Rockwood, Green, and Wilkins' Fractures, 10e Package

128

SECTION ONE • General Principles

had more difficulty distinguishing between complex patterns, the CNN found this task to be less challenging. The benefit of the CNN is the ability to be trained with a near-infinite number of samples, vastly more than any surgeon will experience in a lifetime of training. 7 A meta-analysis by Langerhuizen et al. examined 10 studies utilizing AI in the detection and classification of fractures, utilizing multiple modalities including XR and CT imaging. It was found across multiple studies that regarding the use of fracture detection AI had near perfect prediction (range 0.95–1.0); more impres sively, pretrained CNN models outperformed humans in both proximal humerus and hip fracture classification. However, these comparisons are based only upon one imaging modality; whereas physicians have the benefit of utilizing clinical acumen and phys ical examination correlation to improve accuracy for fractures that present more subtly on imaging, such as scaphoid fractures. 33 Tarzi et al. used a new computer-aided diagnosis tool based on vision transformers, another deep learning technique, to classify femur fractures. 61 They noted that the technique was able to pre dict 83% of the images correctly, and that the clinician’s diagnos tic improvement was 29% when supported by the technique. 61 Olczak et al. used deep learning with the OTA/AO 2018 compendium for malleolar fracture identification and found a high degree of correct classification. 43 The authors stated that their approach could be scaled up to other anatomic areas and that this approach was “an important step toward computer- assisted decision making.” 43

SUMMARY

The OTA/AO 2018 Fracture Compendium classification, advances in CT and MRI imaging, and the progression of AI presents a tremendous opportunity to accurately and system atically classify fractures with linkage to epidemiology, imag ing, fracture morphology, pathoanatomy, treatment, cost, and outcomes. Issues of reliability and validity are improved but remain. This process will require a tremendous amount of data to be further analyzed and understood. Technologic improve ments in advanced imaging have also made rapid assessment of fracture morphology and the associated soft tissue injury a reality. Classification of fractures in the future may occur so quickly and accurately that the process itself of “classifying” becomes an anachronism. It may become a small, initial step in larger analytical pathways and processes of AI with machine learn ing. The additional data could become increasingly linked with accurately classified fracture fractures, providing tremen dous power and an ability to improve clinical decision making and fracture treatment. Advances in technology and clinical research on a global scale, aligned with better analysis of frac tures, are already driving exponential progress in the classifi cation of fractures, fracture care, and the orthopaedic trauma patients we serve. Six videos of open fractures were reviewed by both attending surgeons (91) and residents (45) at multiple centers. Fractures were classified using the OTA-Open Fracture Classification. The study demonstrated moderate to excellent interobserver reliability. Acute compartment syndrome (ACS) occurred in 4.7 of 2,885 tibia fractures in 2,778 patients. OTA/AO 41-C injuries were 5.5 times more likely to advance to ACS compared with OTA/AO 41-A. OTA/AO 43 injuries were at least 4.0 times less likely to foster ACS versus OTA/AO 41 or 42 injures. A retrospective review of 673 open fractures and a prospective review of 352 open fractures from Hennepin County Medical Center, Minnesota, were performed. The infection rate was as high as 44% in type III fractures. Nineteen patients out of 512 (3.7%) who sustained an open fracture underwent amputation. Based on the review the OTA-Open Fracture Classification was superior to the Gustilo–Anderson classification system for prediction of postoperative complications and treatment outcomes in patients with open long bone fractures. Annotation

Annotated References Reference

Agel J, Evans AR, Marsh JL, et al. The OTA open fracture classification: a study of reliability and agreement. J Orthop Trauma . 2013;27(7):379–384; discussion 384–385.

Beebe MJ, Auston DA, Quade JH, et al. AO/OTA Classification is highly predictive of acute compartment syndrome after tibia fracture: a cohort of 2885 fractures. J Orthop Trauma. 2017;31(11):600–605.

Gustilo RB, Anderson JT. Prevention of infection in the treatment of one thousand and twenty-five open fractures of long bones: retrospective and prospective analyses. J Bone Joint Surg Am. 1976;58(4):453–458.

Hao J, Cuellar DO, Herbert B, et al. Does the OTA open fracture classification predict the need for limb amputation? A retrospective observational cohort study on 512 patients. J Orthop Trauma. 2016; 30(4):194–198.

Meinberg E, Agel J, Roberts C, et al. Fracture and dislocation classification compendium—2018. J Orthop Trauma. 2018;32(suppl 1):S1–S10. Neer CS II. Displaced proximal humeral fractures. I. Classification and evaluation. J Bone Joint Surg Am. 1970;52(6):1077–1089.

Introduction of the updated OTA/AO Fracture and Dislocation Classification Compendium.

Review of 300 displaced proximal humerus fractures selected at random. Precise relationships of the major segments were charted and reviewed. Distinct anatomical categories became evident and hence the Neer classification evolved.