are not feasible, cricothyroidotomy may be necessary (4,14,20). Accurately distinguishing between non-threatening post-operative imaging changes and early indicators of potentially adverse outcomes may be challenging due to the considerable variability in the normal appearance and thickness of pre-vertebral soft tissue during the early postoperative period (19). The aim of this study is to develop novel artificial intelligence (Al) software for predicting and identifying post-operative airway obstruction at an early stage based on routine postoperative imaging.

[0072] Reference is now made to Figure 2, depicting a flowchart of the study design, and to Figure 4, depicting an Al Model's Training Workflow.

[0073] Methods

[0074] Following the institutional review board's (IRB) approval for this study, all adult patients operated for anterior cervical fusion between 2012-2019 at a single tertiary care medical center for any clinical indication were identified using keyword-based query in the electronic medical record (EMR). Patients experiencing respiratory adverse events in the immediate and intermediate post-operative period were recognized through individual review of all hospitalization and outpatient follow-up clinical notes. In addition to baseline characteristics and operative parameters, the clinical course following surgery, including emergent wound opening, intubation, and surgical exploration, were documented. Exclusion criteria were:

[0075] 1. high cervical spinal cord injuries causing severe respiratory muscle dysfunction (ASIA A- C; n=20) (Figure 2, block 2100),

[0076] 2. respiratory complications necessitating mechanical ventilation unrelated to airway obstruction (n=24), and

[0077] 3. the absence of post-operative imaging prior to airway obstruction due to emergent wound opening (n=2).

[0078] Surgical technique and post-operative assessment

[0079] In an inpatient setting, surgeries involved general anesthesia, endotracheal intubation, nasogastric tube insertion, and conventional prevertebral dissection. Bilateral fixation of sharp retractors to longus coli muscles was performed using Caspar devices (Aesculap Corp, Tuttlingen, Germany). Multilevel stenosis was managed with either multiple discectomies or hybrid constructs using corpectomyand discectomy. The posterior longitudinal ligament (PLL) was consistently resected to achieve complete dural release. Standard static or dynamic plates were used for fusion by various manufacturing companies. In all cases, a prevertebral drain was placed and was subsequently removed on the first postoperative day if its output was less than 50cc over an 8-hour period. Immediately following surgery, patients were monitored for two hours in the post-anesthesia care unit (PACU), then transferred to the ward. Routine cervical imaging of either CT or anterior-posterior and lateral radiograph was performed on the first post-operative day (POD1). The choice between imaging modalities was directed by or at the most caudal operated level and the individual patient's size, to ensure acceptable visualization of the construct. Patients with clinically significant airway obstruction underwent bedside wound opening and revision. If feasible, awake fiberoptic intubation was performed in the operating room before surgical revision. Those with mild symptoms, or those who displayed asymptomatic but narrowed airways on imaging, received conservative management and were monitored in a neurosurgical ICU.

[0080] Image Processing, Optimization, and Mitigation of Overfitting

[0081] Digital Imaging and Communications in Medicine (DICOM) files from post-operative imaging studies were anonymized and prepared for segmentation. A semi-supervised automated segmentation process was performed, focusing on the following structures: the airway from Cl downwards, retropharyngeal space, implanted hardware, and osseous anatomy.

[0082] For computed tomography (CT) scans, a three-dimensional reconstruction of the airway and retropharyngeal space was achieved to facilitate a comprehensive volumetric assessment (in voxels). Figures 3A and 3B shows screenshots of different views of three-dimensional visualization of the automated segmentation output, showing three-dimensional reconstruction of anatomical regions, including the spine, pre-vertebral space, hardware, and airway segmented by the Al model For radiographs, measurements were conducted within a two-dimensional environment (pixels).

[0083] Data augmentation techniques were used to bolster data accuracy and model robustness (Medical Image Processing Toolbox, MATLAB 2022b, The MathWorks, Inc. Massachusetts, United States). To mitigate the inherent risk of overfitting in this Al-driven model, several measures were implemented. The data were systematically partitioned into training and validation subsets at a 4:1 ratio. In addition, a balanced cohort was maintained by adhering to a 1:40-50 ratio between patients with and without airway obstructions. From the n=9 patients with airway obstructions included in this model, an additional n=411 unaffected patients were randomly chosen, resulting in a dataset of n=420 patients (Figure 2). Additionally, the training set underwent five rounds of random data allocation between the training and validation groups (Figure 4).

[0084] Following dataset anonymization, various image processing augmentation techniques were applied to create a segmentation model. 3D reconstructions were used for CT scans, while pixel analysis was conducted for radiographs. Multiple training and validation sets were created and augmented through repetitive randomization. The CatBoost algorithm was employed to optimize decision trees and generate accurate predictions, which were integrated into a graphical user interface (GUI) for user- friendly interaction.

[0085] Gradient Boosting Algorithm

[0086] CatBoost (Yandex, Inc. Moscow, Russia), a gradient boosting algorithm, was utilized to process the automatically segmented voxel data from CT scans and pixel data from radiographic scans. Within the Catboost model, the decision trees were designed to make sequential selections between radiographs and CT scans. This flexibility was incorporated to cater to diverse clinical scenarios, accounting for instances where either a CT or radiograph were the primary post-operative imaging method. The model generated and permutated multiple decision trees to address data variations. In cases where data points, like post-operative radiographs, were absent for certain patients, CatBoost diverged from the common gradient boosting approach of creating synthetic data. Instead, it treated the missing data as a distinct category, allowing for continued learning and prediction even in the presence of data gaps. In addition, k- fold cross-validation approach was utilized 1, wherein the data was split into k segments and the model was subsequently trained on each. This approach aimed to bolster its dependability and curtail the likelihood of overfitting.

[0087] Statistical Analysis

[0088] Descriptive statistics were utilized to represent patient demographics, the type of surgery performed, the duration of the surgery, and the imaging studies conducted. Continuous variables were portrayed as mean, standard deviation (SD), and interquartile range (IQR), while categorical variables were depicted using counts and percentages. The performance of the models was assessed using key metrics, such as positive predictive value (PPV), negative predictive value (NPV), Balanced Fl score, and overall accuracy.

[0089] Results

[0090] A total of 815 anterior cervical fusion procedures were performed between 2012-2019 (Figure 2, block 2050). After the exclusion of patients attributed to SCI ASIA A-C (n=20) (Figure 2, block 2100, resulting in remaining 795, shown in block 2200), those with respiratory complications requiring mechanical ventilation not related to airway obstruction (n=24), and patients lacking post-operative imaging due to the urgency of emergent wound opening (n=2), the study's cohort was consolidated to 769 individuals. From the 760 unaffected participants, a random subset (n=411) was designated (block2300) as the control group for model training (Figure 2). The rate of overall respiratory complications was 4.4% (35 / 795) (block 2250). The rate of respiratory distress due to airway compromise was 1.38% (11 / 795) (block 2260). The mean follow-up period was 7.65 1 0.6 months. Patient characteristics are displayed in Table 1:Characteristic Negative AOPAC Positive AOPACNo. of patients 411 9Age in yrs 57.8217.8 61.4615.73Female n, % (45) 185 3 (33.3)LOS (Days) 3.32± 2.6 12.31 11.9Number of Levels 2.2 ± 1.2 2.051 1.8(Average)Type of surgery corpectomy □ lo / Discectomy 4hybrid 91 2Surgery Duration (Min) 13.7 88.951 1031 20.99Imaging resultsEdema 3Hematoma 6

[0091] In two patients, the airway obstruction resulted in sustained hypoxemia that culminated in anoxic brain damage. There were nine patients who were successfully intubated without any permanent damage related to hypoxemia. While cricothyroidotomy was not performed in any of the patients, urgent wound opening was performed in nine patients. Following endotracheal extubation, all patients underwent surgical exploration, with three revealing edema and six revealing frank hematoma.

[0092] Al Algorithm Performance

[0093] The model's Confusion Matrix is displayed in Figure 5. Overall, the model demonstrated a PPV, NPV, sensitivity, and specificity of 0.98, 0.9, 0.91, and 0.99, respectively. The model's Decision Boundary plot featuring the airway volume, retropharyngeal space volume, and weighted radiograph-based findings, is displayed in Figure 6.

[0094] Figure 6 visualizes how the model classifies instances based on retropharyngeal space volume, airway volume in the operated levels, and an integrated radiograph findings parameter. While two axes represent retropharyngeal space volume and airway volume, the third axis shows the integrated radiograph findings, which combines multiple two-dimensional parameters. The decision boundary in this three-dimensional space demonstrates the model's classification patterns and how it separates instances with different combinations of these features.

[0095] Discussion

[0096] This study presented an Al model designed to detect and predict early airway obstruction following anterior cervical spinal fusion, while discerning harmless postoperative soft tissue changes from potential life-threatening airway obstruction. A total of 420 patients were included in the model's training (Figure 2, block 2400), yielding positive predictive value, negative predictive value, sensitivity, and specificity of 0.98, 0.9, 0.91, and 0.99, respectively.

[0097] Recently, the Symposium on Al in Orthopedic Surgery by the American Orthopedic Association laid out common Al-related terminology and highlighted the distinctions between supervised and unsupervised learning (18). While supervised learning typically employs manually labeled data, unsupervised learning relies on unlabeled data and identifies patterns autonomously. In addition to the aforementioned learning techniques, semi-supervised learning is another well-established strategy (3). This method integrates both manually labeled data and an advanced automated learning process. For this study, a two-tiered strategy was employed. Initially, a semi-supervised segmentation was executed, laying the basis for the subsequent supervised gradient boosting algorithm. In this latter stage, the outcome (specifically, airway obstruction) was manually evaluated by investigators through a thorough review of patient records. Our preference to employ CatBoost stems from its improved performance in managing categorical attributes and absent data, innate immunity to overfitting, and its resilience against diverse data distributions (5,7,8). These attributes were particularly useful considering the variability and unpredictability inherent in post-operative radiographic data, as well as the scarcity of post-operative airway obstruction overall (17,19).

[0098] Predicting Pending Post-Surgical Airway Obstruction

[0099] In attempt to predict pending post-operative airway obstruction, Chen et al (2) sought to predict the need for tracheostomy in patients with deep neck infections, utilizing the K-Nearest Neighbor (KNN) method based on manually derived CT measurements. This approach produced relatively moderate sensitivity and specificity values of 62.50% and 80.60%, respectively. In another retrospective review of 774 patients operated for anterior cervical surgery, the authors reported 1.8% rate of postoperative airway obstruction necessitating reintubation. With the majority (85.7%) occurring within 48 hours of the surgery and the rest presenting 9-11 days after surgery. The study identified several risk factors associated with an increased likelihood of reintubation: advanced age, smoking, elevated BMI, preference for corpectomies over discectomies, surgical interventions on segments above C5, extended operation durations, and a greater number of operated surgical segments (11). Lastly, one study explored the use of ultrasound to detect increased soft tissue swelling that leads to airway obstruction after ACDF, as opposed to relying on standard post-operative imaging (14). While these studies relied on retrospective evaluations to identify risk factors for airway obstruction following anterior approach to the cervical spine, the present study strived to target the prediction of post-operative airway obstruction in Al-driven approach.

[0100] Overfitting in Al Models

[0101] Al-powered algorithms designed for diagnostic and predictive purposes are inherently susceptible to overfitting (1,15). At its core, overfitting is characterized by an excessive adaptation of the model to its training data. Such a model may perform impressively on the training dataset, but may struggle significantly when presented with novel, unseen data, often leading to unreliable predictions and reduced generalizability (9). The main culprits behind overfitting include over-training on limited datasets, the presence of noisy or irrelevant data, lack of a proper validation mechanism, and improper feature selection. Given the low incidence rate of airway obstruction at 1.4%, this study is potentially susceptible to overfitting. To counteract these inherent vulnerabilities, several strategies were employed:

[0102] 1. The dataset was systematically partitioned into training and validation subsets, adhering to a 4:1 proportion. This was done to provide a robust training foundation while allocating a significant portion for model validation, striving to mitigate the influence of noise on model training.

[0103] 2. To ensure a balanced and representative sample in each fold, a ratio of 1:45 was maintained between patients diagnosed with airway obstructions and those without. In line with this, from the n=9 identified cases of airway obstruction (Figure 2, block 2500), a random selection of n=411 unaffected patients was incorporated, culminating in a dataset of n=420 patients (Figure 2, block 2400). The dataset set was split into a Testing Set of n=84 (block 2600) and a Training Set of N=336 (block 2700).

[0104] 3. Lastly, the training set's diversity was enhanced by randomizing data allocation between the training and validation groups on five separate folds (Figure 4), thereby increasing its generalizability to a wider range of post-operative images.

[0105] Study Limitations

[0106] First, despite the CatBoost algorithm's general resistance to overfitting, its potential remains a concern given the low number of airway-compromised patients who met the inclusion criteria for this study. Second, the high percentage of corpectomies in the patient sample may limit the generalizability of our findings as many centers more commonly perform cervical discectomies. Additionally, our model relies on post-operative imaging for predicting airway obstruction, with direct clinical manifestations of airway obstructions falling outside the scope of this model. This constraint is particularly important considering that patients with airway obstructions often experience rapid deterioration, making the acquisition of timely imaging before intervention implausible (13). Moreover, it is advised against performing imaging on patients showing signs of airway obstruction to ensure that emergent management takes place in a controlled environment.

[0107] Consequently, patients who deteriorated and required treatment before imaging could be completed are not applicable for this tool.

[0108] Conclusions:

[0109] The Al model designed in this study showed promising potential in predicting pending airway obstruction following anterior cervical spine surgery, while discerning normal postoperative changes from rapidly deteriorating complications like edema or retropharyngeal hematoma. The model demonstrated adaptability to varying radiographic environments, encompassing both radiographs and CT scans. While this tool may assist in the early detection of these life-threatening complications, further studies are necessary to validate our initial findings and integrate this modality into real-life clinical practice. To our knowledge, this is the first publication attempting to predict airway obstruction following anterior cervical spine surgery in Al-driven tools.

[0110] Additional reference is made to Figure 7. A method for detecting and / or predicting Postoperative Airway (POA) Complications Detection may include, in some embodiments, receiving data that are descriptive of at least one post-operatively acquired (POA) image of a mammalian's airway (block 7100).[Ill] In some embodiments, the method may include processing the received data, e.g., by an airways analysis engine (block 7200).

[0112] In some embodiments, the method may include, based on the processing, determining for the received data, e.g., by the airway analysis engine, a probability that the mammalian is expected to experience a POA-related complication (block 7300).

[0113] In some embodiments, the processing of the data is performed by a trained ML-module.

[0114] The methods described herein and illustrated in the accompanying diagrams shall not be construed in a limiting manner. For example, methods described herein may include additional or even fewer processes or operations in comparison to what is described herein and / or illustrated in the diagrams. In addition, method steps are not necessarily limited to the chronological order as illustrated and described herein.

[0115] Any digital computer system, apparatus, unit, device, module and / or engine exemplified herein can be configured or otherwise programmed to implement a method disclosed herein, and to the extent that the system, apparatus, module and / or engine is configured to implement such a method, it is within the scope and spirit of the disclosure. Once the system, apparatus, module, and / or engine are programmed to perform particular functions pursuant to computer readable and executable instructions from program software that implements a method disclosed herein, it in effect becomes a special purpose computer particular to embodiments of the method disclosed herein. The methods and / or processes disclosed herein may be implemented as a computer program product that may be tangibly embodied in an information carrier including, for example, in a non-transitory tangible computer-readable and / or non- transitory tangible machine-readable storage device. The computer program product may be directly loadable into an internal memory of a digital computer, comprising software code portions for performing the methods and / or processes as disclosed herein.

[0116] The methods and / or processes disclosed herein may be implemented as a computer program that may be intangibly embodied by a computer readable signal medium. A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a non-transitory computer or machine-readable storage device and that can communicate, propagate, or transport a program for use by or in connection with apparatuses, systems, platforms, methods, operations and / or processes discussed herein.

[0117] The terms "non-transitory computer-readable storage device" and "non-transitory machine- readable storage device" encompasses distribution media, intermediate storage media, execution memory of a computer, and any other medium or device capable of storing for later reading by a computerprogram implementing embodiments of a method disclosed herein. A computer program product can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by one or more communication networks.

[0118] These computer readable and executable instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable and executable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0119] The computer readable and executable instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0120] The term "engine" may comprise one or more computer modules, wherein a module may be a self-contained hardware and / or software component that interfaces with a larger system. A module may comprise a machine or machines executable instructions. A module may be embodied by a circuit or a controller programmed to cause the systems, apparatuses, and / or platforms to implement the method, process and / or operation as disclosed herein. For example, a module may be implemented as a hardware circuit comprising, e.g., custom VLSI circuits or gate arrays, an Application-Specific Integrated Circuit (ASIC), off-the-shelf semiconductors such as logic chips, transistors, and / or other discrete components. A module may also be implemented in programmable hardware devices such as field programmable gate arrays, programmable array logic, programmable logic devices and / or the like.

[0121] In the discussion, unless otherwise stated, adjectives such as "substantially" and "about" that modify a condition or relationship characteristic of a feature or features of an embodiment of the invention, are to be understood to mean that the condition or characteristic is defined to within tolerances that are acceptable for operation of the embodiment for an application for which it is intended.

[0122] Unless otherwise specified, the terms "substantially", "'about" and / or "close" with respect to a magnitude or a numerical value may imply to be within an inclusive range of -10% to +10% of the respective magnitude or value.

[0123] "Coupled with" can mean indirectly or directly "coupled with".

[0124] It is important to note that the method is not limited to those diagrams or to the corresponding descriptions. For example, the method may include additional or even fewer processes or operations in comparison to what is described in the figures. In addition, embodiments of the method are not necessarily limited to the chronological order as illustrated and described herein.

[0125] Discussions herein utilizing terms such as, for example, "processing", "computing", "calculating", "determining", "establishing", "analyzing", "checking", "estimating", "deriving", "selecting", "inferring" or the like, may refer to operation(s) and / or process(es) of a computer, a computing platform, a computing system, or other electronic computing device, that manipulate and / or transform data represented as physical (e.g., electronic) quantities within the computer's registers and / or memories into other data similarly represented as physical quantities within the computer's registers and / or memories or other information storage medium that may store instructions to perform operations and / or processes. The term determining may, where applicable, also refer to "heuristically determining".

[0126] It should be noted that where an embodiment refers to a condition of "above a threshold", this should not be construed as excluding an embodiment referring to a condition of "equal or above a threshold". Analogously, where an embodiment refers to a condition "below a threshold", this should not be construed as excluding an embodiment referring to a condition "equal or below a threshold". It is clear that should a condition be interpreted as being fulfilled if the value of a given parameter is above a threshold, then the same condition is considered as not being fulfilled if the value of the given parameter is equal or below the given threshold. Conversely, should a condition be interpreted as being fulfilled if the value of a given parameter is equal or above a threshold, then the same condition is considered as not being fulfilled if the value of the given parameter is below (and only below) the given threshold.

[0127] It should be understood that where the claims or specification refer to "a" or "an" element and / or feature, such reference is not to be construed as there being only one of that element. Hence, reference to "an element" or "at least one element" for instance may also encompass "one or more elements".

[0128] Terms used in the singular shall also include the plural, except where expressly otherwise stated or where the context otherwise requires.

[0129] In the description and claims of the present application, each of the verbs, "comprise" "include" and "have", and conjugates thereof, are used to indicate that the data portion or data portions of the verbare not necessarily a complete listing of components, elements or parts of the subject or subjects of the verb.

[0130] Unless otherwise stated, the use of the expression "and / or" between the last two members of a list of options for selection indicates that a selection of one or more of the listed options is appropriate and may be made. Further, the use of the expression "and / or" may be used interchangeably with the expressions "at least one of the following", "any one of the following" or "one or more of the following", followed by a listing of the various options.

[0131] As used herein, the phrase "A,B,C, or any combination of the aforesaid" should be interpreted as meaning all of the following: (i) A or B or C or any combination of A, B, and C, (ii) at least one of A, B, and C; (iii) A, and / or B and / or C, and (iv) A, B and / or C. Where appropriate, the phrase A, B and / or C can be interpreted as meaning A, B or C. The phrase A, B or C should be interpreted as meaning "selected from the group consisting of A, B and C". This concept is illustrated for three elements (i.e., A, B, C), but extends to fewer and greater numbers of elements (e.g., A, B, C, D, etc.).

[0132] It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments or example, may also be provided in combination with a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, example and / or option, may also be provided separately or in any suitable sub-combination or as suitable in any other described embodiment, example or option of the invention. Certain features described in the context of various embodiments, examples and / or optional implementation are not to be considered essential features of those embodiments, unless the embodiment, example and / or optional implementation is inoperative without those elements.

[0133] It is noted that the terms "in some embodiments", "according to some embodiments", "for example", "e.g.", "for instance" and "optionally" may herein be used interchangeably.

[0134] The number of elements shown in the Figures should by no means be construed as limiting and is for illustrative purposes only.

[0135] It is noted that the terms "operable to" can encompass the meaning of the term "modified or configured to". In other words, a machine "operable to" perform a task can in some embodiments, embrace a mere capability (e.g., "modified") to perform the function and, in some other embodiments, a machine that is actually made (e.g., "configured") to perform the function.

[0136] Throughout this application, various embodiments may be presented in and / or relate to a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the embodiments. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subrangesas well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the range.

[0137] The phrases "ranging / ranges between" a first indicate number and a second indicate number and "ranging / ranges from" a first indicate number "to" a second indicate number are used herein interchangeably and are meant to include the first and second indicated numbers and all the fractional and integral numerals there between.

[0138] While the invention has been described with respect to a limited number of embodiments, these should not be construed as limitations on the scope of the invention, but rather as exemplifications of some of the embodiments.

[0139] Additional Examples:

[0140] Embodiments pertain to a computer program (also: computer program product), comprising program instructions for the execution of a method comprising the following steps, when the computer program is run on a computer:

[0141] processing data descriptive of at least one post-operatively acquired image of a mammalian's airway, wherein the data is stored on the computer,

[0142] wherein the data processing results in determining a probability that the mammalian is expected to experience a POA-related complication.

[0143] In embodiments, the method comprises receiving data descriptive of at least one post-operatively acquired image of a mammalian's airway, for processing.

[0144] In embodiments, the computer program product is configured to perform the following, when run on a computer:

[0145] In embodiments, the computer program product is configured to provide, when run on a computer, an output relating to the determined probability.

[0146] In embodiments, the computer program product is configured to implement, when run on a computer, an Airway Analysis Engine that comprises a trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating airways of other mammalians.

[0147] In embodiments, the computer program product is configured to implement, when run on a computer, an ML module that was trained with image datasets comprising images obtained through one or more imaging modalities.

[0148] In embodiments, the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

[0149] In embodiments, the computer program product is configured to implement, when run on a computer, at least one procedure including a Gradient Boosting Module.

[0150] In embodiments, the computer program product is configured, when run on a computer, to predict a probability that the mammalian is expected to experience acute POA related complications.

[0151] In embodiments, the computer program product is configured, when run on a computer, to distinguish between acute and non-acute POA-related complications.

[0152] In embodiments, the computer program product is configured, when run on a computer, to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute POA-related complication.

[0153] In embodiments, the computer program product is configured, when run on a computer, to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non-acute POA-related complication.

[0154] In embodiments, the computer program product is configured, when run on a computer, to identify an anatomical location that is related to an acute and / or non-acute POA complication.

[0155] In embodiments, the computer program product is configured to implement, when run on a computer, a classifier for classifying a POA complication into one of the following classes: "acute POA complication" and "non-acute POA complication".

[0156] In embodiments, the computer program product is configured to implement, when run on a computer, an Airway Analysis Engine.

[0157] Embodiments pertain to a system for detecting and / or predicting Post-Operative Airway (POA) Complications, the system comprising:

[0158] at least one processor; and

[0159] at least one memory configured to store data and software code portions executable by the at least one processor to cause to perform:

[0160] receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway,

[0161] providing an Airway Analysis Engine with the received data; and

[0162] determining for and / or based on the received data, by the Airway Analysis Engine, a probability that the mammalian is expected to experience a POA-related complication.

[0163] In embodiments, the system (e.g., the Airway Analysis Engine) is configured to provide an output relating to the determined probability.

[0164] In embodiments, the system (e.g., Airway Analysis Engine) comprises a trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating to airways of other mammalians.

[0165] In embodiments, the ML module of the system (e.g., the Airway Analysis Engine) was trained with image datasets comprising images obtained through one or more imaging modalities.

[0166] In embodiments, the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

[0167] In embodiments, the ML module of the system (e.g., of the Airway Analysis Engine) implements at least one procedure including a Gradient Boosting Module.

[0168] In embodiments, the system (e.g., the Airway Analysis Engine) is configured to predict a probability that the mammalian is expected to experience acute POA-related complications.

[0169] In embodiments, the system (e.g., the Airway Analysis Engine) is configured to distinguish between acute and non-acute POA-related complications.

[0170] In embodiments, the system (e.g., the Airway Analysis Engine), is configured to provide an output relating to and / or indicative of a probability that the mammalian is expected to experience an acute POA- related complication.

[0171] In embodiments, the system (e.g., the Airway Analysis Engine), is configured to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non- acute POA-related complication.

[0172] In embodiments, the system (e.g., the Airway Analysis Engine), is configured to identify an anatomical location that is related to an acute and / or non-acute POA complication.

[0173] In embodiments, the system (e.g., the Airway Analysis Engine), comprises at least one classifier configured to classify a POA complication into one of the following classes: "acute POA complication" and "non-acute POA complication".

[0174] Embodiments pertain to a method for detecting and / or predicting Post-Operative Airway (POA) Complications, the method comprising:

[0175] receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway,

[0176] providing an Airway Analysis Engine with the received data; and

[0177] determining for the received data, by the Airway Analysis Engine, a probability that the mammalian is expected to experience a POA-related complication.

[0178] In embodiments, the method comprises providing an output relating to the determined probability.

[0179] In embodiments, the method comprises providing trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating airways of other mammalians.

[0180] In embodiments, the method comprises implementing an ML module that was trained with image datasets comprising images obtained through one or more imaging modalities.

[0181] In embodiments, the method comprises training an ML module with image datasets comprising images obtained through one or more imaging modalities.

[0182] In embodiments, the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

[0183] In embodiments, the method comprises implementing an ML module by a Gradient Boosting Module.

[0184] In embodiments, the method comprises predicting a probability that the mammalian is expected to experience acute POA related complications.

[0185] In embodiments, the method comprises: distinguishing between acute and non-acute POA- related complications.

[0186] In embodiments, the method comprises providing an output relating to and / or indicative of a probability that the mammalian is expected to experience an acute POA-related complication.

[0187] In embodiments, the method comprises providing an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non-acute POA-related complication.

[0188] In embodiments, the method comprises identifying an anatomical location that is related to an acute and / or non-acute POA complication.

[0189] In embodiments, the method comprises classifying a POA complication into one of the following classes: "acute POA complication" and "non-acute POA complication".

[0190] References

[0191] 1. Castiglioni I, Rundo L, Codari M, Di Leo G, Salvatore C, Interlenghi M, et al: Al applications to medical images: From machine learning to deep learning. Physica Medica 83:9-24, 2021

[0192] 2. Chen S-L, Chin S-C, Ho C-Y: Deep Learning Artificial Intelligence to Predict the Need for Tracheostomy in Patients of Deep Neck Infection Based on Clinical and Computed Tomography Findings— Preliminary Data and a Pilot Study. Diagnostics 12:1943, 2022

[0193] 3. Cheplygina V, de Bruijne M, Pluim JP: Not-so-supervised: a survey of semi-supervised, multi-instance, and transfer learning in medical image analysis. Medical image analysis 54:280-296, 2019

[0194] 4. Debkowska MP, Butterworth JF, Moore J E, Kang S, Appelbaum EN, Zuelzer WA: Acute post-operative airway complications following anterior cervical spine surgery and the role for cricothyrotomy. Journal of Spine Surgery 5:142, 2019

[0195] 5. Dorogush AV, Ershov V, Gulin A: CatBoost: gradient boosting with categorical features support. arXiv preprint arXiv:1810.11363, 2018

[0196] 6. Fountas KN, Kapsalaki EZ, Nikolakakos LG, Smisson HF, Johnston KW, Grigorian AA, et al: Anterior cervical discectomy and fusion associated complications. Spine 32:2310-2317, 2007

[0197] 7. Galbusera F, Casaroli G, Bassani T: Artificial intelligence and machine learning in spine research. JOR spine 2:el044, 2019

[0198] 8. Hancock JT, Khoshgoftaar TM: CatBoost for big data: an interdisciplinary review. Journal of big data 7:1-45, 2020

[0199] 9. Kernbach JM, Staartjes VE: Foundations of machine learning-based clinical prediction modeling: Part ii— generalization and overfitting. Machine Learning in Clinical Neuroscience:Foundations and Applications:15-21, 2022

[0200] 10. Lee S-H, Kim K-T, Suk K-S, Park K-J, Oh K-l: Effect of retropharyngeal steroid on prevertebral soft tissue swelling following anterior cervical discectomy and fusion: a prospective, randomized study. Spine 36:2286-2292, 2011

[0201] 11. Li H, Huang Y, Shen B, Ba Z, Wu D: Multivariate analysis of airway obstruction and reintubation after anterior cervical surgery: a retrospective cohort study of 774 patients. International Journal of Surgery 41:28-33, 2017

[0202] 12. Lopez CD, Boddapati V, Lombardi JM, Lee NJ, Mathew J, Danford NC, et al: Artificial learning and machine learning applications in spine surgery: a systematic review. Global Spine Journal 12:1561-1572, 2022

[0203] 13. Lynch J, Crawley S: Management of airway obstruction. BJA education 18:46, 2018

[0204] 14. Murata S, Iwasaki H, Oka H, Hashizume H, Yukawa Y, Minamide A, et al: A novel technique using ultrasonography in upper airway management after anterior cervical decompression and fusion. BMC Medical Imaging 22:67, 2022

[0205] 15. Navarro CLA, Damen JA, TakadaT, Nijman SW, Dhiman P, Ma J, et al: Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review, bmj 375, 2021

[0206] 16. O'Neill KR, Neuman B, Peters C, Riew KD: Risk factors for postoperative retropharyngeal hematoma after anterior cervical spine surgery. Spine 39:E246-E252, 2014

[0207] 17. Palumbo MA, Aidlen JP, Daniels AH, Thakur NA, Caiati J: Airway compromise due to wound hematoma following anterior cervical spine surgery. The open orthopaedics journal 6, 2012

[0208] 18. Patel AA, Schwab JH, Amanatullah DF, Divi SN: AOA Critical Issues Symposium: Shaping the Impact of Artificial Intelligence within Orthopaedic Surgery. JBJS:10.2106, 2023

[0209] 19. Rojas C, Vermess D, Bertozzi J, Whitlow J, Guidi C, Martinez C: Normal thickness and appearance of the prevertebral soft tissues on multidetector CT. American journal of neuroradiology 30:136-141, 2009

[0210] 20. Song K-J, Choi B-W, Lee D-H, Lim D-J, Oh S-Y, Kim S-S: Acute airway obstruction due to postoperative retropharyngeal hematoma after anterior cervical fusion: a retrospective analysis. Journal of Orthopaedic Surgery and Research 12:1-7, 2017

[0211] 21. Yagi K, Nakagawa H, Okazaki T, Irie S, Inagaki T, Saito O, et al: Noninfectious prevertebral soft-tissue inflammation and hematoma eliciting swelling after anterior cervical discectomy and fusion.Journal of Neurosurgery: Spine 26:459-465, 2017

Claims

CLAIMSWhat is claimed is:

1. A computer program comprising program instructions for the execution of a method comprising the following steps, when the computer program is run on a computer: processing data descriptive of at least one post-operatively acquired image of a mammalian's airway, wherein the data is stored on the computer, wherein the data processing results in determining a probability that the mammalian is expected to experience a POA-related complication.

2. The computer program of claim 1, further configured to perform the following, when run on a computer: to provide an output relating to the determined probability.

3. The computer program of any one or more of claim 1 and / or claim 2, configured to implement an Airway Analysis Engine that comprises a trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating airways of other mammalians.

4. The computer program of claim 3, wherein the ML module was trained with image datasets comprising images obtained through one or more imaging modalities.

5. The computer program of claim 4, wherein the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

6. The computer program of any one or more of the claims 3 to 5, wherein the ML module implements at least one procedure including a Gradient Boosting Module.

7. The computer program of any one or more of the preceding claims configured, when run on a computer, to predict a probability that the mammalian is expected to experience acute POA related complications.

8. The computer program of any one or more of the preceding claims configured, when run on a computer, to distinguish between acute and non-acute POA-related complications.

9. The computer program of any one or more of the preceding claims configured, when run on a computer, to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute POA-related complication.

10. The computer program of any one or more of the preceding claims configured, when run on a computer, to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non-acute POA-related complication.

11. The computer program of any one or more of the preceding claims configured, when run on a computer, to identify an anatomical location that is related to an acute and / or non-acute POA complication.

12. The computer program of any one or more of the preceding claims, implementing, when run on a computer, a classifier for classifying a POA complication into one of the following classes: "acute POA complication" and "non-acute POA complication".

13. A system for detecting and / or predicting Post-Operative Airway (POA) Complications, the system comprising: at least one processor; and at least one memory configured to store data and software code portions executable by the at least one processor to cause to perform: receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway, providing an Airway Analysis Engine with the received data; and determining for the received data, by the Airway Analysis Engine, a probability that the mammalian is expected to experience a POA-related complication.

14. The system of claim 13, further configured to provide an output relating to the determined probability.

15. The system of any one or more of the claims 13 and / or 14, wherein the Airway Analysis Engine comprises a trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating to airways of other mammalians.

16. The system of claim 15, wherein the ML module was trained with image datasets comprising images obtained through one or more imaging modalities.

17. The system of claim 16, wherein the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

18. The system of any one or more of the claims 15 to 17, wherein the ML module implements at least one procedure including a Gradient Boosting Module.

19. The system of any one or more of the claims 13 to 18, wherein the Airway Analysis Engine is configured to predict a probability that the mammalian is expected to experience acute POA related complications.

20. The system of any one or more of the claims 13 to 19, wherein the Airway Analysis Engine is configured to distinguish between acute and non-acute POA-related complications.

21. The system of any one or more of the claims 13 to 20, wherein the Airway Analysis Engine is configured to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute POA-related complication.

22. The system of any one or more of the claims 13 to 21, wherein the Airway Analysis Engine is configured to provide an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non-acute POA-related complication.

23. The system of any one or more of the claims 13 to 22, wherein the Airway Analysis Engine identifies an anatomical location that is related to an acute and / or non-acute POA complication.

24. The system of any one or more of the claims 13 to 23, wherein the Airway Analysis Engine includes a classifier for classifying a POA complication into one of the following classes: "acute POA complication" and "non-acute POA complication".

25. A method for detecting and / or predicting Post-Operative Airway (POA) Complications, the method comprising: receiving data that are descriptive of at least one post-operatively acquired image of a mammalian's airway, providing an Airway Analysis Engine with the received data; and determining for the received data, by the Airway Analysis Engine, a probability that the mammalian is expected to experience a POA-related complication.

26. The method of claim 25, further comprising: providing an output relating to the determined probability.

27. The method of any one or more of the claims 25 and / or 26, further comprising: providing trained Machine- Learning (ML) module for predicting probabilities relating to POA complications, wherein the ML module was trained with image data sets relating to airways of other mammalians.

28. The method of claim 27, wherein the ML module was trained with image datasets comprising images obtained through one or more imaging modalities.

29. The method of claim 28, wherein the one or more imaging modalities comprise X-ray imaging techniques, CT imaging techniques, or both.

30. The method of any one or more of the claims 27 and / or 29, wherein the ML module implements at least one procedure including a Gradient Boosting Module.

31. The method of any one or more of the claims 27 to 30, further comprising: predicting a probability that the mammalian is expected to experience acute POA related complications.

32. The method of any one or more of the claims 27 to 31, further comprising: distinguishing between acute and non-acute POA-related complications.

33. The method of any one or more of the claims 27 to 32, further comprising: providing an output relating to or indicative of a probability that the mammalian is expected to experience an acute POA-related complication.

34. The method of any one or more of the claims 27 to 33, further comprising: providing an output relating to or indicative of a probability that the mammalian is expected to experience an acute and non-acute POA-related complication.

35. The method of any one or more of the claims 27 to 34, further comprising: identifying an anatomical location that is related to an acute and / or non-acute POA complication.

36. The method of any one or more of the claims 27 to 35, further comprising: classifying a POA complication into one of the following classes: "acute POA complication" and "non- acute POA complication".

Images