PP-NAS: Searching for Plug-and-Play Blocks on Convolutional Neural Networks.

Multiscale features are of great importance in modern convolutional neural networks, showing consistent performance gains on numerous vision tasks. Therefore, many plug-and-play blocks are introduced to upgrade existing convolutional neural networks for stronger multiscale representation ability. However, the design of plug-and-play blocks is getting more and more complex, and these manually designed blocks are not optimal. In this work, we propose PP-NAS to develop plug-and-play blocks based on neural architecture search (NAS). Specifically, we design a new search space PPConv and develop a search algorithm consisting of one-level optimization, zero-one loss, and connection existence loss. PP-NAS minimizes the optimization gap between super-net and subarchitectures and can achieve good performance even without retraining. Extensive experiments on image classification, object detection, and semantic segmentation verify the superiority of PP-NAS over state-of-the-art CNNs (e.g., ResNet, ResNeXt, and Res2Net). Our code is available at https://github.com/ainieli/PP-NAS.

Differentiable RandAugment: Learning Selecting Weights and Magnitude Distributions of Image Transformations.

Automatic data augmentation is a technique to automatically search for strategies for image transformations, which can improve the performance of different vision tasks. RandAugment (RA), one of the most widely used automatic data augmentations, achieves great success in different scales of models and datasets. However, RA randomly selects transformations with equivalent probabilities and applies a single magnitude for all transformations, which is suboptimal for different models and datasets. In this paper, we develop Differentiable RandAugment (DRA) to learn selecting weights and magnitudes of transformations for RA. The magnitude of each transformation is modeled following a normal distribution with both learnable mean and standard deviation. We also introduce the gradient of transformations to reduce the bias in gradient estimation and KL divergence as part of the loss to reduce the optimization gap. Experiments on CIFAR-10/100 and ImageNet demonstrate the efficiency and effectiveness of DRA. Searching for only 0.95 GPU hours on ImageNet, DRA can reach a Top-1 accuracy of 78.19% with ResNet-50, which outperforms RA by 0.28% under the same settings. Transfer learning on object detection also demonstrates the power of DRA. The proposed DRA is one of the few that surpasses RA on ImageNet and has great potential to be integrated into modern training pipelines to achieve state-of-the-art performance. Our code will be made publicly available for out-of-the-box use.

Intraoperative Glioma Grading Using Neural Architecture Search and Multi-Modal Imaging.

Glioma grading during surgery can help clinical treatment planning and prognosis, but intraoperative pathological examination of frozen sections is limited by the long processing time and complex procedures. Near-infrared fluorescence imaging provides chances for fast and accurate real-time diagnosis. Recently, deep learning techniques have been actively explored for medical image analysis and disease diagnosis. However, issues of near-infrared fluorescence images, including small-scale, noise, and low-resolution, increase the difficulty of training a satisfying network. Multi-modal imaging can provide complementary information to boost model performance, but simultaneously designing a proper network and utilizing the information of multi-modal data is challenging. In this work, we propose a novel neural architecture search method DLS-DARTS to automatically search for network architectures to handle these issues. DLS-DARTS has two learnable stems for multi-modal low-level feature fusion and uses a modified perturbation-based derivation strategy to improve the performance on the area under the curve and accuracy. White light imaging and fluorescence imaging in the first near-infrared window (650-900 nm) and the second near-infrared window (1,000-1,700 nm) are applied to provide multi-modal information on glioma tissues. In the experiments on 1,115 surgical glioma specimens, DLS-DARTS achieved an area under the curve of 0.843 and an accuracy of 0.634, which outperformed manually designed convolutional neural networks including ResNet, PyramidNet, and EfficientNet, and a state-of-the-art neural architecture search method for multi-modal medical image classification. Our study demonstrates that DLS-DARTS has the potential to help neurosurgeons during surgery, showing high prospects in medical image analysis.

Excitation-based fully connected network for precise NIR-II fluorescence molecular tomography.

Fluorescence molecular tomography (FMT) is a novel imaging modality to obtain fluorescence biomarkers' three-dimensional (3D) distribution. However, the simplified mathematical model and complicated inverse problem limit it to achieving precise results. In this study, the second near-infrared (NIR-II) fluorescence imaging was adopted to mitigate tissue scattering and reduce noise interference. An excitation-based fully connected network was proposed to model the inverse process of NIR-II photon propagation and directly obtain the 3D distribution of the light source. An excitation block was embedded in the network allowing it to autonomously pay more attention to neurons related to the light source. The barycenter error was added to the loss function to improve the localization accuracy of the light source. Both numerical simulation and in vivo experiments showed the superiority of the novel NIR-II FMT reconstruction strategy over the baseline methods. This strategy was expected to facilitate the application of machine learning in biomedical research.

Real-time intraoperative glioma diagnosis using fluorescence imaging and deep convolutional neural networks.

PURPOSE: Surgery is the predominant treatment modality of human glioma but suffers difficulty on clearly identifying tumor boundaries in clinic. Conventional practice involves neurosurgeon's visual evaluation and intraoperative histological examination of dissected tissues using frozen section, which is time-consuming and complex. The aim of this study was to develop fluorescent imaging coupled with artificial intelligence technique to quickly and accurately determine glioma in real-time during surgery. METHODS: Glioma patients (N = 23) were enrolled and injected with indocyanine green for fluorescence image-guided surgery. Tissue samples (N = 1874) were harvested from surgery of these patients, and the second near-infrared window (NIR-II, 1000-1700 nm) fluorescence images were obtained. Deep convolutional neural networks (CNNs) combined with NIR-II fluorescence imaging (named as FL-CNN) were explored to automatically provide pathological diagnosis of glioma in situ in real-time during patient surgery. The pathological examination results were used as the gold standard. RESULTS: The developed FL-CNN achieved the area under the curve (AUC) of 0.945. Comparing to neurosurgeons' judgment, with the same level of specificity >80%, FL-CNN achieved a much higher sensitivity (93.8% versus 82.0%, P < 0.001) with zero time overhead. Further experiments demonstrated that FL-CNN corrected >70% of the errors made by neurosurgeons. FL-CNN was also able to rapidly predict grade and Ki-67 level (AUC 0.810 and 0.625) of tumor specimens intraoperatively. CONCLUSION: Our study demonstrates that deep CNNs are better at capturing important information from fluorescence images than surgeons' evaluation during patient surgery. FL-CNN is highly promising to provide pathological diagnosis intraoperatively and assist neurosurgeons to obtain maximum resection safely. TRIAL REGISTRATION: ChiCTR ChiCTR2000029402. Registered 29 January 2020, retrospectively registered.

Attention mechanism-based locally connected network for accurate and stable reconstruction in Cerenkov luminescence tomography.

Cerenkov luminescence tomography (CLT) is a novel and highly sensitive imaging technique, which could obtain the three-dimensional distribution of radioactive probes to achieve accurate tumor detection. However, the simplified radiative transfer equation and ill-conditioned inverse problem cause a reconstruction error. In this study, a novel attention mechanism based locally connected (AMLC) network was proposed to reduce barycenter error and improve morphological restorability. The proposed AMLC network consisted of two main parts: a fully connected sub-network for providing a coarse reconstruction result, and a locally connected sub-network based on an attention matrix for refinement. Both numerical simulations and in vivo experiments were conducted to show the superiority of the AMLC network in accuracy and stability over existing methods (MFCNN, KNN-LC network). This method improved CLT reconstruction performance and promoted the application of machine learning in optical imaging research.