Convolutional neural network algorithm trained on lumbar spine radiographs to predict outcomes of transforaminal epidural steroid injection for lumbosacral radicular pain from spinal stenosis.

Little is known about the therapeutic outcomes of transforaminal epidural steroid injection (TFESI) in patients with lumbosacral radicular pain due to lumbar spinal stenosis (LSS). Using lumbar spine radiographs as input data, we trained a convolutional neural network (CNN) to predict therapeutic outcomes after lumbar TFESI in patients with lumbosacral radicular pain caused by LSS. We retrospectively recruited 193 patients for this study. The lumbar spine radiographs included anteroposterior, lateral, and bilateral (left and right) oblique views. We cut each lumbar spine radiograph image into a square shape that included the vertebra corresponding to the level at which the TFESI was performed and the vertebrae juxta below and above that level. Output data were divided into "favorable outcome" (≥ 50% reduction in the numeric rating scale [NRS] score at 2 months post-TFESI) and "poor outcome" (< 50% reduction in the NRS score at 2 months post-TFESI). Using these input and output data, we developed a CNN model for predicting TFESI outcomes. The area under the curve of our model was 0.920. Its accuracy was 87.2%. Our CNN model has an excellent capacity for predicting therapeutic outcomes after lumbar TFESI in patients with lumbosacral radicular pain induced by LSS.

Deep learning algorithm for automatically measuring Cobb angle in patients with idiopathic scoliosis.

PURPOSE: The Cobb angle is a standard measurement to qualify and track the progression of scoliosis. However, the Cobb angle has high inter- and intra-observer variability. Consequently, its measurement varies with vertebrae and may even differ when the same vertebra is measured. Therefore, it is not constant and differs with measurements. This study aimed to develop a deep learning model that automatically measures the Cobb angle. The deep learning model for identifying vertebrae on spine radiographs was developed. METHODS: The dataset consisted of 297 images that were divided into two subsets for training and validation. Two hundred and twenty-seven images (76.4%) were used to train the model, while 70 images (23.6%) were used as the validation dataset. Absolut error between the measurements by the observer and developed deep learning model and intraclass correlation coefficient (ICC). RESULTS: The average absolute error between the measurements was 1.97° with a standard deviation of 1.57°. In addition, 95.9% of the angles had an absolute error of less than 5°. The ICC was calculated to assess the model's reliability further. The ICC was 0.981, indicating excellent reliability. CONCLUSIONS: The authors believe the model will be useful in clinical practice by relieving clinicians of the burden of having to manually compute the Cobb angle. Further studies are needed to enhance the accuracy and versatility of this deep learning model.

Identification of L5 vertebra on lumbar spine radiographs using deep learning.

OBJECTIVE: Deep learning is an advanced machine-learning approach that is used in several medical fields. Here, we developed a deep learning model using an object detection algorithm to identify the L5 vertebra on anteroposterior lumbar spine radiographs, and assessed its detection accuracy. METHODS: We retrospectively recruited 150 participants for whom both anteroposterior whole-spine and lumbar spine radiographs were available. The anteroposterior lumbar spine radiographs of these patients were used as the input data. Of the 150 images, 105 (70%) were randomly selected as the training set, and the remaining 45 (30%) were assigned to the validation set. YOLOv5x, of the YOLOv5 family model, was used to detect the L5 vertebra area. RESULTS: The mean average precisions 0.5 and 0.75 of the trained L5 detection model were 99.2% and 96.9%, respectively. The model's precision was 95.7% and its recall was 97.8%. Furthermore, 93.3% of the validation data were correctly detected. CONCLUSION: Our deep learning model showed an outstanding ability to identify L5 vertebrae.

Deep Learning Algorithm Trained on Oblique Cervical Radiographs to Predict Outcomes of Transforaminal Epidural Steroid Injection for Pain from Cervical Foraminal Stenosis.

INTRODUCTION: We developed a convolutional neural network (CNN) model to predict treatment outcomes of transforaminal epidural steroid injection (TFESI) for controlling cervical radicular pain due to cervical foraminal stenosis. METHODS: We retrospectively recruited 293 patients with cervical TFESI due to cervical radicular pain caused by cervical foraminal stenosis. We obtained a single oblique cervical radiograph from each patient. We cut each oblique cervical radiograph image into a square shape, including the foramen that was targeted for TFESI, the intervertebral disc, the facet joint of the corresponding level with the targeted foramen, and the pedicles of the vertebral bodies just above and below the targeted foramen. Therefore, images including the targeted foramen and structures around the targeted foramen were used as input data. A favorable outcome was defined as a ≥ 50% reduction in the numeric rating scale (NRS) score at 2 months post TFESI compared to the pretreatment NRS score. A poor outcome was defined as a < 50% reduction in the NRS score at 2 months post TFESI vs. the pretreatment score. RESULTS: The area under the curve of our developed model for predicting the treatment outcome of cervical TFESI in patients with cervical foraminal stenosis was 0.823. CONCLUSION: A CNN model trained using oblique cervical radiographs can be helpful in predicting treatment outcomes after cervical TFESI in patients with cervical foraminal stenosis. If the predictive accuracy is increased, we believe that the deep learning model using cervical radiographs as input data can be easily and widely used in clinics or hospitals.

The Use of Artificial Intelligence to Predict the Prognosis of Patients Undergoing Central Nervous System Rehabilitation: A Narrative Review.

Applications of machine learning in the healthcare field have become increasingly diverse. In this review, we investigated the integration of artificial intelligence (AI) in predicting the prognosis of patients with central nervous system disorders such as stroke, traumatic brain injury, and spinal cord injury. AI algorithms have shown promise in prognostic assessment, but challenges remain in achieving a higher prediction accuracy for practical clinical use. We suggest that accumulating more diverse data, including medical imaging and collaborative efforts among hospitals, can enhance the predictive capabilities of AI. As healthcare professionals become more familiar with AI, its role in central nervous system rehabilitation is expected to advance significantly, revolutionizing patient care.

Deep Learning Algorithm Trained on Cervical Magnetic Resonance Imaging to Predict Outcomes of Transforaminal Epidural Steroid Injection for Radicular Pain from Cervical Foraminal Stenosis.

Purpose: A convolutional neural network (CNN) is one of the representative deep learning (DL) model that is especially useful for image recognition and classification. In the current study, using cervical axial magnetic resonance imaging (MRI) data obtained prior to transforaminal epidural steroid injection (TFESI), we developed a CNN model to predict the therapeutic outcome of cervical TFESI in patients with cervical foraminal stenosis. Patients and Methods: We retrospectively recruited 288 patients with cervical foraminal stenosis who received cervical TFESI due to cervical radicular pain. We collected single T2-axial spine MR image obtained from each patient. The image showing narrowest width of the neural foramen in the level at which TFESI was performed was used for input data. A "favor outcome" was defined as a ≥ 50% reduction in the NRS score at 2 months post-TFESI vs the pretreatment NRS score. A "poor outcome" was defined as a < 50% reduction in the NRS score at 2 months post-TFESI vs the pretreatment score. Results: The area under the curve of our developed model for predicting therapeutic outcome of cervical TFESI in patients with cervical spinal stenosis was 0.801. Conclusion: We showed that a CNN model trained using cervical axial MRI could be helpful for predicting therapeutic outcome after cervical TFESI in patients with cervical foraminal stenosis.

Development of convolutional neural network model for diagnosing osteochondral lesions of the talus using anteroposterior ankle radiographs.

Deep learning is an advanced machine learning technique that is used in several medical fields to diagnose diseases and predict therapeutic outcomes. In this study, using anteroposterior ankle radiographs, we developed a convolutional neural network (CNN) model to diagnose osteochondral lesions of the talus (OLTs) using ankle radiographs as input data. We evaluated whether a CNN model trained on anteroposterior ankle radiographs could help diagnose the presence of OLT. We retrospectively collected 379 cases (OLT cases = 133, non-OLT cases = 246) of anteroposterior ankle radiographs taken at a university hospital between January 2010 and December 2020. The OLT was diagnosed using ankle magnetic resonance images of each patient. Among the 379 cases, 70% of the included data were randomly selected as the training set, 10% as the validation set, and the remaining 20% were assigned to the test set to evaluate the model performance. To accurately classify OLT and non-OLT, we cropped the area of the ankle on anteroposterior ankle radiographs, resized the image to 224 × 224, and used it as the input data. We then used the Visual Geometry Group Network model to determine whether the input image was OLT or non-OLT. The performance of the CNN model for the area under the curve, accuracy, positive predictive value, and negative predictive value on the test data were 0.774 (95% confidence interval [CI], 0.673-0.875), 81.58% (95% CI, 0.729-0.903), 80.95% (95% CI, 0.773-0.846), and 81.82% (95% CI, 0.804-0.832), respectively. A CNN model trained on anteroposterior ankle radiographs achieved meaningful accuracy in diagnosing OLT and demonstrated that it could help diagnose OLT.

Deep Learning Algorithm Trained on Lumbar Magnetic Resonance Imaging to Predict Outcomes of Transforaminal Epidural Steroid Injection for Chronic Lumbosacral Radicular Pain.

BACKGROUND: Transforaminal epidural steroid injections (TFESI) are widely used to alleviate lumbosacral radicular pain. Knowledge of the therapeutic outcomes of TFESI allows clinicians to elucidate therapeutic plans for managing lumbosacral radicular pain. Deep learning (DL) can outperform traditional machine learning techniques and learn from unstructured and perceptual data. A convolutional neural network (CNN) is a representative DL model. OBJECTIVES: We developed and investigated the accuracy of a CNN model for predicting therapeutic outcomes after TFESI for controlling chronic lumbosacral radicular pain using T2-weighted sagittal lumbar spine magnetic resonance (MR) images as input data. STUDY DESIGN: Imaging study using DL. SETTING: At the spine center of a university hospital. METHODS: We collected whole T2-weighted sagittal lumbar spine MR images from 503 patients with chronic lumbosacral radicular pain due to a herniated lumbar disc (HLD) and spinal stenosis. A "good outcome" was defined as a >= 50% reduction in the numeric rating scale (NRS-11) score at 2 months after TFESI vs the pretreatment NRS-11 score. A "poor outcome" was defined as a < 50% decrease in the NRS-11 score at 2 months after TFESI vs pretreatment. RESULTS: In the prediction of therapeutic outcomes after TFESI on the validation dataset, the area under the curve was 0.827. LIMITATIONS: Our study was limited in that we used a small amount of lumbar spine MR imaging data to train the CNN model. CONCLUSIONS: We demonstrated that a CNN model trained, using whole lumbar spine sagittal T2-weighted MR images, could help determine outcomes after TFESI in patients with chronic lumbosacral radicular pain due to an HLD or spinal stenosis.

Assessment of Fusion After Anterior Cervical Discectomy and Fusion Using Convolutional Neural Network Algorithm.

BACKGROUND: A convolutional neural network (CNN) is a deep learning (DL) model specialized for image processing, analysis, and classification. OBJECTIVE: In this study, we evaluated whether a CNN model using lateral cervical spine radiographs as input data can help assess fusion after anterior cervical discectomy and fusion (ACDF). STUDY DESIGN: Diagnostic imaging study using DL. PATIENT SAMPLE: We included 187 patients who underwent ACDF and fusion assessment with postoperative one-year computed tomography and neutral and dynamic lateral cervical spine radiographs. OUTCOME MEASURES: The performance of the CNN-based DL algorithm was evaluated in terms of accuracy and area under the curve. MATERIALS AND METHODS: Fusion or nonunion was confirmed by cervical spine computed tomography. Among the 187 patients, 69.5% (130 patients) were randomly selected as the training set, and the remaining 30.5% (57 patients) were assigned to the validation set to evaluate model performance. Radiographs of the cervical spine were used as input images to develop a CNN-based DL algorithm. The CNN algorithm used three radiographs (neutral, flexion, and extension) per patient and showed the diagnostic results as fusion (0) or nonunion (1) for each radiograph. By combining the results of the three radiographs, the final decision for a patient was determined to be fusion (fusion ≥2) or nonunion (fusion ≤1). By combining the results of the three radiographs, the final decision for a patient was determined as fusion (fusion ≥2) or nonunion (nonunion ≤1). RESULTS: The CNN-based DL model demonstrated an accuracy of 89.5% and an area under the curve of 0.889 (95% confidence interval, 0.793-0.984). CONCLUSION: The CNN algorithm for fusion assessment after ACDF trained using lateral cervical radiographs showed a relatively high diagnostic accuracy of 89.5% and is expected to be a useful aid in detecting pseudarthrosis.

Prediction of Motor Outcome of Stroke Patients Using a Deep Learning Algorithm with Brain MRI as Input Data.

BACKGROUND: Deep learning techniques can outperform traditional machine learning techniques and learn from unstructured and perceptual data, such as images and languages. We evaluated whether a convolutional neural network (CNN) model using whole axial brain T2-weighted magnetic resonance (MR) images as input data can help predict motor outcomes of the upper and lower limbs at the chronic stage in stroke patients. METHODS: We collected MR images taken at the early stage of stroke in 1,233 consecutive stroke patients. We categorized modified Brunnstrom classification (MBC) scores of ≥5 and functional ambulatory category (FAC) scores of ≥4 at 6 months after stroke as favorable outcomes in the upper and lower limbs, respectively, and MBC scores of <5 and FAC scores of <4 as poor outcomes. We applied a CNN to train the image data. Of the 1,233 patients, 70% (863 patients) were randomly selected for the training set and the remaining 30% (370 patients) were assigned to the validation set. RESULTS: In the prediction of upper limb motor function on the validation dataset, the area under the curve (AUC) was 0.768, and for lower limb motor function, the AUC was 0.828. CONCLUSION: We showed that a CNN model trained using whole-brain axial T2-weighted MR images of stroke patients would help predict upper and lower limb motor function at the chronic stage.

Practical Machine Learning Model to Predict the Recovery of Motor Function in Patients with Stroke.

BACKGROUND: Machine learning (ML) is an artificial intelligence technique in which a system learns patterns and rules from a given data. OBJECTIVES: The objective of the study was to investigate the potential of ML for predicting motor recovery in stroke patients. METHODS: We analyzed data from 833 consecutive stroke patients using 3 ML algorithms: deep neural network (DNN), random forest, and logistic regression. We created a practical ML model using the most common data measured in almost all rehabilitation hospitals as input data. Demographic and clinical data, including modified Brunnstrom classification (MBC) and functional ambulation classification (FAC), were collected when patients were transferred to the rehabilitation unit (8-30 days) and 6 months after stroke onset and were used as input data. Motor outcomes at 6 months after stroke onset of the affected upper and lower extremities were classified according to MBC and FAC, respectively. Patients with an MBC of <5 and an FAC of <4 at 6 months after stroke onset were considered to have a "poor" outcome, whereas those with MBC ≥5 and FAC ≥4 were considered to have a "good" outcome. RESULTS: The area under the curve (AUC) for the DNN model for predicting motor function was 0.836 for the upper and lower limb motor functions. For the random forest and logistic regression models, the AUCs were 0.736 and 0.790 for the upper and lower limb motor functions, respectively. The AUCs for the random forest and logistic regression models were 0.741 and 0.795 for the upper and lower limb motor functions, respectively. CONCLUSION: Although we used simple and common data that can be obtained in clinical practice as variables, our DNN algorithm was useful for predicting motor recovery of the upper and lower extremities in stroke patients during the recovery phase.

Deep Learning Analysis to Automatically Detect the Presence of Penetration or Aspiration in Videofluoroscopic Swallowing Study.

BACKGROUND: Videofluoroscopic swallowing study (VFSS) is currently considered the gold standard to precisely diagnose and quantitatively investigate dysphagia. However, VFSS interpretation is complex and requires consideration of several factors. Therefore, considering the expected impact on dysphagia management, this study aimed to apply deep learning to detect the presence of penetration or aspiration in VFSS of patients with dysphagia automatically. METHODS: The VFSS data of 190 participants with dysphagia were collected. A total of 10 frame images from one swallowing process were selected (five high-peak images and five low-peak images) for the application of deep learning in a VFSS video of a patient with dysphagia. We applied a convolutional neural network (CNN) for deep learning using the Python programming language. For the classification of VFSS findings (normal swallowing, penetration, and aspiration), the classification was determined in both high-peak and low-peak images. Thereafter, the two classifications determined through high-peak and low-peak images were integrated into a final classification. RESULTS: The area under the curve (AUC) for the validation dataset of the VFSS image for the CNN model was 0.942 for normal findings, 0.878 for penetration, and 1.000 for aspiration. The macro average AUC was 0.940 and micro average AUC was 0.961. CONCLUSION: This study demonstrated that deep learning algorithms, particularly the CNN, could be applied for detecting the presence of penetration and aspiration in VFSS of patients with dysphagia.

Prediction of Motor Function in Stroke Patients Using Machine Learning Algorithm: Development of Practical Models.

BACKGROUND: Machine learning (ML) techniques are being increasingly adopted in the medical field. OBJECTIVE: We developed a deep neural network (DNN) model and applied 2 well-known ML algorithms, logistic regression and random forest, in predicting motor outcome at 6 months after stroke. METHODS: In the present study, by using 14 input variables which are easily measured by clinicians, we developed ML models and investigated their applicability to predicting motor outcome in hemiplegic stroke patients. We retrospectively analyzed data of 1,056 consecutive stroke patients. Favorable outcomes of the upper and lower limbs were defined as a modified Brunnstrom classification (MBC) score of ≥5 (able to perform activities of daily living with the affected upper limb) and a functional ambulation category (FAC) score of ≥4 (able to walk without guardian's assistance), respectively. Poor outcomes of the upper and lower limbs were defined as MBC and FAC scores of <5 and <4, respectively. We developed 3 ML algorithms, namely the DNN, logistic regression, and random forest. RESULTS: Regarding the prediction of upper limb function, for the DNN model, the area under the curve (AUC) was 0.906. For the logistic regression and random forest models, the AUC were 0.874 and 0.882, respectively. For the prediction of lower limb function, for the DNN, logistic regression, and random forest models, the AUCs were 0.822, 0.768, and 0.802, respectively. CONCLUSIONS: We demonstrated that the ML algorithms, particularly the DNN, can be useful for predicting motor outcomes in the upper and lower limbs at 6 months after stroke.

Prediction of ambulatory outcome in patients with corona radiata infarction using deep learning.

Deep learning (DL) is an advanced machine learning approach used in diverse areas such as bioinformatics, image analysis, and natural language processing. Here, using brain magnetic resonance imaging (MRI) data obtained at early stages of infarcts, we attempted to develop a convolutional neural network (CNN) to predict the ambulatory outcome of corona radiata infarction at six months after onset. We retrospectively recruited 221 patients with corona radiata infarcts. A favorable outcome of ambulatory function was defined as a functional ambulation category (FAC) score of ≥ 4 (able to walk without a guardian's assistance), and a poor outcome of ambulatory function was defined as an FAC score of < 4. We used a CNN algorithm. Of the included subjects, 69.7% (n = 154) were assigned randomly to the training set and the remaining 30.3% (n = 67) were assigned to the validation set to measure the model performance. The area under the curve was 0.751 (95% CI 0.649-0.852) for the prediction of ambulatory function with the validation dataset using the CNN model. We demonstrated that a CNN model trained using brain MRIs captured at an early stage after corona radiata infarction could be helpful in predicting long-term ambulatory outcomes.

Machine learning analysis to predict the need for ankle foot orthosis in patients with stroke.

We investigated the potential of machine learning techniques, at an early stage after stroke, to predict the need for ankle-foot orthosis (AFO) in stroke patients. We retrospectively recruited 474 consecutive stroke patients. The need for AFO during ambulation (output variable) was classified according to the Medical Research Council (MRC) score for the ankle dorsiflexor of the affected limb. Patients with an MRC score of < 3 for the ankle dorsiflexor of the affected side were considered to require AFO, while those with scores ≥ 3 were considered not to require AFO. The following demographic and clinical data collected when patients were transferred to the rehabilitation unit (16.20 ± 6.02 days) and 6 months after stroke onset were used as input data: age, sex, type of stroke (ischemic/hemorrhagic), motor evoked potential data on the tibialis anterior muscle of the affected side, modified Brunnstrom classification, functional ambulation category, MRC score for muscle strength for shoulder abduction, elbow flexion, finger flexion, finger extension, hip flexion, knee extension, and ankle dorsiflexion of the affected side. For the deep neural network model, the area under the curve (AUC) was 0.887. For the random forest and logistic regression models, the AUC was 0.855 and 0.845, respectively. Our findings demonstrate that machine learning algorithms, particularly the deep neural network, are useful for predicting the need for AFO in stroke patients during the recovery phase.

Deep Learning Algorithm Trained on Brain Magnetic Resonance Images and Clinical Data to Predict Motor Outcomes of Patients With Corona Radiata Infarct.

The early and accurate prediction of the extent of long-term motor recovery is important for establishing specific rehabilitation strategies for stroke patients. Using clinical parameters and brain magnetic resonance images as inputs, we developed a deep learning algorithm to increase the prediction accuracy of long-term motor outcomes in patients with corona radiata (CR) infarct. Using brain magnetic resonance images and clinical data obtained soon after CR infarct, we developed an integrated algorithm to predict hand function and ambulatory outcomes of the patient 6 months after onset. To develop and evaluate the algorithm, we retrospectively recruited 221 patients with CR infarct. The area under the curve of the validation set of the integrated modified Brunnstrom classification prediction model was 0.891 with 95% confidence interval (0.814-0.967) and that of the integrated functional ambulatory category prediction model was 0.919, with 95% confidence interval (0.842-0.995). We demonstrated that an integrated algorithm trained using patients' clinical data and brain magnetic resonance images obtained soon after CR infarct can promote the accurate prediction of long-term hand function and ambulatory outcomes. Future efforts will be devoted to finding more appropriate input variables to further increase the accuracy of deep learning models in clinical applications.

Essential Medical Information for Stroke Patients Undergoing Interhospital Transfer: A Delphi Study.

OBJECTIVE: For improving the efficiency of transferring medical records for stroke patients undergoing interhospital transfer, we evaluated what is the essential medical information for physicians using the Delphi method. DESIGN: We set up an expert panel of 31 physiatrists, who participated in this study. The 32 preliminary items of the transferred medical information were listed by a physiatrist for the first round of the Delphi method, and degree of necessity for these 32 items was evaluated using a 3-point scale ("very necessary," "necessary," and "not necessary"). We considered "very necessary" and "necessary" as "agreed to its necessity." According to the Delphi method, a satisfactory level of consensus can be achieved with the agreement of significant majority (≥80%) in the expert panel. RESULTS: Based on the experts' feedback, some items were added as the preliminary items. After the second round of the Delphi method, the items confirmed to be necessary information during interhospital transfer were motor and sensory evoked potentials, Barthel Index, Mini-Mental State Examination/Global Deterioration Scale, Motor-Free Visual Perception Test, Manual Function Test, Purdue Pegboard Test, hand grip power, monofilament, 2-point discrimination test, Manual Muscle Test, Nottingham Scale, modified Brunnstrom Classification, functional ambulation category, Glasgow Coma Scale, language function test, imaging study, videofluoroscopic swallowing study, rehabilitation goal, previous medical history, comorbidity, and medication information. CONCLUSIONS: If the previously mentioned necessary items are presented together at once during interhospital transfer, physicians who receive new stroke patients can evaluate patients' medical information more easily and completely.