DCNAS-Net: deformation convolution and neural architecture search detection network for bone marrow oedema.

BACKGROUND: Lumbago is a global disease that affects more than 500 million people worldwide. Bone marrow oedema is one of the main causes of the condition and clinical diagnosis is mainly made by radiologists manually reviewing MRI images to determine whether oedema is present. However, the number of patients with Lumbago has risen dramatically in recent years, which has brought a huge workload to radiologists. In order to improve the efficiency of diagnosis, this paper is devoted to developing and evaluating a neural network for detecting bone marrow edema in MRI images. RELATED WORK: Inspired by the development of deep learning and image processing techniques, we design a deep learning detection algorithm specifically for the detection of bone marrow oedema from lumbar MRI images. We introduce deformable convolution, feature pyramid networks and neural architecture search modules, and redesign the existing neural networks. We explain in detail the construction of the network and illustrate the setting of the network hyperparameters. RESULTS AND DISCUSSION: The detection accuracy of our algorithm is excellent. And its accuracy of detecting bone marrow oedema reached up to 90.6[Formula: see text], an improvement of 5.7[Formula: see text] compared to the original. The recall of our neural network is 95.1[Formula: see text], and the F1-measure also reaches 92.8[Formula: see text]. And our algorithm is fast in detecting it, taking only 0.144 s per image. CONCLUSION: Extensive experiments have demonstrated that deformable convolution and aggregated feature pyramid structures are conducive for the detection of bone marrow oedema. Our algorithm has better detection accuracy and good detection speed compared to other algorithms.

Establishment and validation of an AI-aid method in the diagnosis of myocardial perfusion imaging.

BACKGROUND: This study aimed to develop and validate an AI (artificial intelligence)-aid method in myocardial perfusion imaging (MPI) to differentiate ischemia in coronary artery disease. METHODS: We retrospectively selected 599 patients who had received gated-MPI protocol. Images were acquired using hybrid SPECT-CT systems. A training set was used to train and develop the neural network and a validation set was used to test the predictive ability of the neural network. We used a learning technique named "YOLO" to carry out the training process. We compared the predictive accuracy of AI with that of physician interpreters (beginner, inexperienced, and experienced interpreters). RESULTS: Training performance showed that the accuracy ranged from 66.20% to 94.64%, the recall rate ranged from 76.96% to 98.76%, and the average precision ranged from 80.17% to 98.15%. In the ROC analysis of the validation set, the sensitivity range was 88.9 ~ 93.8%, the specificity range was 93.0 ~ 97.6%, and the AUC range was 94.1 ~ 96.1%. In the comparison between AI and different interpreters, AI outperformed the other interpreters (most P-value < 0.05). CONCLUSION: The AI system of our study showed excellent predictive accuracy in the diagnosis of MPI protocols, and therefore might be potentially helpful to aid radiologists in clinical practice and develop more sophisticated models.