Ultrasound carotid plaque segmentation via image reconstruction-based self-supervised learning with limited training labels.

Carotid total plaque area (TPA) is an important contributing measurement to the evaluation of stroke risk. Deep learning provides an efficient method for ultrasound carotid plaque segmentation and TPA quantification. However, high performance of deep learning requires datasets with many labeled images for training, which is very labor-intensive. Thus, we propose an image reconstruction-based self-supervised learning algorithm (IR-SSL) for carotid plaque segmentation when few labeled images are available. IR-SSL consists of pre-trained and downstream segmentation tasks. The pre-trained task learns region-wise representations with local consistency by reconstructing plaque images from randomly partitioned and disordered images. The pre-trained model is then transferred to the segmentation network as the initial parameters in the downstream task. IR-SSL was implemented with two networks, UNet++ and U-Net, and evaluated on two independent datasets of 510 carotid ultrasound images from 144 subjects at SPARC (London, Canada) and 638 images from 479 subjects at Zhongnan hospital (Wuhan, China). Compared to the baseline networks, IR-SSL improved the segmentation performance when trained on few labeled images (n = 10, 30, 50 and 100 subjects). For 44 SPARC subjects, IR-SSL yielded Dice-similarity-coefficients (DSC) of 80.14-88.84%, and algorithm TPAs were strongly correlated (r=0.962-0.993, p < 0.001) with manual results. The models trained on the SPARC images but applied to the Zhongnan dataset without retraining achieved DSCs of 80.61-88.18% and strong correlation with manual segmentation (r=0.852-0.978, p < 0.001). These results suggest that IR-SSL could improve deep learning when trained on small labeled datasets, making it useful for monitoring carotid plaque progression/regression in clinical use and trials.

An improved self-supervised learning for EEG classification.

Motor Imagery EEG (MI-EEG) classification plays an important role in different Brain-Computer Interface (BCI) systems. Recently, deep learning has been widely used in the MI-EEG classification tasks, however this technology requires a large number of labeled training samples which are difficult to obtain, and insufficient labeled training samples will result in a degradation of the classification performance. To address the degradation problem, we investigate a Self-Supervised Learning (SSL) based MI-EEG classification method to reduce the dependence on a large number of labeled training samples. The proposed method includes a pretext task and a downstream classification one. In the pretext task, each MI-EEG is rearranged according to the temporal characteristic. A network is pre-trained using the original and rearranged MI-EEGs. In the downstream task, a MI-EEG classification network is firstly initialized by the network learned in the pretext task and then trained using a small number of the labeled training samples. A series of experiments are conducted on Data sets 1 and 2b of BCI competition IV and IVa of BCI competition III. In the case of one third of the labeled training samples, the proposed method can obtain an obvious improvement compared to the baseline network without using SSL. In the experiments under different percentages of the labeled training samples, the results show that the designed SSL strategy is effective and beneficial to improving the classification performance.

Role of the Intravoxel Incoherent Motion Diffusion Weighted Imaging in the Pre-treatment Prediction and Early Response Monitoring to Neoadjuvant Chemotherapy in Locally Advanced Breast Cancer.

The aim of this study was to explore whether intravoxel incoherent motion (IVIM) diffusion-weighted imaging (DWI) can probe pre-treatment differences or monitor early response in patients with locally advanced breast cancer receiving neoadjuvant chemotherapy (NAC). Thirty-six patients with locally advanced breast cancer were imaged using multiple-b DWI with 12 b values ranging from 0 to 1000  s/mm(2) at the baseline, and 28 patients were repeatedly scanned after the second cycle of NAC. Subjects were divided into pathologic complete response (pCR) and nonpathologic complete response (non-pCR) groups according to the surgical pathologic specimen. Parameters (D, D*, f, maximum diameter [MD] and volume [V]) before and after 2 cycles of NAC and their corresponding change (Δparameter) between pCR and non-pCR groups were compared using the Student t test or nonparametric test. The diagnostic performance of different parameters was judged by the receiver-operating characteristic curve analysis. Before NAC, the f value of pCR group was significantly higher than that of non-pCR (32.40% vs 24.40%, P = 0.048). At the end of the second cycle of NAC, the D value was significantly higher and the f value was significantly lower in pCR than that in non-pCR (P = 0.001; P = 0.015, respectively), whereas the D* value and V of the pCR group was slightly lower than that of the non-pCR group (P = 0.507; P = 0.676, respectively). ΔD was higher in pCR (-0.45 × 10(-3)  mm(2)/s) than that in non-pCR (-0.07 × 10(-3)  mm(2)/s) after 2 cycles of NAC (P < 0.001). Δf value in the pCR group was significantly higher than that in the non-pCR group (17.30% vs 5.30%, P = 0.001). There was no significant difference in ΔD* between the pCR and non-pCR group (P = 0.456). The prediction performance of ΔD value was the highest (AUC [area under the curve] = 0.924, 95% CI [95% confidence interval] = 0.759-0.990). When the optimal cut-off was set at -0.163 × 10(-3)  mm(2)/s, the values for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were up to 100% (95% CI = 66.4-100), 73.7% (95% CI = 48.8-90.9), 64.3% (95% CI = 35.6-86.0), and 100% (95% CI = 73.2-99.3), respectively. IVIM-derived parameters, especially the D and f value, showed potential value in the pre-treatment prediction and early response monitoring to NAC in locally advanced breast cancer. ΔD value had the best prediction performance for pathologic response after NAC.

[Prospective comparison of ultrasonography, helical computed tomography, magnetic resonance imaging, and endoscopic ultrasonography in assessing locoregional invasion of primary pancreatic carcinoma].

OBJECTIVE: To evaluate prospectively the efficacy and clinical significance of ultrasonography (US), helical computed tomography (HCT), endoscopic ultrasonography (EUS) and magnetic resonance imaging (MRI) in assessing locoregional invasion to the surrounding tissue or organs of primary pancreatic carcinoma. METHODS: Sixty-eight consecutive patients with pancreatic carcinoma underwent US, HCT, EUS and MRI examinations before surgical exploration. All imaging results in terms of tumor size and locoregional invasion were assessed separately by two diagnostic radiologists and compared with the surgical and pathological findings. RESULTS: Among the HCT, US, EUS and MRI examinations, EUS had the highest accuracy in assessing tumor size with a regression coefficient for the maximal and minimal diameter of 1.0250 (P = 0.0426) and 0.9873 (P < 0.0001), respectively. In the assessment of locoregional invasion to the surrounding tissue or organs, EUS also had the highest accuracy (75.8%) and sensitivity (80.0%), but MRI had the highest positive predicting value (97.4%). None of these four imaging techniques was significantly correlated with the surgical findings when analyzed by univariate logistic regression. CONCLUSION: Endoscopic ultrasonography may be the most useful imaging technique in assessing tumor size, but for assessing loco-regional invasion of primary pancreatic carcinoma, combination of more than one imaging techniques may be necessary.