Breast histology image classification is a crucial step in the early diagnosis of breast cancer. In breast pathological diagnosis, Convolutional Neural Networks (CNNs) have demonstrated great success using digitized histology slides. However, tissue classification is still challenging due to the high visual variability of the large-sized digitized samples and the lack of contextual information. In this paper, we propose a novel CNN, called Multi-level Context and Uncertainty aware ( MCUa ) dynamic deep learning ensemble model. MCUa model consists of several multi-level context-aware models to learn the spatial dependency between image patches in a layer-wise fashion. It exploits the high sensitivity to the multi-level contextual information using an uncertainty quantification component to accomplish a novel dynamic ensemble model. MCUa model has achieved a high accuracy of 98.11% on a breast cancer histology image dataset. Experimental results show the superior effectiveness of the proposed solution compared to the state-of-the-art histology classification models.

Automated grading systems using deep convolution neural networks (DCNNs) have proven their capability and potential to distinguish between different breast cancer grades using digitized histopathological images. In digital breast pathology, it is vital to measure how confident a DCNN is in grading using a machine-confidence metric, especially with the presence of major computer vision challenging problems such as the high visual variability of the images. Such a quantitative metric can be employed not only to improve the robustness of automated systems, but also to assist medical professionals in identifying complex cases. In this paper, we propose Entropy-based Elastic Ensemble of DCNN models (3E-Net) for grading invasive breast carcinoma microscopy images which provides an initial stage of explainability (using an uncertainty-aware mechanism adopting entropy). Our proposed model has been designed in a way to (1) exclude images that are less sensitive and highly uncertain to our ensemble model and (2) dynamically grade the non-excluded images using the certain models in the ensemble architecture. We evaluated two variations of 3E-Net on an invasive breast carcinoma dataset and we achieved grading accuracy of 96.15% and 99.50%.

Colorectal cancer (CRC) is the second leading cause of cancer-related mortality. Precise diagnosis of CRC plays a crucial role in increasing patient survival rates and formulating effective treatment strategies. Deep learning algorithms have demonstrated remarkable proficiency in the precise categorization of histopathology images. In this paper, we introduce a novel deep learning model, termed DeepCon which incorporates the divide-and-conquer principle into the classification task. DeepCon has been methodically conceived to scrutinize the influence of acquired composition on the learning process, with a specific application to the classification of histology images related to CRC. Our model harnesses pre-trained networks to extract features from both the source and target domains, employing a two-stage transfer learning approach encompassing multiple loss functions. Our transfer learning strategy exploits a learned composition of decomposed images to enhance the transferability of extracted features. The efficacy of the proposed model was assessed using a clinically valid dataset of 5000 CRC images. The experimental results reveal that DeepCon when coupled with the Xception network as the backbone model and subjected to extensive fine-tuning, achieved a remarkable accuracy rate of 98.4% and an F1 score of 98.4%

Background: Placental dysfunction remains a leading cause of stillbirth and neonatal morbidity, yet current monitoring tools provide only indirect and intermittent measures of fetoplacental wellbeing. Near-infrared spectroscopy (NIRS) offers non-invasive, continuous monitoring of tissue oxygenation and metabolism. Objectives: To develop a wearable NIRS system for placental monitoring (FetalSenseM v1 – FSM v1), investigate optical markers of placental oxygenation and metabolism in a population at high risk of adverse pregnancy outcomes such as stillbirth, and to apply machine learning analysis to develop a model for pregnancy outcome prediction. Study design: In this prospective observational study, women with high-risk singleton pregnancies underwent antenatal placental NIRS monitoring for over 40 minutes. FSM v1 incorporates dual-source–detector separations and multiwavelength light sources to derive absolute placental oxygen saturation (PltO ₂ ) and relative cytochrome-c-oxidase (oxCCO) changes. FSM was placed on the abdominal wall following an ultrasound scan locating the placental position. Monte Carlo simulations were performed to estimate placental sensitivity, and a minimum placental sensitivity (MPS) threshold (>5%) defined a physiologically refined sub-cohort. Outcomes were classified using the In Utero near-miss criteria for stillbirth. Machine learning (ML) analysis evaluated 11 classifiers using nested stratified 5 × 4 cross-validation (5 outer folds for performance estimation and 4 inner folds for hyperparameter tuning). Results: Seventy monitoring sessions from 58 participants were completed across gestational ages (25+2-41+1 weeks’ gestation); 33 recordings from 30 participants met MPS criteria. In the full cohort, mean PltO ₂ was 49.8% and was not related to gestational age or poor outcome based on near-miss stillbirth criteria. In the MPS sub-cohort, higher PltO ₂ was observed in severe fetal growth restriction (FGR) and lower PltO ₂ in gestational diabetes (both p=0.04). Hemodynamic-metabolic coupling (HbD-oxCCO semblance) was increased in severe FGR (p=0.0002). The best performing ML model (SVM) achieved a balanced accuracy of 78%, a recall (sensitivity) of 72% and a specificity of 84% under 5 x 4 nested cross-validation using the top 50 features. Feature importance analysis identified oxCCO-derived and haemodynamic–metabolic coupling features as dominant predictors, whereas static PltO _₂ was non-discriminatory. Conclusion: We describe the first wearable NIRS device to provide simultaneous non-invasive placental haemodynamic and metabolic monitoring. While static oxygenation indices lacked predictive value, ML analysis applied to dynamic NIRS features yielded accurate pregnancy outcome prediction, with metabolic signals emerging as key drivers. These findings support further development of wearable placental NIRS integrated with advanced analytics for antenatal surveillance. Condensation page: Tweetable statement: A wearable placental near-infrared spectroscopy device enabled real-time monitoring of placental oxygenation and metabolism; machine learning of dynamic signals predicted risk of adverse pregnancy outcomes with 78% balanced accuracy. At a Glance: A. Why was the study conducted? Current antenatal surveillance for assessment of fetal wellbeing is suboptimal. We evaluated a wearable near-infrared spectroscopy device for real-time placental monitoring and outcome prediction B. What are the key findings? Machine learning applied to dynamic haemodynamic and metabolic optical signals of placental function identified pregnancies at risk with 78% balance accuracy Placental oxygenation was higher in severe fetal growth restriction (FGR) and lower in participants with gestational diabetes (GDM) C. What does this study add to what is already known This is the first wearable near-infrared spectroscopy system to simultaneously monitor real-time changes in placental oxygenation and metabolism in vivo. This is also the first application of machine learning analysis to placental NIRS signals. Dynamic features of placental metabolism and oxygenation levels may provide clinically meaningful placental biomarkers.

Placental dysfunction is a major cause of adverse pregnancy outcomes such as stillbirth and neonatal hypoxic‐ischaemic encephalopathy. Currently available fetal monitoring tools (e.g. cardiotocography) lack the sensitivity and specificity to accurately predict pregnancy outcomes. Near‐infrared spectroscopy (NIRS) is emerging as a promising non‐invasive tool for monitoring of placental function. FetalSenseM (FSM) v1 is a semi‐portable wearable continuous‐wave NIRS device developed to monitor both placental oxygenation and metabolism. In this study, deep learning (DL) models were trained to predict pregnancy outcomes associated with placental dysfunction based on the optical signals using FSM v1 collected from 61 pregnant women at gestations after 24 weeks (mean age 36.01 years, standard deviation 5.70). Several DL architectures were explored to predict pregnancy outcomes: convolutional neural networks (CNNs), recurrent neural networks, hybrid and ensembles. Model performance was evaluated using balanced accuracy, recall (sensitivity), specificity, and receiver operating characteristic‐area under the curve (ROC‐AUC). Based on nested cross‐validation, the parallel CNN + LSTM model demonstrated the highest balanced accuracy of 0.615 (confidence interval (CI) = 0.48–0.74), followed by parallel CNN + bidirectional gated recurrent unit (GRU) of 0.577(CI = 0.41–0.75) and bidirectional GRU of 0.575 (CI = 0.42–0.73). This study highlights the potential of applying DL techniques to NIRS signals to inform antenatal risk assessment.