Classification of mild and severe adolescent idiopathic scoliosis (AIS) from healthy subjects via a supervised learning model based on electromyogram and ground reaction force data during gait

Spine Deformity Assessment for Scoliosis Diagnostics Utilizing Image Processing Techniques: A Systematic Review

“…Six out of eighteen studies used common and conventional imaging modalities, such as X-rays, computed tomography (CT), and ultrasounds [18][19][20]28,33,34], while other researchers [21][22][23][24][25][26][27][29][30][31]35] used uncommon instrumentation, like rasterstereography, cameras, EOS imaging, scanners, and 3D laser profilemeters. Special mention goes to Sikidar et al [32], as the study did not collect data as images that used electromyogram (EMG) and ground reaction force (GRF) data.…”

“…Study [18] provided no evaluated parameter, as the study used image processing for the assessment, whereas the authors constructed new parameters or improvised from the current parameter to evaluate scoliosis. The new parameters used for scoliosis evaluation are the vertebrae as a landmark [22,24], rasterstereographic measurements [25], the center of laminae (COL) [28], the Scolioscan angle [29], a digital image-based postural assessment (DIPA) [31], the bending asymmetry index (BAI) [33], EMG and GRF [32] data, and a 3D scoliosis angle [34].…”

“…Three possible approaches have been employed for the assessment method of scoliosis diagnosis where five scholarly articles devoted to the study of artificial intelligence [18,19,25,27,32], nine articles studied image processing [20,22,24,26,28,31,33,34], and four articles studied building a system to diagnose scoliosis [21,23,29,30]. Digital image processing has become the most common form in the medical field, as it is the most efficient and cheapest method.…”

“…The number of subjects or data that participated in the studies varied with the highest being 3240 images and the lowest being 10 patients. Three studies involved a wide age range of individuals (aged between 11 and 86) [20,22,23], and ten studies focused on adolescents and middle-aged individuals (aged between 10 and 30 years old) [18,[24][25][26][27][28][29][30][31][32], while four studies did not provide the age information of the dataset. Fourteen studies provided the required details of the data included in the investigations, such as the inclusion and exclusion criteria.…”

“…Based on the reviewed articles, the outcome measures showed the reliability and validity of the method of the scoliosis assessment, which can be proven in the findings in Table 4. A total of five studies [18,25,27,30,32] were evaluated in their methodology through the application of performance classification, which involved the utilization of measures such as accuracy, sensitivity, and specificity. When classifying scoliosis, the majority of artificial intelligence studies [18,25,27,32] achieved above 70%, where the highest accuracy was 90.7%, and the lowest accuracy was 55%.…”