Federated Partially Supervised Learning With Limited Decentralized Medical Images

Data government has played an instrumental role in securing the privacy-critical infrastructure in the medical domain and has led to an increased need of federated learning (FL). While decentralization can limit the effectiveness of standard supervised learning, the impact of decentralization on partially supervised learning remains unclear. Besides, due to data scarcity, each client may have access to only limited partially labeled data. As a remedy, this work formulates and discusses a new learning problem federated partially supervised learning (FPSL) for limited decentralized medical images with partial labels. We study the impact of decentralized partially labeled data on deep learning-based models via an exemplar of FPSL, namely, federated partially supervised learning multi-label classification. By dissecting FedAVG, a seminal FL framework, we formulate and analyze two major challenges of FPSL and propose a simple yet robust FPSL framework, FedPSL, which addresses these challenges. In particular, FedPSL contains two modules, task-dependent model aggregation and task-agnostic decoupling learning, where the first module addresses the weight assignment and the second module improves the generalization ability of the feature extractor. We provide a comprehensive empirical understanding of FSPL under data scarcity with simulated experiments. The empirical results not only indicate that FPSL is an under-explored problem with practical value but also show that the proposed FedPSL can achieve robust performance against baseline methods on data challenges such as data scarcity and domain shifts. The findings of this study also pose a new research direction towards label-efficient learning on medical images.

On the use of multi–objective evolutionary classifiers for breast cancer detection

Purpose

Breast cancer is one of the most common tumours in women, nevertheless, it is also one of the cancers that is most usually treated. As a result, early detection is critical, which can be accomplished by routine mammograms. This paper aims to describe, analyze, compare and evaluate three image descriptors involved in classifying breast cancer images from four databases.

Approach

Multi–Objective Evolutionary Algorithms (MOEAs) prove themselves as being efficient methods for selection and classification problems. This paper aims to study combinations of well–known classification objectives in order to compare the results of their application in solving very specific learning problems. The experimental results undergo empirical analysis which is supported by a statistical approach. The results are illustrated on a collection of medical image databases, but with a focus on the MOEAs’ performance in terms of several well–known measures. The databases were chosen specifically to feature reliable human annotations, so as to measure the correlation between the gold standard classifications and the various MOEA classifications.

Results

We have seen how different statistical tests rank one algorithm over the others in our set as being better. These findings are unsurprising, revealing that there is no single gold standard for comparing diverse techniques or evolutionary algorithms. Furthermore, building meta-classifiers and evaluating them using a single, favorable metric is both extremely unwise and unsatisfactory, as the impact is to skew the results.

Conclusions

The best method to address these flaws is to select the right set of objectives and criteria. Using accuracy-related objectives, for example, is directly linked to maximizing the number of true positives. If, on the other hand, accuracy is chosen as the generic metric, the primary classification goal is shifted to increasing the positively categorized data points.

Quadruple Augmented Pyramid Network for Multi-class COVID-19 Segmentation via CT

COVID-19, a new strain of coronavirus disease, has been one of the most serious and infectious disease in the world. Chest CT is essential in prognostication, diagnosing this disease, and assessing the complication. In this paper, a multi-class COVID-19 CT segmentation is proposed aiming at helping radiologists estimate the extent of effected lung volume. We utilized four augmented pyramid networks on an encoder-decoder segmentation framework. Quadruple Augmented Pyramid Network (QAP-Net) not only enable CNN capture features from variation size of CT images, but also act as spatial inter-connections and down-sampling to transfer sufficient feature information for semantic segmentation. Experimental results achieve competitive performance in segmentation with the Dice of 0.8163, which outperforms other state-of-the-art methods, demonstrating the proposed framework can segment of consolidation as well as glass, ground area via COVID-19 chest CT efficiently and accurately.

A Single RGB Camera Based Gait Analysis With A Mobile Tele-Robot For Healthcare

With the increasing awareness of high-quality life, there is a growing need for health monitoring devices running robust algorithms in home environment. Health monitoring technologies enable real-time analysis of users' health status, offering long-term healthcare support and reducing hospitalization time. The propose of this work is twofold, the software focuses on the analysis of gait, which is widely adopted for joint correction and assessing any lower limb, or spinal problem. On the hardware side, a novel marker-less gait analysis device using a low-cost RGB camera mounted on a mobile tele-robot is designed. As gait analysis with a single camera is much more challenging compared to previous works utilizing multi-cameras, a RGB-D camera or wearable sensors, we propose using vision-based human pose estimation approaches. More specifically, based on the out-put of state-of-the-art human pose estimation models, we devise measurements for four bespoke gait parameters: inversion/eversion, dorsiflexion/plantarflexion, ankle and foot progression angles. We thereby classify walking patterns into normal, supination, pronation and limp. We also illustrate how to run the proposed machine learning models in low-resource environments such as a single entry-level CPU. Experiments show that our single RGB camera method achieves competitive performance compared to multi-camera motion capture systems, at smaller hardware costs.

Validation of an automated scoring program for a digital complex figure copy task within healthy aging and stroke

OBJECTIVE

Complex Figure Copy Tasks are one of the most commonly employed neuropsychological tests. However, manual scoring of this test is time-consuming, requires training, and can then still be inconsistent between different examiners. We aimed to develop and evaluate a novel, automated method for scoring a tablet-based Figure Copy Task.

METHOD

A cohort of 261 healthy adults and 203 stroke survivors completed the digital Oxford Cognitive Screen-Plus (OCS-Plus) Figure Copy Task. Responses were independently scored by two trained human raters and by a novel automated scoring program.

RESULTS

Overall, the Automated Scoring Program was able to reliably extract and identify the separate figure elements (average sensitivity and specificity of 92.10% and 90.20%, respectively) and assigned total scores which agreed well with manual scores (intraclass correlation coefficient [ICC] = .83). Receiver Operating Curve analysis demonstrated that, compared to overall impairment categorizations based on manual scores, the Automated Scoring Program had an overall sensitivity and specificity of 80% and 93.40%, respectively (Area Under the Curve; AUC = 86.70%). Automated total scores also reliably distinguished between different clinical impairment groups with subacute stroke survivors scoring significantly worse than longer-term survivors, which in turn scored worse than neurologically healthy adults.

CONCLUSIONS

These results demonstrate that the novel Automated Scoring Program was able to reliably extract and accurately score Figure Copy Task data, even in cases where drawings were highly distorted due to comorbid fine-motor deficits. This represents a significant advancement as this novel technology can be employed to produce immediate, unbiased, and reproducible scores for Figure Copy Task responses in clinical and research environments. Key Points-Question: We aimed to develop and evaluate a novel, automated method for scoring a tablet-based Figure Copy Task.

FINDINGS

The novel Automated Scoring Program was able to reliably extract and accurately score Figure Copy Task data, even in cases where drawings were highly distorted due to comorbid fine-motor deficits.

IMPORTANCE

This represents a significant advancement as this novel technology can be employed to produce immediate, unbiased, and reproducible scores for Figure Copy Task responses in clinical and research environments. Next Steps: Trialing the Automated Scoring Program in clinical environments. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Deep learning for detection and segmentation of artefact and disease instances in gastrointestinal endoscopy

The Endoscopy Computer Vision Challenge (EndoCV) is a crowd-sourcing initiative to address eminent problems in developing reliable computer aided detection and diagnosis endoscopy systems and suggest a pathway for clinical translation of technologies. Whilst endoscopy is a widely used diagnostic and treatment tool for hollow-organs, there are several core challenges often faced by endoscopists, mainly: 1) presence of multi-class artefacts that hinder their visual interpretation, and 2) difficulty in identifying subtle precancerous precursors and cancer abnormalities. Artefacts often affect the robustness of deep learning methods applied to the gastrointestinal tract organs as they can be confused with tissue of interest. EndoCV2020 challenges are designed to address research questions in these remits. In this paper, we present a summary of methods developed by the top 17 teams and provide an objective comparison of state-of-the-art methods and methods designed by the participants for two sub-challenges: i) artefact detection and segmentation (EAD2020), and ii) disease detection and segmentation (EDD2020). Multi-center, multi-organ, multi-class, and multi-modal clinical endoscopy datasets were compiled for both EAD2020 and EDD2020 sub-challenges. The out-of-sample generalization ability of detection algorithms was also evaluated. Whilst most teams focused on accuracy improvements, only a few methods hold credibility for clinical usability. The best performing teams provided solutions to tackle class imbalance, and variabilities in size, origin, modality and occurrences by exploring data augmentation, data fusion, and optimal class thresholding techniques.

Effects of nickel content on the microstructure, microhardness and corrosion behavior of high-entropy AlCoCrFeNix alloys

In this study the effect of three different nickel concentration on the microstructure, hardness and corrosion properties of high entropy alloys (HEAs) from AlCrFeCoNi system as an alternative material for medical instruments fabrication was investigated. The analyzed HEAs were AlCrFeCoNix obtained by vacuum arc remelting from high purity raw materials and having nickel atomic ratio x = 1.0, 1.4 and 1.8. The microscopy examination revealed the dendritic morphology for the reference alloy (AlCrFeCoNi) and that the extent of the interdendritic areas increased with the concentration of nickel while Cr was more segregated in the interdendritic areas than in dendrites. Hardness values decreased as the percentage of nickel increased due to the dissolution of the precipitates in a nickel-rich matrix and consequently the formation of continuous solid solutions. The corrosion properties of the synthesized HEAs were evaluated using a potentiodynamic polarization method. The alloys were immersed in Simulated Body Fluid during one week and the corrosion parameters were recorded. The low corrosion rates, low corrosion currents and high polarization resistance attest the good stability of these HEAs in simulated biological environment indicating their possible use for surgical and dental instruments.

Segmentation of hip cartilage in compositional magnetic resonance imaging: A fast, accurate, reproducible, and clinically viable semi-automated methodology

Manual segmentation is a significant obstacle in the analysis of compositional MRI for clinical decision-making and research. Our aim was to produce a fast, accurate, reproducible, and clinically viable semi-automated method for segmentation of hip MRI. We produced a semi-automated segmentation method for cartilage segmentation of hip MRI sequences consisting of a two step process: (i) fully automated hierarchical partitioning of the data volume generated using a bespoke segmentation approach applied recursively, followed by (ii) user selection of the regions of interest using a region editor. This was applied to dGEMRIC scans at 3T taken from a prospective longitudinal study of individuals considered at high-risk of developing osteoarthritis (SibKids) which were also manually segmented for comparison. Fourteen hips were segmented both manually and using our semi-automated method. Per hip, processing time for semi-automated and manual segmentation was 10-15, and 60-120 min, respectively. Accuracy and Dice similarity coefficient (DSC) for the comparison of semi-automated and manual segmentations was 0.9886 and 0.8803, respectively. Intra-observer and inter-observer reproducibility of the semi-automated segmentation method gave an accuracy of 0.9997 and 0.9991, and DSC of 0.9726 and 0.9354, respectively. We have proposed a fast, accurate, reproducible, and clinically viable semi-automated method for segmentation of hip MRI sequences. This enables accurate anatomical and biochemical measurements to be obtained quickly and reproducibly. This is the first such method that shows clinical applicability, and could have large ramifications for the use of compositional MRI in research and clinically. © 2018 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res.

Family of boundary overlap metrics for the evaluation of medical image segmentation

All medical image segmentation algorithms need to be validated and compared, yet no evaluation framework is widely accepted within the imaging community. None of the evaluation metrics that are popular in the literature are consistent in the way they rank segmentation results: they tend to be sensitive to one or another type of segmentation error (size, location, and shape) but no single metric covers all error types. We introduce a family of metrics, with hybrid characteristics. These metrics quantify the similarity or difference of segmented regions by considering their average overlap in fixed-size neighborhoods of points on the boundaries of those regions. Our metrics are more sensitive to combinations of segmentation error types than other metrics in the existing literature. We compare the metric performance on collections of segmentation results sourced from carefully compiled two-dimensional synthetic data and three-dimensional medical images. We show that our metrics: (1) penalize errors successfully, especially those around region boundaries; (2) give a low similarity score when existing metrics disagree, thus avoiding overly inflated scores; and (3) score segmentation results over a wider range of values. We analyze a representative metric from this family and the effect of its free parameter on error sensitivity and running time.

A lab-on-chip cell-based biosensor for label-free sensing of water toxicants

This paper presents a lab-on-chip biosensor containing an enclosed fluidic cell culturing well seeded with live cells for rapid screening of toxicants in drinking water. The sensor is based on the innovative placement of the working electrode for the electrical cell-substrate impedance sensing (ECIS) technique as the top electrode of a quartz crystal microbalance (QCM) resonator. Cell damage induced by toxic water will cause a decrease in impedance, as well as an increase in the resonant frequency. For water toxicity tests, the biosensor's unique capabilities of performing two complementary measurements simultaneously (impedance and mass-sensing) will increase the accuracy of detection while decreasing the false-positive rate. Bovine aortic endothelial cells (BAECs) were used as toxicity sensing cells. The effects of the toxicants, ammonia, nicotine and aldicarb, on cells were monitored with both the QCM and the ECIS technique. The lab-on-chip was demonstrated to be sensitive to low concentrations of toxicants. The responses of BAECs to toxic samples occurred during the initial 5 to 20 minutes depending on the type of chemical and concentrations. Testing the multiparameter biosensor with aldicarb also demonstrated the hypothesis that using two different sensors to monitor the same cell monolayer provides cross validation and increases the accuracy of detection. For low concentrations of aldicarb, the variations in impedance measurements are insignificant in comparison with the shifts of resonant frequency monitored using the QCM resonator. A highly linear correlation between signal shifts and chemical concentrations was demonstrated for each toxicant.

Doing a good turn: the use of quaternions for rotation in molecular docking

Much work has been done on algorithms for structure-based drug modeling in silico, and almost all these systems have a core need for three-dimensional geometric models. The manipulation of these models, particularly their transformation from one position to another, is a substantial computational task with design questions of its own. Solid body rotation is an important part of these transformations, and we present here a careful comparison of two established techniques: Euler angles and quaternions. The relative superiority of the quaternion method when applied to molecular docking is demonstrated by practical experiment, as is the crucial importance of proper adjustment calculations in search methods.

De novo isochromosome 18p in two patients: cytogenetic diagnosis and confirmation by chromosome painting

This report concerns two patients with clinical features typical for tetrasomy 18p syndrome. Chromosomal analysis revealed a male karyotype in both cases, with an additional small metacentric marker chromosome, putatively an i(18p). Fluorescent in situ hybridization with a chromosome 18-specific paint confirmed that the marker chromosome consisted of chromosome 18 material in both cases.

A retrospective CISS hybridization analysis of a case with de novo translocation t(18;22) resulting in an 18p- syndrome

An unbalanced de novo translocation t(18;22) leading to a severely malformed liveborn girl with 18p- syndrome is described. Using the chromosomal in situ suppression (CISS) hybridization technique on 4-year-old G-banded chromosome preparations, it could be demonstrated that the translocation chromosome is composed of the long arm including the centromere of a chromosome 22 and the long arm of a chromosome 18. Consequently, the patient described here has lost the short arm including the centromere of chromosome 18. The possibility of restudying cytogenetically unsolved cases in clinical cytogenetics using older G-banded chromosome preparations with the fluorescence in situ hybridization techniques is pointed out.

A putative gene family in 15q11-13 and 16p11.2: possible implications for Prader-Willi and Angelman syndromes

The genetic defects in Prader-Willi syndrome (PWS) and Angelman syndrome (AS) map to 15q11-13. Using microdissection, we have recently isolated several DNA probes for the critical region. Here we report that microclone MN7 detects multiple loci in 15q11-13 and 16p11.2. Eight yeast artificial chromosome (YAC) clones, two genomic phage clones, and two placenta cDNA clones were isolated to analyze these loci in detail. Two of the YAC clones map to 16p. Six YAC clones and two genomic phage clones contain a total of four or five different MN7 copies, which are spread over a large distance within 15q11-13. One cDNA clone is from chromosome 15 and one is from chromosome 16. The chromosome 15 cDNA detects transcripts of 14 and 8 kilobases in various human tissues. The presence of multiple copies of the MN7 gene family in proximal 15q may conceivably be related to the instability of this region and thus to the etiology of associated disorders.

Homozygous condition for a BrdU-requiring fragile site on chromosome 12

A rare BrdU-sensitive fragile site, designated FRA12C*RQ24.2 has a relatively high frequency in the normal population. It can be demonstrated in a heterozygous and homozygous condition. There is no evidence that a phenotypic abnormality is associated with the expression of this site. A comparison with the fragile site FRA10B*RQ25.2 has revealed common features with FRA12C*RQ24.2.

Familial ring (20) chromosomal mosaicism

Ring (20) chromosomal mosaicism defined by two cell lines (one normal and the other with the ring) has been demonstrated in lymphocyte and fibroblast cultures from three members of a family through two generations. Two carriers of the ring chromosome were affected and showed the typical signs of r(20) syndrome including mental retardation, microcephaly, behavioral disorders, and epilepsy. The epilepsy is characterized by complex partial seizures sometimes evolving secondarily into generalized tonic-clonic seizures and is poorly controlled by or resistant to medical treatment. The mother of the two patients, also a carrier of ring (20) chromosomal mosaicism, was clinically and phenotypically normal and did not exhibit any signs of epilepsy. Lymphocyte and fibroblast cultures from the most severely affected sib, the proband, contained the highest percentage of cells with ring (20) chromosome and revealed the greatest instability of the ring. Though it is assumed that the ring (20) chromosome arose from terminal breakage and reunion in both arms, no loss of genetic material could be documented cytogenetically. Yet the question arises of how ring chromosomal mosaicism can be passed on. One explanation might be that a chromosome 20 predisposed to terminal lesions or breaks is transmitted from the mother to her offspring. Inherited instability of this type might lead to de novo formation of the ring.

A BrdU-requiring fragile site on chromosome 12

A BrdU-requiring fragile site, fra(12)(q24.2), on human chromosome 12 of some individuals is reported. This fragile site is inherited in a Mendelian codominant fashion and does not seem to be associated with any physical or mental abnormality in carriers. It was mostly observed as a chromatid gap: no acentric fragments, triradials or deleted chromosomes were found. The fra(12)(q24.2) was expressed in 34%-48% of metaphases in lymphocyte cultures from carriers when BrdU and FdU were added 6.5 h before harvest, while the expression ranged between 5% and 20% when the cultures were treated with BrdU alone. The fra(12)(q24.2) represents the second BrdU-requiring rare fragile site described on human chromosomes.

Trisomy 22 in a newborn with multiple malformations

A case of complete trisomy 22 in a live-born female child with multiple malformations is reported. The karyotype of the index patient had 46 chromosomes, with one chromosome 22 missing and one supranumerary metacentric chromosome. Different banding methods and in situ hybridization revealed that the extra chromosome consists of the long arms and a part of the short arms of two chromosomes 22. Our report supplies further proof that a fetus with complete trisomy 22 can occasionally survive to term, but the condition is not compatible with life over a long period.

Familial pericentric inversion of chromosome 12

A pericentric inversion in one of the chromosomes 12, found in two families living in the same region, is described. This inversion was detected during routine chromosomal analysis in two separate laboratories. The breakpoints were at 12p112 and 12q13. The inverted segment represented approximately 20% of the length of chromosome 12. Twenty nine descendants of carriers of the inversion were investigated, and the inversion was present in 23 of them. The other six descendants showed a normal karyotype. After correction for sample bias with the single selection scheme, a segregation ratio of 3:1 was estimated, indicating that the inverted chromosome 12 was preferentially transmitted. All the carriers of the inversion were phenotypically normal, without noticeable fertility disturbances.

On the development of a standard two-dimensional polypeptide map of the human X chromosome

Two-dimensional gel electrophoresis analysis of a rodent-human somatic cell hybrid containing the X as the only human chromosome reveals three polypeptides that are absent in the parental cell line. The presence of these spots in human female fibroblasts indicates their human origin. The polypeptides have molecular weights and isoelectric points of 30,000/5.8,37,000/5.4, and 57,000/4.7 and are designated as PFHG 1, PFHG 2, and PFHG 3. Comparison of their molecular characteristics with those of polypeptides assigned to the human X in two other investigations shows that some but not all polypeptides are similar. Factors are discussed that might interfere with the rapid development of a standardized polypeptide map of the human X.