Novel Approaches to Detection of Cerebral Microbleeds: Single Deep Learning Model to Achieve a Balanced Performance

PurposeCerebral microbleeds (CMBs) are considered essential indicators for the diagnosis of cerebrovascular disease and cognitive disorders. Traditionally, CMBs are manually interpreted based on criteria including the shape, diameter, and signal characteristics after an MR examination, such as susceptibility-weighted imaging or gradient echo imaging (GRE). In this paper, an efficient method for CMB detection in GRE scans is presented.Materials and methodsThe proposed framework consists of the following phases: (1) pre-processing (skull extraction), (2) the first training with the ground truth labeled using CMB, (3) the second training with the ground truth labeled with CMB mimicking the same subjects, and (4) post-processing (cerebrospinal fluid (CSF) filtering). The proposed technique was validated on a dataset of 1133 CBMs that consisted of 5284 images for training and 1737 images for testing. We applied a two-stage approach using a region-based CNN method based on You Only Look Once (YOLO) to investigate a novel CMB detection technique.ResultsThe sensitivity, precision, F1-score and false positive per person (FP_avg) were evaluated as 80.96, 60.98, 69.57 and 6.57, 59.69, 62.70, 61.16 and 4.5, 66.90, 79.75, 72.76 and 2.15 for YOLO with a single label, YOLO with double labels, and YOLO + CSF filtering, respectively, and YOLO + CSF filtering showed the highest precision performance, F1-score and lowest FP_avg.ConclusionsUsing proposed framework, we developed an optimized CMB learning model with low false positives and a balanced performance in clinical practice.     

Workflow and Outcomes of Endovascular Thrombectomy for In-Hospital Stroke a Systematic Review and Meta-Analysis

Background and PurposeAcute strokes due to large vessel occlusion in hospitalized patients is not uncommon. We performed a systematic review and meta-analysis to investigate the timing and outcome of endovascular thrombectomy (EVT) for in-hospital stroke.MethodsWe conducted a meta-analysis of clinical studies published in English until September 2020 in the MEDLINE and Cochrane databases. Studies reporting original data on the characteristics and outcomes of in-hospital stroke patients treated with EVT were included. We extracted data on the time-metrics from last known well (LKW) until reperfusion was achieved. We also collected data on procedural and functional outcomes.ResultsOut of 5093 retrieved studies, 8 were included (2,622 patients). The median age was 71.4 years and median NIHSS score on admission was 16. Patients were mostly admitted to the cardiology service (27.3%). The pooled time from LKW to recognition by staff was 72.9 min (95% CI: 40.7 to 105 min). 25.6% received IV tPA. The mean time from stroke recognition to arterial puncture was 134.5 min (95% CI: 94.9 to 174.1). Successful reperfusion occurred in 82.8.% with a pooled mean time from detection to reperfusion of 193.1 min (95% CI: 139.5 to 246.7). The 90-day independent functional outcome was reported in 42% of patients (95% CI 29 to 55%).ConclusionEVT can be performed safely and successfully for in-hospital strokes. Noticeable delays from LKW to detection and then to puncture are noted. This calls for better stroke pathways to identify and treat these patients.BackgroundStroke in hospitalized patients, referred to as in-hospital stroke (IHS), accounts for 2.2–17% of all strokes.¹ The majority of these are ischemic while intracranial hemorrhage represents 2–11% of all IHS.¹ These patients are expected to have a rapid diagnosis and treatment given the ongoing medical supervision, and therefore favorable outcomes.^1–3 However, existing studies report poor outcomes in patients with IHS with a mortality risk that exceeds that of community-onset stroke (COS): 24.7% vs 9.6%.⁴ Surviving IHS patients are also less likely to be discharged home compared to COS (27.7% vs 49.9%) and to be functionally independent at 3 months (31.0% vs 50.4%).^1–4     

Exploring the potential for age estimation using facial image sensing technology for postmortem investigation

In this study, we examine the potential for using face imaging sensing technology in place of a human forensic practitioner to estimate the age of cadavers. We used the age estimation software FieldAnalyst for Signage Ver. 6.0 AW32. To validate the usefulness of its age estimation for living subjects, images of 28 subjects were taken at three angles (+30°, 0°, and −30°) with respect to the horizontal plane, with their eyes open and closed. The highest positive correlation between mean the estimated age and the actual age (y = 1.02x − 0.35, and Spearman’s rank correlation coefficient of 0.78, P < 0.001) was obtained when the subjects had their eyes closed and the image was captured at an angle of 0°. The ages of 93% of the subjects were estimated within ±10 years of their actual ages. We then applied this procedure to 61 cadavers with their eyes closed. Facial images were taken at an angle of 0° with respect to the horizontal plane and used to estimate the ages of the cadavers. Although a positive correlation between the actual and mean estimated ages was obtained (y = 1.28x + 0.43, Pearson’s correlation coefficient of 0.69, P < 0.001), the mean estimated ages of only 39.3% of the subjects were within ±10 years of their actual ages. It appears that this technology is not accurate enough to use to determine the age of a cadaver. Therefore, medical inspectors with adequate knowledge and experience are still required for postmortem examination.     

Aging is only significant factor causing CPR-induced injuries and serious injuries

BackgroundCardiopulmonary resuscitation (CPR) sometime cause severe injuries and can affect quality of life, lead to long-term disabilities or death of the patient. The aim of this study is to identify the risk factors causing CPR-induced injuries and those of serious injuries.MethodsThis was a retrospective forensic autopsy study in a single institution. Among 885 forensic autopsies undertaken between 2011 and 2018, those in which the victim had undergone CPR immediately after cardiac arrest were recorded. ‘Serious injuries’ were defined as an Abbreviated Injury Scale (AIS) score ≥ 3. CPR-induced injuries were evaluated by three experienced forensic pathologists. With the background and history of the patient, the circumstances of cardiac arrest and risks of causing CPR-induced injuries were determined by multivariate analyses.ResultsSeventy-five victims comprised the study cohort. CPR-induced injuries were found in 52 victims (69.3%). Rib fracture was the most common (60.0%), followed by sternal fracture (37.3%), heart injury (21.3%) and liver injury (8.0%). Multivariate analysis revealed higher age to be an independent factor causing CPR-induced injuries (odds ratio [OR], 1.07, P < 0.001). Thirty-six victims had 39 serious injuries in the chest or abdomen: fracture of ≥ 3 ribs (35 cases), aortic dissection (two), lung contusion (one) and rupture of the heart (one). Multivariate analysis revealed higher age to be an independent factor causing CPR-induced serious injuries (OR, 1.09; P < 0.001).ConclusionAging was the significant factor causing CPR-induced injuries and serious injuries.     

Forensic characteristics and phylogenetic analysis of the Chinese Han population from Chongqing Municipality,Southwest China

Chongqing Han is an important southern Han group, but investigations on its paternal genetic structure are still limited. Here, we analyzed the forensic and phylogenetic characteristics of the Chongqing Han population based on 27 Y-STR and predicted Y-SNP markers. Based on AMOVA, haplogroup distribution and network analysis, we explored the genetic relationship between Chongqing Han, other Chinese groups and some southern indigenous groups (speaking Kra-Dai, Austronesian, etc).     

调强放射治疗“end to end”质量控制核查中应用针尖电离室测量小野剂量的初步研究

目的:在调强放射治疗“end to end”质量核查中,探讨应用针尖电离室对调强放射治疗小野照射进行绝对剂量测量的研究。方法:选择3省20家医院,将放有热释光剂量计TLD（距模体表面距离约7.5 cm）和胶片的国际原子能机构（IAEA）模体进行CT扫描,图像导入放射治疗计划系统（TPS）中,设计治疗计划,进行7野等中心调强照射,MLC照射野大小>2 cm×2 cm且<4 cm×4 cm。同时针尖电离室（0.015 cc）放在固体水模体距模体表面7.5 cm下进行点剂量绝对剂量验证:（1）将治疗计划中射野角度归零平移到固体水模体中进行剂量验证;（2）治疗计划射野角度不归零时为实际治疗照射方向,平移到固体水模体中进行绝对剂量验证。结果:在调强放射治疗多叶光栅小野照射的固体水模体中,用针尖电离室测量的绝对剂量与TPS计算得到的绝对剂量比较,7野照射方向归为零度时,比较偏差<5%;实际照射方向时,比较偏差<5%。验证后的计划,在IAEA模体上进行实际7野调强治疗,模体中的高剂量靶区胶片（Gafchromic EBT3 film）绝对剂量通过率均≥90%（Gamma分析:3%, 3 mm）,TLD偏差<7%。均符合IAEA提出的标准。结论:在调强放射治疗多叶光栅小野照射时,可以应用针尖电离室作为绝对剂量验证的一个方法。     

人体赤足形态特征接触面积与身高体质量相关性统计分析

背景:当前法医人类学、人体工程学及足迹学研究均发现足部各信息(如足长、足宽等)与身高、体质量呈现出显著相关性,但国内外文献对有关赤足形态特征接触面积与人体身高、体质量的量化研究尚不完善。目的:分析人体赤足形态特征接触面积与身高、体质量之间相关性,创新法医人类学、人体工程学等相关学科理论体系,为利用赤足足迹进行人身个体识别提供参考。方法:采用足迹学传统平面足迹油墨捺印方法进行样本收集,再利用MATLAB数字软件对足迹对象进行图像处理。测试对象为随机抽取身高分布在165-190 cm之间健康的青壮年男性100名,记录其身高、体质量,获取正常行走步态下的赤足形态特征接触面积并进行统计分析。试验的实施符合《赫尔辛基宣言》和中国刑事警察学院对人体研究的相关伦理要求。结果与结论:在健康青壮年男性正常弓类型(排除中断弓和膨胀弓两种弓区类型)条件下:①人体赤足形态特征接触面积与身高、体质量之间均呈显著正相关(P<0.01),其与人体身高之间关系可被应用于法医人类学、人体工程学、足迹学的身高分析;②身高相较于体质量而言,对人体赤足形态特征接触面积的影响更加显著,更具有直接相关性。     

高能光源储存环及同步加速器屏蔽计算与探讨

目的验证和探讨在高能同步辐射光源辐射屏蔽计算中半经验公式和蒙特卡罗模拟方法的一致性和适用性。方法分别采用半经验公式和蒙特卡罗模拟独立计算单电子打靶时屏蔽体外产生的周围剂量当量。结果Jenkins半经验公式计算结果与蒙特卡罗模拟结果比值范围为111%~153%,Sakano半经验公式计算结果与蒙特卡罗模拟结果比值为201%。结论对单一屏蔽材料,半经验公式可简单、保守地完成对高能电子加速器的屏蔽计算。对于多种屏蔽材料,宜采用蒙特卡罗模拟方法。     

全身照射后大鼠血浆代谢特征分析

目的 探索全身照射大鼠血浆代谢物变化,为辐射生物标志物研究提供实验依据。方法 应用代谢组学方法筛选辐射差异代谢物并进行初步验证。筛选研究中,大鼠50只（发现集）,用⁶⁰Co γ射线对大鼠进行全身照射,照射剂量为0、1、2、3、5、8 Gy;验证研究中,大鼠25只（验证集）,照射剂量为0、0.5、2.5、4、6 Gy。照射后4 h采集外周血液,分离血浆进行代谢组学分析,并测定辐射差异代谢物浓度,用代谢物组合受试者工作特征曲线（ROC）对特定剂量进行分类。结果 共筛选出8个血浆辐射差异代谢物,其中4个（胞嘧啶、己酰基肉碱、十八碳二烯酰基肉碱及棕榈酰基肉碱）在受照后发生上调,变化趋势与验证集一致,区分特定剂量样本曲线下面积（AUC）>0.75。将上述4个代谢物进行组合后,区分0 Gy与>0 Gy、<2 Gy与≥2 Gy、<5 Gy与≥5 Gy样本的AUC值分别为0.96、1和0.94。结论 大鼠受到全身照射后4 h,血浆代谢物发生显著变化,共鉴定出8个辐射差异代谢物,其中胞嘧啶、己酰基肉碱、十八碳二烯酰基肉碱及棕榈酰基肉碱具有良好的稳定性,其组合具有较高的分类准确性,有可能成为特定剂量分类的辐射敏感标志物。     

水中总α放射性测量的制源厚度及效率校准研究

目的 研究制源厚度及效率校准对水中总α放射性测量结果的影响。方法 分别采用²⁴¹Am和天然铀标准溶液加入水样制备样品源、²⁴¹Am标准物质粉末直接铺样制备标准源,研究质量厚度与探测效率的关系、质量厚度对样品源净计数率的影响、有效厚度的测定以及有效厚度法和厚源法对同一水样测量结果的影响。结果 在水样中分别加入²⁴¹Am和天然铀标准溶液测得水样的有效厚度均与经验值4 mg/cm²基本一致;水样的α计数率在质量厚度为2A~5A mg/cm²时近似处于线性增长趋势,>10A mg/cm²时计数率基本平稳不再增加（A为样品盘面积,cm²）;分别采用有效厚度法和厚源法测量同一水样的总α放射性,铺样在质量厚度为10A mg/cm²时,两种方法的结果最一致。结论 在分析水样总α放射性时,为减少因样品源质量厚度不同引起的总α计数率偏差、因标准源质量厚度不同所致的探测效率偏差,铺样厚度均选择为10A mg/cm²较为合理。