Mitigating Diagnostic Errors in Lung Cancer Classification: A Multi-Eyes Principle to Uncertainty Quantification.

In radiology, particularly in lung cancer diagnosis, diagnostic errors and cognitive biases pose substantial challenges. These issues, including perceptual errors, interpretive mistakes, and cognitive biases such as anchoring and premature closure, are often unnoticed by experienced radiologists. To address these challenges, we propose the Multi-Eyes principle approach, which utilises multiple deep learning models to reduce bias and potentially improve diagnostic accuracy. Inspired by the Four-Eyes principle in business and cybersecurity, this methodology employs various 3D and 2D (for validation) deep learning architectures and three uncertainty quantification techniques: Monte Carlo Dropout, Deep Ensemble, and Ensemble Monte Carlo Dropout. Each model functions as an independent reviewer, similar to blind reviews. With entropy selected as the uncertainty measurement, it is averaged, followed by ensemble averaging of predictions. The effectiveness of this approach was demonstrated using the LIDC-IDRI dataset for lung cancer classification. Statistical analysis of the uncertainty's distribution reveals that with more models, uncertainty in incorrect predictions becomes more peaked and left skewed, indicating consensus on uncertainty levels. This results in accuracy and F1 score improvements, even with the best performing model, addressing overconfidence in single-model systems. These findings highlight the potential of the Multi-Eyes principle to significantly improve diagnostic performance in computer-aided diagnostic systems. Future research may explore different uncertainty quantification methods and feedback mechanisms for further advancement.

Inhibition of PDIs Downregulates Core LINC Complex Proteins, Promoting the Invasiveness of MDA-MB-231 Breast Cancer Cells in Confined Spaces In Vitro.

Eukaryotic cells tether the nucleoskeleton to the cytoskeleton via a conserved molecular bridge, called the LINC complex. The core of the LINC complex comprises SUN-domain and KASH-domain proteins that directly associate within the nuclear envelope lumen. Intra- and inter-chain disulphide bonds, along with KASH-domain protein interactions, both contribute to the tertiary and quaternary structure of vertebrate SUN-domain proteins. The significance of these bonds and the role of PDIs (protein disulphide isomerases) in LINC complex biology remains unclear. Reducing and non-reducing SDS-PAGE analyses revealed a prevalence of SUN2 homodimers in non-tumorigenic breast epithelia MCF10A cells, but not in the invasive triple-negative breast cancer MDA-MB-231 cell line. Furthermore, super-resolution microscopy revealed SUN2 staining alterations in MCF10A, but not in MDA-MB-231 nuclei, upon reducing agent exposure. While PDIA1 levels were similar in both cell lines, pharmacological inhibition of PDI activity in MDA-MB-231 cells led to SUN-domain protein down-regulation, as well as Nesprin-2 displacement from the nucleus. This inhibition also caused changes in perinuclear cytoskeletal architecture and lamin downregulation, and increased the invasiveness of PDI-inhibited MDA-MB-231 cells in space-restrictive in vitro environments, compared to untreated cells. These results emphasise the key roles of PDIs in regulating LINC complex biology, cellular architecture, biomechanics, and invasion.

Uncertainty-aware image classification on 3D CT lung.

Early detection is crucial for lung cancer to prolong the patient's survival. Existing model architectures used in such systems have shown promising results. However, they lack reliability and robustness in their predictions and the models are typically evaluated on a single dataset, making them overconfident when a new class is present. With the existence of uncertainty, uncertain images can be referred to medical experts for a second opinion. Thus, we propose an uncertainty-aware framework that includes three phases: data preprocessing and model selection and evaluation, uncertainty quantification (UQ), and uncertainty measurement and data referral for the classification of benign and malignant nodules using 3D CT images. To quantify the uncertainty, we employed three approaches; Monte Carlo Dropout (MCD), Deep Ensemble (DE), and Ensemble Monte Carlo Dropout (EMCD). We evaluated eight different deep learning models consisting of ResNet, DenseNet, and the Inception network family, all of which achieved average F1 scores above 0.832, and the highest average value of 0.845 was obtained using InceptionResNetV2. Furthermore, incorporating the UQ demonstrated significant improvement in the overall model performance. Upon evaluation of the uncertainty estimate, MCD outperforms the other UQ models except for the metric, URecall, where DE and EMCD excel, implying that they are better at identifying incorrect predictions with higher uncertainty levels, which is vital in the medical field. Finally, we show that using a threshold for data referral can greatly improve the performance further, increasing the accuracy up to 0.959.

Computed tomography-based radiomic markers are independent prognosticators of survival in advanced laryngeal cancer: a pilot study.

OBJECTIVE: Advanced laryngeal cancers are clinically complex; there is a paucity of modern decision-making models to guide tumour-specific management. This pilot study aims to identify computed tomography-based radiomic features that may predict survival and enhance prognostication. METHODS: Pre-biopsy, contrast-enhanced computed tomography scans were assembled from a retrospective cohort (n = 72) with advanced laryngeal cancers (T3 and T4). The LIFEx software was used for radiomic feature extraction. Two features: shape compacity (irregularity of tumour volume) and grey-level zone length matrix - grey-level non-uniformity (tumour heterogeneity) were selected via least absolute shrinkage and selection operator-based Cox regression and explored for prognostic potential. RESULTS: A greater shape compacity (hazard ratio 2.89) and grey-level zone length matrix - grey-level non-uniformity (hazard ratio 1.64) were significantly associated with worse 5-year disease-specific survival (p < 0.05). Cox regression models yielded a superior C-index when incorporating radiomic features (0.759) versus clinicopathological variables alone (0.655). CONCLUSIONS: Two radiomic features were identified as independent prognostic biomarkers. A multi-centre prospective study is necessary for further exploration. Integrated radiomic models may refine the treatment of advanced laryngeal cancers.

Biomedical Data Annotation: An OCT Imaging Case Study.

In ophthalmology, optical coherence tomography (OCT) is a widely used imaging modality, allowing visualisation of the structures of the eye with objective and quantitative cross-sectional three-dimensional (3D) volumetric scans. Due to the quantity of data generated from OCT scans and the time taken for an ophthalmologist to inspect for various disease pathology features, automated image analysis in the form of deep neural networks has seen success for the classification and segmentation of OCT layers and quantification of features. However, existing high-performance deep learning approaches rely on huge training datasets with high-quality annotations, which are challenging to obtain in many clinical applications. The collection of annotations from less experienced clinicians has the potential to alleviate time constraints from more senior clinicians, allowing faster data collection of medical image annotations; however, with less experience, there is the possibility of reduced annotation quality. In this study, we evaluate the quality of diabetic macular edema (DME) intraretinal fluid (IRF) biomarker image annotations on OCT B-scans from five clinicians with a range of experience. We also assess the effectiveness of annotating across multiple sessions following a training session led by an expert clinician. Our investigation shows a notable variance in annotation performance, with a correlation that depends on the clinician's experience with OCT image interpretation of DME, and that having multiple annotation sessions has a limited effect on the annotation quality.

Expansile gas kinetics for pneumatic retinopexy.

PURPOSE: To study the behaviour of expansile intravitreal gases and air used in treating rhegmatogenous retinal detachment. METHODS: A validated mathematical model of gas expansion and absorption in human eyes was used to simulate the effect of varying volumes of pure air, SF6, C2F6 and C3F8 injected into the vitreous cavity. Variation in axial length was accounted for by using three different vitreous cavity volumes to represent hypermetropic, emmetropic and myopic eyes. RESULTS: The time course of varying volumes of pure air and fluorinated gases injected into the vitreous cavity were tabulated, with calculated parameters including volume of gas, percentage gas fills and corresponding retinal contact angles at different time points. CONCLUSION: We produced a comprehensive compilation of expansive gas kinetics aiming to facilitate surgeon selection of the most suitable choice of gas and volume to use, tailored to an individual patient's clinical need.

Big data for human security: The case of COVID-19.

The COVID-19 epidemic has changed the world dramatically since societies are changing their behaviour according to the new normal, which comes along with numerous challenges and uncertainties. These uncertainties have led to instabilities in several facets of society, most notably health, economy and public order. Measures to contain the pandemic by governments have occasionally met with increasing discontent from societies and have triggered social unrest, imposing serious threats to human security. Big Data Analytics can provide a powerful force multiplier to support policy and decision makers to contain the virus while at the same time dealing with such threats to human security. This paper presents the utilisation of a big data forecasting and analytics framework and its utilisation to deal with COVID-19 triggered social unrest. The paper is an extended version of paper Cárdenas et al. (2021) presented at the 2021 International Conference on Computational Science.

Benchmarking automated detection of the retinal external limiting membrane in a 3D spectral domain optical coherence tomography image dataset of full thickness macular holes.

In this article, we present a new benchmark for the segmentation of the retinal external limiting membrane (ELM) using an image dataset of spectral domain optical coherence tomography (OCT) scans in a patient population with idiopathic full-thickness macular holes. Specifically, the dataset used contains OCT images from one eye of 107 patients with an idiopathic full-thickness macular hole. In total, the dataset contains 5243 individual 2-dimensional (2-D) OCT image slices, with each patient contributing 49 individual spectral-domain OCT tagged image slices. We display precise image-wise binary annotations to segment the ELM line. The OCT images present high variations in image contrast, motion, brightness, and speckle noise which can affect the robustness of applied algorithms, so we performed an extensive OCT imaging and annotation data quality analysis. Imaging data quality control included noise, blurriness and contrast scores, motion estimation, darkness and average pixel scores, and anomaly detection. Annotation quality was measured using gradient mapping of ELM line annotation confidence, and idiopathic full-thickness macular hole detection. Finally, we compared qualitative and quantitative results with seven state-of-the-art machine learning-based segmentation methods to identify the ELM line with an automated system.

Segmentation of Preretinal Space in Optical Coherence Tomography Images Using Deep Neural Networks.

This paper proposes an efficient segmentation of the preretinal area between the inner limiting membrane (ILM) and posterior cortical vitreous (PCV) of the human eye in an image obtained with the use of optical coherence tomography (OCT). The research was carried out using a database of three-dimensional OCT imaging scans obtained with the Optovue RTVue XR Avanti device. Various types of neural networks (UNet, Attention UNet, ReLayNet, LFUNet) were tested for semantic segmentation, their effectiveness was assessed using the Dice coefficient and compared to the graph theory techniques. Improvement in segmentation efficiency was achieved through the use of relative distance maps. We also show that selecting a larger kernel size for convolutional layers can improve segmentation quality depending on the neural network model. In the case of PVC, we obtain the effectiveness reaching up to 96.35%. The proposed solution can be widely used to diagnose vitreomacular traction changes, which is not yet available in scientific or commercial OCT imaging solutions.

Biallelic SYNE2 Missense Mutations Leading to Nesprin-2 Giant Hypo-Expression Are Associated with Intellectual Disability and Autism.

Autism spectrum disorder (ASD) is a group of neurological and developmental disabilities characterised by clinical and genetic heterogeneity. The current study aimed to expand ASD genotyping by investigating potential associations with SYNE2 mutations. Specifically, the disease-causing variants of SYNE2 in 410 trios manifesting neurodevelopmental disorders using whole-exome sequencing were explored. The consequences of the identified variants were studied at the transcript level using quantitative polymerase chain reaction (qPCR). For validation, immunofluorescence and immunoblotting were performed to analyse mutational effects at the protein level. The compound heterozygous variants of SYNE2 (NM_182914.3:c.2483T>G; p.(Val828Gly) and NM_182914.3:c.2362G>A; p.(Glu788Lys)) were identified in a 4.5-year-old male, clinically diagnosed with autism spectrum disorder, developmental delay and intellectual disability. Both variants reside within the nesprin-2 giant spectrin repeat (SR5) domain and are predicted to be highly damaging using in silico tools. Specifically, a significant reduction of nesprin-2 giant protein levels is revealed in patient cells. SYNE2 transcription and the nuclear envelope localisation of the mutant proteins was however unaffected as compared to parental control cells. Collectively, these data provide novel insights into the cardinal role of the nesprin-2 giant in neurodevelopment and suggest that the biallelic hypomorphic SYNE2 mutations may be a new cause of intellectual disability and ASD.

The application of radiomics in laryngeal cancer.

OBJECTIVES: Radiomics is the conversion of medical images into quantitative high-dimensional data. Laryngeal cancer, one of the most common head and neck cancers, has risen globally by 58.7%. CT, MRI and PET are acquired during the diagnostic process providing potential data for radiomic analysis and correlation with outcomes.This review aims to examine the applications of this technique to laryngeal cancer and the future considerations for translation into clinical practice. METHODS: A comprehensive systematic review-informed search of the MEDLINE and EMBASE databases was undertaken. Keywords "laryngeal cancer" OR "larynx" OR "larynx cancer" OR "head and neck cancer" were combined with "radiomic" OR "signature" OR "machine learning" OR "artificial intelligence". Additional articles were obtained from bibliographies using the "snowball method". RESULTS: The included studies (n = 15) demonstrated that radiomic features are significantly associated with various clinical outcomes (including stage, overall survival, treatment response, progression-free survival) and that predictive models incorporating radiomic features are superior to those that do not. Two studies demonstrated radiomics could improve laryngeal cancer staging whilst 12 studies affirmed its predictive capability for clinical outcomes. CONCLUSIONS: Radiomics has potential for improving multiple aspects of laryngeal cancer care; however, the heterogeneous cohorts and lack of data on laryngeal cancer exclusively inhibits firm conclusions. Large prospective well-designed studies in laryngeal cancer are required to progress this field. Furthermore, to implement radiomics into clinical practice, a unified research effort is required to standardise radiomics practice. ADVANCES IN KNOWLEDGE: This review has highlighted the value of radiomics in enhancing laryngeal cancer care (including staging, prognosis and predicting treatment response).

Three-dimensional data capture and analysis of intact eye lenses evidences emmetropia-associated changes in epithelial cell organization.

Organ and tissue development are highly coordinated processes; lens growth and functional integration into the eye (emmetropia) is a robust example. An epithelial monolayer covers the anterior hemisphere of the lens, and its organization is the key to lens formation and its optical properties throughout all life stages. To better understand how the epithelium supports lens function, we have developed a novel whole tissue imaging system using conventional confocal light microscopy and a specialized analysis software to produce three-dimensional maps for the epithelium of intact mouse lenses. The open source software package geometrically determines the anterior pole position, the equatorial diameter, and three-dimensional coordinates for each detected cell in the epithelium. The user-friendly cell maps, which retain global lens geometry, allow us to document age-dependent changes in the C57/BL6J mouse lens cell distribution characteristics. We evidence changes in epithelial cell density and distribution in C57/BL6J mice during the establishment of emmetropia between postnatal weeks 4-6. These epithelial changes accompany a previously unknown spheroid to lentoid shape transition of the lens as detected by our analyses. When combined with key findings from previous mouse genetic and cell biological studies, we suggest a cytoskeleton-based mechanism likely underpins these observations.

Macular hole morphology and measurement using an automated three-dimensional image segmentation algorithm.

OBJECTIVE: Full-thickness macular holes (MH) are classified principally by size, which is one of the strongest predictors of anatomical and visual success. Using a three-dimensional (3D) automated image processing algorithm, we analysed optical coherence tomography (OCT) images of 104 MH of patients, comparing MH dimensions and morphology with clinician-acquired two-dimensional measurements. METHODS AND ANALYSIS: All patients underwent a high-density central horizontal scanning OCT protocol. Two independent clinicians measured the minimum linear diameter (MLD) and maximum base diameter. OCT images were also analysed using an automated 3D segmentation algorithm which produced key parameters including the respective maximum and minimum diameter of the minimum area (MA) of the MH, as well as volume and surface area. RESULTS: Using the algorithm-derived values, MH were found to have significant asymmetry in all dimensions. The minima of the MA were typically approximately 90° to the horizontal, and differed from their maxima by 55 µm. The minima of the MA differed from the human-measured MLD by a mean of nearly 50 µm, with significant interobserver variability. The resultant differences led to reclassification using the International Vitreomacular Traction Study Group classification in a quarter of the patients (p=0.07). CONCLUSION: MH are complex shapes with significant asymmetry in all dimensions. We have shown how 3D automated analysis of MH describes their dimensions more accurately and repeatably than human assessment. This could be used in future studies investigating hole progression and outcome to help guide optimum treatments.

The multiscale top-hat tensor enables specific enhancement of curvilinear structures in 2D and 3D images.

Quantification and modelling of curvilinear structures in 2D and 3D images is a common challenge in a wide range of biomedical applications. Image enhancement is a crucial pre-processing step for curvilinear structure quantification. Many of the existing state-of-the-art enhancement approaches still suffer from contrast variations and noise. In this paper, we propose to address such problems via the use of a multiscale image processing approach, called Multiscale Top-Hat Tensor (MTHT). MTHT produces a better quality enhancement of curvilinear structures in low contrast and noisy images compared with other approaches in a range of 2D and 3D biomedical images. The proposed approach combines multiscale morphological filtering with a local tensor representation of curvilinear structure. The MTHT approach is validated on 2D and 3D synthetic and real images, and is also compared to the state-of-the-art curvilinear structure enhancement approaches. The obtained results demonstrate that the proposed approach provides high-quality curvilinear structure enhancement, allowing high accuracy segmentation and quantification in a wide range of 2D and 3D image datasets.

Multi-Scale Segmentation and Surface Fitting for Measuring 3-D Macular Holes.

Macular holes are blinding conditions, where a hole develops in the central part of retina, resulting in reduced central vision. The prognosis and treatment options are related to a number of variables, including the macular hole size and shape. High-resolution spectral domain optical coherence tomography allows precise imaging of the macular hole geometry in three dimensions, but the measurement of these by human observers is time-consuming and prone to high inter- and intra-observer variability, being characteristically measured in 2-D rather than 3-D. We introduce several novel techniques to automatically retrieve accurate 3-D measurements of the macular hole, including: surface area, base area, base diameter, top area, top diameter, height, and minimum diameter. Specifically, we introduce a multi-scale 3-D level set segmentation approach based on a state-of-the-art level set method, and we introduce novel curvature-based cutting and 3-D measurement procedures. The algorithm is fully automatic, and we validate our extracted measurements both qualitatively and quantitatively, where our results show the method to be robust across a variety of scenarios. Our automated processes are considered a significant contribution for clinical applications.

Quantitation of ER Structure and Function.

The plant endoplasmic reticulum forms a network of tubules connected by three-way junctions or sheet-like cisternae. Although the network is three-dimensional, in many plant cells, it is constrained to a thin volume sandwiched between the vacuole and plasma membrane, effectively restricting it to a 2-D planar network. The structure of the network, and the morphology of the tubules and cisternae can be automatically extracted following intensity-independent edge-enhancement and various segmentation techniques to give an initial pixel-based skeleton, which is then converted to a graph representation. Collectively, this approach yields a wealth of quantitative metrics for ER structure and can be used to describe the effects of pharmacological treatments or genetic manipulation. The software is publicly available.

Threonine 150 Phosphorylation of Keratin 5 Is Linked to Epidermolysis Bullosa Simplex and Regulates Filament Assembly and Cell Viability.

A characteristic feature of the skin blistering disease epidermolysis bullosa simplex is keratin filament (KF) network collapse caused by aggregation of the basal epidermal keratin type II (KtyII) K5 and its type I partner keratin 14 (K14). Here, we examine the role of keratin phosphorylation in KF network rearrangement and cellular functions. We detect phosphorylation of the K5 head domain residue T150 in cytoplasmic epidermolysis bullosa simplex granules containing R125C K14 mutants. Expression of phosphomimetic T150D K5 mutants results in impaired KF formation in keratinocytes. The phenotype is enhanced upon combination with other phosphomimetic K5 head domain mutations. Remarkably, introduction of T150D K5 mutants into KtyII-lacking (KtyII-/-) keratinocytes prevents keratin network formation altogether. In contrast, phosphorylation-deficient T150A K5 leads to KFs with reduced branching and turnover. Assembly of T150D K5 is arrested at the heterotetramer stage coinciding with increased heat shock protein association. Finally, reduced cell viability and elevated response to stressors is noted in T150 mutant cells. Taken together, our findings identify T150 K5 phosphorylation as an important determinant of KF network formation and function with a possible role in epidermolysis bullosa simplex pathogenesis.

Extracting 3D Parametric Curves from 2D Images of Helical Objects.

Helical objects occur in medicine, biology, cosmetics, nanotechnology, and engineering. Extracting a 3D parametric curve from a 2D image of a helical object has many practical applications, in particular being able to extract metrics such as tortuosity, frequency, and pitch. We present a method that is able to straighten the image object and derive a robust 3D helical curve from peaks in the object boundary. The algorithm has a small number of stable parameters that require little tuning, and the curve is validated against both synthetic and real-world data. The results show that the extracted 3D curve comes within close Hausdorff distance to the ground truth, and has near identical tortuosity for helical objects with a circular profile. Parameter insensitivity and robustness against high levels of image noise are demonstrated thoroughly and quantitatively.