Cardiotocography (CTG) is a non-invasive and important tool for diagnosing fetal distress during pregnancy. To meet the needs of intelligent fetal heart monitoring based on deep learning, this paper proposes a TWD-MOAL deep active learning algorithm based on the three-way decision (TWD) theory and multi-objective optimization Active Learning (MOAL). During the training process of a convolutional neural network (CNN) classification model, the algorithm incorporates the TWD theory to select high-confidence samples as pseudo-labeled samples in a fine-grained batch processing mode, meanwhile low-confidence samples annotated by obstetrics experts were also considered. The TWD-MOAL algorithm proposed in this paper was validated on a dataset of 16 355 prenatal CTG records collected by our group. Experimental results showed that the algorithm proposed in this paper achieved an accuracy of 80.63% using only 40% of the labeled samples, and in terms of various indicators, it performed better than the existing active learning algorithms under other frameworks. The study has shown that the intelligent fetal heart monitoring model based on TWD-MOAL proposed in this paper is reasonable and feasible. The algorithm significantly reduces the time and cost of labeling by obstetric experts and effectively solves the problem of data imbalance in CTG signal data in clinic, which is of great significance for assisting obstetrician in interpretations CTG signals and realizing intelligence fetal monitoring.
ObjectiveTo evaluate the efficacy of 30% and 50% dose photodynamic therapy (PDT) for acute central serous chorioretinopathy (CSC). MethodsA retrospective cohort study. Ninety-two eyes of 88 patients with CSC, diagnosed by best corrected visual acuity (BCVA) of logarithm of the minimum angle of resolution (logMAR), indirect ophthalmoscope, fundus colorized photography, fundus fluorescein angiography (FFA), indocyanine green angiography (ICGA)and optical coherence tomography (SD-OCT) treated with 30% and 50% doses of verteporfin respectively between March 2007 and August 2013, were enrolled. The eyes were divided into 50% dose group (49 eyes) and 30% dose group (43 eyes). The differences of age (t=-1.45), gender (χ2=0.011), eyes (χ2=2.140), mean logMAR BCVA (t=-0.40), mean central retinal thickness (CRT) and the maximum thickness of serous retinal detachment (SRD) between two groups were not significant (P > 0.05). The difference of spot size between two groups was significant (t=-2.84, P < 0.05). The follow-up time was ranged from 6 to 68 months, with a mean of (17.16 ±11.30) months. The difference of follow-up between two groups was significant (P > 0.05). The BCVA, cure rate, recurrence rate and the changes of CRT and maximum SRT were observed by SD-OCT. ResultsThe subretinal fluid (SRF) of 31 eyes (72.09%) in the 30% dose group and that of 47 eyes (95.92%) in the 50% dose PDT group was absorbed completely respectively. The cure rates in the 30% dose PDT group was significantly less than that in the 50% dose group (χ2=10.077, P=0.020). There was a significant negative association between the cure rate and spot size by Logistic regression (odds ratio > 1, P=0.040). The difference of changes in the BCVA of logMAR in 50% dose group was better than that in 30% dose group after more than 12 months after PDT (P=0.036). On 3, 6, 12 and more than 12 months after PDT, the difference in CRT in 50% dose group and 30% dose group were not statistically significant (P=0.068, 0.060, 0.082, 0.067). The difference in maximum thickness of SRD was not statically significant (P > 0.05). SRF was appeared in 8 eyes (25.81%) of 31 eyes in the 30% dose group, while SRF was appeared in 1 eye (2.13%) of 47 eyes in the 50% dose group. The recurrence rate of 30% dose group was much higher than that of 50% dose group (P < 0.05). ConclusionsFor acute CSC treated by PDT, the curative effect of 50% dose group is better than the 30% dose group.