Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation

PurposeThe purpose of this study was to report procedures developed to annotate abdominal computed tomography (CT) images from subjects without pancreatic disease that will be used as the input for deep convolutional neural networks (DNN) for development of deep learning algorithms for automatic recognition of a normal pancreas.Materials and methodsDual-phase contrast-enhanced volumetric CT acquired from 2005 to 2009 from potential kidney donors were retrospectively assessed. Four trained human annotators manually and sequentially annotated 22 structures in each datasets, then expert radiologists confirmed the annotation. For efficient annotation and data management, a commercial software package that supports three-dimensional segmentation was used.ResultsA total of 1150 dual-phase CT datasets from 575 subjects were annotated. There were 229 men and 346 women (mean age: 45 ± 12 years; range: 18–79 years). The mean intra-observer intra-subject dual-phase CT volume difference of all annotated structures was 4.27 mL (7.65%). The deep network prediction for multi-organ segmentation showed high fidelity with 89.4% and 1.29 mm in terms of mean Dice similarity coefficients and mean surface distances, respectively.ConclusionsA reliable data collection/annotation process for abdominal structures was developed. This process can be used to generate large datasets appropriate for deep learning.     

BPBSAM: Body part-specific burn severity assessment model

Background and objectiveBurns are a serious health problem leading to several thousand deaths annually, and despite the growth of science and technology, automated burns diagnosis still remains a major challenge. Researchers have been exploring visual images-based automated approaches for burn diagnosis. Noting that the impact of a burn on a particular body part can be related to the skin thickness factor, we propose a deep convolutional neural network based body part-specific burns severity assessment model (BPBSAM).MethodConsidering skin anatomy, BPBSAM estimates burn severity using body part-specific support vector machines trained with CNN features extracted from burnt body part images. Thus BPBSAM first identifies the body part of the burn images using a convolutional neural network in training of which the challenge of limited availability of burnt body part images is successfully addressed by using available larger-size datasets of non-burn images of different body parts considered (face, hand, back, and inner forearm). We prepared a rich labelled burn images datasets: BI & UBI and trained several deep learning models with existing models as pipeline for body part classification and feature extraction for severity estimation.ResultsThe proposed novel BPBSAM method classified the severity of burn from color images of burn injury with an overall average F1 score of 77.8% and accuracy of 84.85% for the test BI dataset and 87.2% and 91.53% for the UBI dataset, respectively. For burn images body part classification, the average accuracy of around 93% is achieved, and for burn severity assessment, the proposed BPBSAM outperformed the generic method in terms of overall average accuracy by 10.61%, 4.55%, and 3.03% with pipelines ResNet50, VGG16, and VGG19, respectively.ConclusionsThe main contributions of this work along with burn images labelled datasets creation is that the proposed customized body part-specific burn severity assessment model can significantly improve the performance in spite of having small burn images dataset. This highly innovative customized body part-specific approach could also be used to deal with the burn region segmentation problem. Moreover, fine tuning on pre-trained non-burn body part images network has proven to be robust and reliable.     

Learning-based parametrized model predictive control for trajectory tracking

This article is concerned with the tracking of nonequilibrium motions with model predictive control (MPC). It proposes to parametrize input and state trajectories of a dynamic system with basis functions to alleviate the computational burden in MPC. As a result of the parametrization, an optimization problem with fewer variables is obtained, and the memory requirements for storing the reference trajectories are reduced. The article also discusses the generation of feasible reference trajectories that account for the system's dynamics, as well as input and state constraints. In order to cope with repeatable disturbances, which may stem from unmodeled dynamics for example, an iterative learning procedure is included. The approach relies on a Kalman filter that identifies the repeatable disturbances based on previous trials. These are then included in the system's model available to the model predictive controller, which compensates them in subsequent trials. The proposed approach is evaluated on a quadcopter, whose task is to balance a pole, while flying a predefined trajectory.     

Evaluación del abordaje laparoscópico en la cistectomía radical desde la implantación hasta su consolidación: validación interna

Introduction and objectivesThe AEU Guidelines of 2017 consider laparoscopic and robot-assisted approaches as investigational procedures. The surgical learning curve is defined as the minimum number of cases that a surgeon has to perform in order to reproduce a technique considered as standard.The aim of this study is to analyze, within our department, the implementation of a laparoscopic radical cystectomy (LRC) program compared with a well consolidated and standardized open radical cystectomy (ORC) program.Material and methodsRetrospective cohort analysis of two cystectomy groups: LRC (n = 196) (20062016) vs. ORC (n = 96) (2003-2005).Comparison of the evolution over time of the following parameters: operative time, blood transfusion rates, resection margins, postoperative complications, hospital stay and recurrence.Three time periods have been defined for LRC: implementation (2006-09) (LRC-I), development (2010-14) (LRC-D) and consolidation (2015-16) (LRC-C); comparing each of them with the control group (ORC).The chi-square test was used for the comparison of the qualitative variables and the Anova test for the numerical ones.ResultsWhen compared to ORC, LRC presented longer operative times in LRC-I and LRC-D periods. We observed a trend toward shorter operative time than ORC in the consolidation period (LRC-C).LRC also presented lower intraoperative transfusion rates in all periods and lower postoperative rates in CRL-D and CRL-C. Overall complications in LRC-D and LRC-C were lower in LRC, having fewer major complications (Clavien≥3) in the 3 periods. A decrease in mortality and hospital stay after the LRC-I phase was also observed. These results were consolidated during the two last periods of the study.We have not observed significant differences between ORC and LRC when comparing surgical margins and recurrence rates, neither in the total series, nor in the comparison between the different periods. These results endorse the oncologic safety of LRC from the beginning of the implementation process.ConclusionsWhen compared to ORC, LRC improves perioperative transfusion rates, complications and hospital stay from its implementation period, maintaining oncological safety. On the contrary, longer operative times during implementation and development were observed. However, in our series, we observed a trend toward shorter operative times than ORC approach in the consolidation period. We have validated the laparoscopic approach for radical cystectomy in our service.     

Surgical training for anatomical endoscopic enucleation of the prostate

Anatomical endoscopic enucleation of the prostate has been proposed as a potentially superior benign prostatic hyperplasia surgery than conventional transurethral resection of prostate. However, the learning curve of the procedure is steep, hence limiting its generalisability worldwide. In order to overcome the learning curve, a proper surgical training is extremely important. This review article discussed about various aspects of surgical training in anatomical endoscopic enucleation of the prostate. In summary, no matter what surgical technique or energy modality you use, the principle of anatomical enucleation should be followed. When one starts to perform prostate enucleation, a 50 to 80 g prostate appears to be the ‘best case’ to begin with. Mentorship is extremely important to shorten the learning curve and to prevent drastic complications from the procedure. A proficiency-based progression training programme with the use of simulation and training models should be the best way to teach and learn about prostate enucleation. Enucleation ratio efficacy is the preferred measure for assessing skill level and learning curve of prostate enucleation. Morcellation efficiency is commonly used to assess morcellation performance, but the importance of safety rather than efficiency must be emphasised.     

基于深度学习情感分类模型的个性化抑郁症护理策略

目的 探讨人工智能技术在个性化抑郁症护理中的应用,实现精准护理以加速抑郁症患者的康复。方法 将60例抑郁症患者按病种和病情分层随机分配为对照组和观察组各30例.对照组采用传统护理方法;观察组采用基于深度学习情感分类模型分类后的个性化护理方案,即利用脑电图像（EEG）采集设备获取大量带标记的脑电信号数据构建EEG情感训练库,标记抑郁症类型;通过深度学习情感分类模型识别抑郁症患者EEG信号对应的情感类别;根据其识别结果,采取相应的个性化护理措施。对两组患者在住院期间进行等间隔的抑郁量化评估和护理满意率调查。结果 干预4周时,观察组汉密尔顿抑郁量表（HAMD）和自评抑郁量表（SDS）的评分显著低于对照组（均P<0.05）;观察组干预8周时的康复率高于对照组,但两组比较,差异无统计学意义（P>0.05）。结论 基于深度学习情感分类模型的个性化护理方法能显著缓减患者的抑郁程度,加快抑郁症患者的康复速度。