Review Article

Automated Bone Age Assessment: Motivation, Taxonomies, and Challenges

Table 1

A comparison of automated approaches in BAA.

ApproachesYearInventorMethodAdvantageDisadvantageReference

HANDX system1989Micheal and NelsonSegmentation and isolated Reduced observation variabilityNo reasonable accuracy [58]
PROI-based system1991Pietka et al.Segmentation of phalanges and epiphysesLow mean difference and error rateEvaluated in small scale [59]
The CASAS system1994Tanner and GibbonsBased on the TW2 RUS methodMore accurate than manual TW methodDid not work for assessing with pathological problem [62]
Middle phalanx of the third finger2002NiemeijerSegmentation of middle phalanx of third finger utilized the active shape modelAccuracy of to compared with an observerOnly covered the children between 9 and 17 years [65]
Neural network system based on linear1995Gross et al.Based on linear distance measuresBetter correlation coefficientsDid not use morphological feature [66]
Phalanges length based system1990Pietka et al.Segmentation of phalangeal length or carpalReduce subjective decisionDepends on the reference population group [67]
The third digit-three epiphyses1999Sato et al. Analyzing the bones of the third digitReasonable accuracyCovered the children between 2 and 15 years [68]
Phalanges, epiphyses, and carpals1999Hsien et al.Based on phalangeal region of interest (PROI)Low error ratePoor image processing techniques [69]
Mahmoodi model2000Mahmoodi et al.Analysis phalangeal and active shape modelReduced the risk in assessing the bone age by using the Bayes risk principleCapability of further progress [70]
Neural network classifiers using RUS and carpal2008Liu et al.Based on RUS and carpal bones Small standard deviation of the differencesHigh image processing loading [71]
Neural network based on the radius and ulna2008Tristàn-Vega and ArribasAdaptive clustering technique for segmentation Improving the bone segmentationLimited to four TW3 levels[72]
Neural network analysis based on the epiphyses and carpal1995Rucci et al.Based on the TW method and using the epiphyses and carpalUseful technique for classification in TW2 method.Neural network system started in dumb state [73]
The royal orthopaedic hospital skeletal ageing System1994Hill and PynsentBased on the 13-bone and 20-bone TW2 methodReliable method for BAASmall group of sample images [74]
BoneXpert system2009Thodberg et al.Based on shape driven and the TW RUS basedHigh accuracyRejects images in poor quality [77]
Web-based system using histogram2012Mansourvar et al.Based on the histogram techniqueRemove the segmentation methodNot reliable for images with poor image quality or abnormal bone structure [78]