ABSTRACT
OBJECTIVE: This study aimed to develop and validate a deep learning (DL) algorithm for the quantitative measurement of thoracolumbar (TL) fracture features, and to evaluate its efficacy across varying levels of clinical expertise. METHODS: Using the pretrained Mask Region-Based Convolutional Neural Networks model, originally developed for vertebral body segmentation and fracture detection, we fine-tuned the model and added a new module for measuring fracture metrics-compression rate (CR), Cobb angle (CA), Gardner angle (GA), and sagittal index (SI)-from lumbar spine lateral radiographs. These metrics were derived from six-point labeling by 3 radiologists, forming the ground truth (GT). Training utilized 1,000 nonfractured and 318 fractured radiographs, while validations employed 213 internal and 200 external fractured radiographs. The accuracy of the DL algorithm in quantifying fracture features was evaluated against GT using the intraclass correlation coefficient. Additionally, 4 readers with varying expertise levels, including trainees and an attending spine surgeon, performed measurements with and without DL assistance, and their results were compared to GT and the DL model. RESULTS: The DL algorithm demonstrated good to excellent agreement with GT for CR, CA, GA, and SI in both internal (0.860, 0.944, 0.932, and 0.779, respectively) and external (0.836, 0.940, 0.916, and 0.815, respectively) validations. DL-assisted measurements significantly improved most measurement values, particularly for trainees. CONCLUSION: The DL algorithm was validated as an accurate tool for quantifying TL fracture features using radiographs. DL-assisted measurement is expected to expedite the diagnostic process and enhance reliability, particularly benefiting less experienced clinicians.
ABSTRACT
This corrects the article on p. 959 in vol. 23, PMID: 36175000.
ABSTRACT
OBJECTIVE: To investigate the agreement and reliability of estimating the volumes and normative percentiles (N%) of segmented brain regions among NeuroQuant (NQ), DeepBrain (DB), and FreeSurfer (FS) software programs, focusing on the comparison between NQ and DB. MATERIALS AND METHODS: Three-dimensional T1-weighted images of 145 participants (48 healthy participants, 50 patients with mild cognitive impairment, and 47 patients with Alzheimer's disease) from a single medical center (SMC) dataset and 130 participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset were included in this retrospective study. All images were analyzed with DB, NQ, and FS software to obtain volume estimates and N% of various segmented brain regions. We used Bland-Altman analysis, repeated measures ANOVA, reproducibility coefficient, effect size, and intraclass correlation coefficient (ICC) to evaluate inter-method agreement and reliability. RESULTS: Among the three software programs, the Bland-Altman plot showed a substantial bias, the ICC showed a broad range of reliability (0.004-0.97), and repeated-measures ANOVA revealed significant mean volume differences in all brain regions. Similarly, the volume differences of the three software programs had large effect sizes in most regions (0.73-5.51). The effect size was largest in the pallidum in both datasets and smallest in the thalamus and cerebral white matter in the SMC and ADNI datasets, respectively. N% of NQ and DB showed an unacceptably broad Bland-Altman limit of agreement in all brain regions and a very wide range of ICC values (-0.142-0.844) in most brain regions. CONCLUSION: NQ and DB showed significant differences in the measured volume and N%, with limited agreement and reliability for most brain regions. Therefore, users should be aware of the lack of interchangeability between these software programs when they are applied in clinical practice.
