Application Value of Deep Learning in Diagnosis of Cervical Spondylosis

Application Value of Deep Learning in Diagnosis of Cervical Spondylosis

Compared with the personal experience judgment of physicians, deep learning can identify something more quickly, efficiently, and accurately The identification and diagnosis of diseases save the energy of clinical and imaging doctors and achieve an individualized diagnosis of patients Diagnosis and evaluation are beneficial to the formulation of clinical surgical methods and the improvement of patients' prognoses.

This study uses deep learning technology, through the big data of cervical spondylosis cases learn, to explore the use of deep learning The feasibility of identifying and analyzing the characteristic imaging findings of cervical CT images that may be suggestive of a diagnosis It is attempted to reach the level of artificial intelligence-assisted diagnosis of cervical spondylosis.

Cervical spondylosis is due to cervical discs and intervertebral joints and their secondary changes to the adjacent spinal cord, nerve roots, and vertebrae Artery and other tissue structure, causing the corresponding clinical symptoms. Cervical spondylosis is a common and frequent disease among middle-aged and elderly people. The incidence of cervical spondylosis is 10 percent in people over 30 years of age and is higher among those who work at a desk.

Two kinds of cervical spondylosis are common: cervical spondylotic radiculopathy and cervical spondylotic myelopathy. The principles of treatment differ greatly, and the prognosis after treatment is also different. Precise preoperative imaging and clinical diagnosis are helpful to accurately estimate the effect of surgery and the prognosis of patients.

Therefore, to achieve early accurate diagnosis is to improve the cervical spine Key to the curative effect of the disease. Generally referred to as deep learning is a kind of pattern analysis method, through the combination of simple and nonlinear module for multi-level "character learning" or "said learning" model of each module to each layer (original input) data into a higher dimension, more abstract representation, when there are enough of these transformations, The initial "low level" feature representation can be transformed into "high level" feature representation, and complex learning tasks such as classification can be completed.

For the classification task, the features that are important to the classification are amplified by increasing the number of layers, and the irrelevant features are ignored.

At present, the main research methods include Convolutional Neural Networks (CNN), self-coding Neural Networks based on multi-layer neurons, and deep confidence Networks. At present, CNN is mainly used in the field of medical image recognition.

Such as AlexNet, Densenet, VGG16, Googlenet and so on. The motivation for deep learning is to build neural networks that mimic the human brain for analytical learning. It imitates the mechanisms of the human brain to interpret data, such as images, sounds, and text and has made significant progress in solving many problems that are difficult for traditional machine learning to solve.

Deep learning methods have proven to be very good at discovering complex structures in high-dimensional data.

In the field of medical and health care, artificial intelligence technology has gradually solved the problem of data interpretation of various medical images, discovered more information "hidden" behind the images, such as X-ray, CT, MRI, etc., and will gradually realize the recognition, screening, and diagnosis of a specific case, reaching the expert level of diagnosis.

At the same time, the application of artificial intelligence technology can reduce the load of doctors, reduce medical misdiagnosis, and improve the efficiency of diagnosis and treatment.

At present, deep learning has been widely used in the diagnosis and prediction of various diseases in various systems of the body.

Hemke et al. conducted machine learning research after manual segmentation of pelvic CT images and found that the deep learning model can be used to automatically recognize and segment a variety of different types of tissues in the pelvic cavity, including muscle, bone, fat, subcutaneous adipose tissue, etc.

Koichiro Yasaka et al. [8] conducted a deep learning study on the lumbar spine and found that deep learning can be used to predict the bone density of the lumbar spine and further evaluate the severity of osteoporosis.

Spampinato et al. conducted a deep learning study on hand X-rays to predict bone age.

Therefore, it is of great significance to apply deep learning, an emerging analytical technique, to the recognition of imaging features of cervical CT images. The research significance is as follows:

Accurate and efficient preoperative diagnosis of cervical spondylosis can be realized to avoid excessive medical treatment or disease delay caused by missed diagnosis and misdiagnosis. Individualized diagnosis and evaluation can be achieved for patients, which helps clinically to reasonably specify treatment strategies, and is conducive to the formulation of clinical surgical methods and the improvement of patient prognosis.

Inclusion Criteria:

1. No surgical treatment was performed before the imaging scan. Imaging report and clinical diagnosis of cervical spondylosis with or without ossification of the posterior longitudinal ligament.
2. Patients who visited the orthopedics department and emergency department of our hospital without any surgical treatment before image scan and no obvious abnormalities were found in cervical imaging CT.

Exclusion Criteria:

1. Surgery before image data acquisition;
2. Cervical cancer, tuberculosis, and fracture;
3. The lack of image data, the image is not clear.