ABSTRACT Developing radiomic biomarkers remains challenging due to the variability in imaging protocols across centres and the lack of standardised methodologies. This study evaluates the impact of different technical decisions in radiomics pipelines using multiparametric magnetic resonance imaging data from six centres for predicting therapy response in rectal cancer. Preprocessing, feature extraction, normalisation strategies, and machine learning (ML) models were assessed for robustness and generalisability. Key findings demonstrated that the preprocessing significantly enhanced the feature reproducibility and the importance of the selected ones over the complexity of the ML classifier. This analysis highlights the necessity of a unique pipeline and the complexity of radiomics biomarkers in real-world settings, particularly when handling highly imbalanced datasets. Several insights and methodologies have been presented that may support towards more conscious decisions when implementing radiomic systems. Future efforts should focus on integrating clinical/genomic/pathomics data to improve the predictive capabilities and facilitate the introduction into clinical practice.

ABSTRACT The rising up-on-the-toes (UTT) 30-second test is used clinically to assess ankle muscle strength and endurance. Typically, the test is subjectively evaluated by counting how many UTT movements are completed. We have recently shown that the UTT test can be objectively assessed using signals from small inertial measurement units (IMUs). The current study investigates whether computer vision (CV) analysis of the UTT test gives comparable outcomes to IMU analysis. A CV-based system was applied to video recordings of 29 older adult participants (76.0 ± 4.3 years) performing the UTT test with IMUs attached to their feet. Angular velocity time series signals were generated using both IMU and video object detection of the right foot landmarks, enabling peak plantarflexion velocities during the ascent and descent phases to be extracted. Findings demonstrate that the CV-based approach produces closely aligned output metrics with IMU data, with coefficient of determination (R 2) values of ≥0.91.

ABSTRACT Introduction Alzheimer’s disease (AD) is a severe brain disorder, which comprises almost 75% of all types of dementia cases. As per the latest report, the global count of Alzheimer’s disease patients is 131 million, due to which the treatment of Alzheimer’s disease is becoming a major burden to the world healthcare system. However, Alzheimer’s disease still remains incurable due to its multifactorial nature of symptoms. Therefore, early-stage diagnosis of Alzheimer’s disease is essential, which helps in the treatment and recovery of patients to a greater extent. Background In the existing literature, few research attempts are carried out for Alzheimer’s disease prediction using machine learning algorithms; yet, most of the existing techniques fail to focus on aspects such as gain-optimisation, dementia severity identification and accuracy enhancement. Methodology To solve these issues, this article presents a light gradient boosting machine-based framework, which employs a gain-optimisation algorithm using multi-category factors for the early stage prediction of Alzheimer’s disease. Further, the proposed framework introduces a detailed feature analysis for identifying the key features needed for the disease prediction and thereby classifies the Alzheimer’s disease patients into different dementia severity types. Results The extensive evaluations of the proposed framework on different dataset sizes using real-world MRI datasets of patients, clearly demonstrate the performance of the proposed work in terms of gain-optimised accuracy metrics. Further, experimental evaluations clearly prove the better performance of the proposed framework when compared to that of other ML techniques including LR, RFC, SVR, SVC and existing methods in terms of Recall, F1-score and accuracy metrics.

ABSTRACT Patient-specific applications in biomechanics and orthopaedics call for segmentation of volumetric medical images. This task has historically been performed manually or semi-automatically, which entails significant effort by trained experts. Computer vision algorithms based on machine learning have emerged as an effective tool for accelerating medical image processing tasks through automation. The current work describes fully no-code software which supports development of custom machine learning models for automatic segmentation of volumetric medical images. This cloud-integrated desktop application allows researchers and biomedical engineers to define segmentation algorithms, configure annotated datasets, and train personalised machine learning algorithms with cloud compute in just a few clicks. A presented case study demonstrates an ideal workflow in the software using femoral cartilage segmentation from magnetic resonance imaging as a representative use case. Models developed in the software demonstrated mean dice similarity coefficients of up to 0.886 on a test cohort, which is competitive with previously reported methods developed with significantly larger data sets. The mean surface difference between digitised models reconstructed from ground truth and predicted segmentation was 0.33 mm. Results suggest the described software enables creation of accurate machine learning models with limited engineering effort.

ABSTRACT The rapid identification and categorisation of breast cancers using low-contrast MRI images presents a significant challenge due to the disease’s prevalence among women of all ages. Nowadays, it is very difficult to classify and generalise models from low-contrast MRI datasets due to the prevalence of class imbalance caused by a wide range of symptoms and untrustworthy data sources. Using ensemble learning with optimised k-means helps with these problems; it reduces overfitting and improves reliability by using methods like k-means clustering, frog feap algorithm (FLA), boosting, and bagging to balance datasets and show the complexity of symptoms. In this research, we present a new strategy that makes use of a hybrid optimised K-Means algorithm combined with the FLA, along with thresholding and morphological approaches, and ensemble learning. There are two main stages to our work. In the initial phase, we preprocess low-contrast MRI images and delineate the breast tumour area from the segmented MRI images. Furthermore, we employ Discrete Wavelet Transformation (DWT) to extract the most salient features. In the next step, the retrieved features will be used as input parameters for an ensemble-learned breast tumour classifier. To determine whether a low-contrast breast tumour MRI picture has a benign or malignant tumour after model training, we use a variety of classifiers, such as k-nearest neighbours (KNN), decision trees (DT), gradient boosting (GB), random forests (RF), and artificial neural networks (ANN). Through a voting mechanism that averages the accuracy of all models, our suggested framework achieves a remarkable 98.8% accuracy rate. In addition, proposed model is quite efficient & reliable to detect the kind of breast tumour or cancer with a sensitivity of 98.02% and a specificity of 97.5%.

ABSTRACT The limited availability of medical images is a major limitation when using deep learning, which requires large amounts of data to improve performance. To address this problem, transfer learning has become the de facto standard, using convolutional neural networks (CNNs) previously trained on natural images, and fine-tuned on medical images. Recently, vision transformers (ViT), which require large annotated medical images, have been studied from various perspectives. In this study, we investigated an effective pre-training method for binary classification of COVID-19 using chest radiography (CXR) images. Our results showed that pre-training on natural images outperformed CXR images using the fine-tuning method. Pre-training on natural images was also able to capture more global features. Furthermore, this trend increased as the number of pre-trained on natural images increased. In summary, these results suggest that the fine-tuning method with a large number of natural images as pre-training had the best discrimination performance.

ABSTRACT Computer-aided diagnosis of ocular disorders like glaucoma can be effectively performed with retinal fundus images. Since the advent of machine learning and later deep learning techniques in medical image analysis, research focusing on automated and early detection of glaucoma has gained prominence. In this paper, we show how a deep Convolutional Neural Network model can be used in visualising the features related to Glaucoma that are consistent with (a subset of) those used by ophthalmologists for diagnosis. However, since the doctors do not necessarily depend entirely on these features but back them up with other important measurements such as visual field testing (HVF), any deep neural network such as CNN may not necessarily be expected to yield 100% accuracy. Further, other important parameters such as Intraocular Pressure (IOP), which may be built into a deep learning model, may not necessarily correlate well with the presence of Glaucoma. In this work, a deep CNN model has been trained and tested on a large number of high-quality fundus images, both normal and glaucomatous. We compare the results with transfer learning models such as VGG16, ResNet50, and MobileNetV2. On an average, we obtained an accuracy of 93.75% in identifying glaucoma, focusing only on the features of the fundus image.

ABSTRACT The goal of this study was to develop an image analysis algorithm for quantifying the effects of remodelling on cortical bone during early fracture healing. An adaptive thresholding technique with boundary curvature and tortuosity control was developed to automatically identify the endocortical and pericortical boundaries in the presence of high-gradient bone mineral density (BMD) near the healing zone. The algorithm successfully detected boundaries in more than 47,000 microCT images from 12 pairs of healing ovine osteotomies and intact contralateral tibiae. Resampling techniques were used to achieve data dimensionality reduction on the segmented images, allowing characterisation of radial, circumferential, and axial distributions of cortical BMD. Local (transverse slice) and total (whole bone) remodelling scores were produced. These surrogate measures of cortical remodelling derived from BMD revealed that cortical changes were detectable throughout the region covered by callus and that the localised loss of cortical BMD was highest near the osteotomy. Total remodelling score was moderately and significantly correlated with callus volume and mineral composition (r > 0.64, p < 0.05), suggesting that the cortex may be a source of mineral needed to build callus.

ABSTRACT Purpose Intra-cerebral hemorrhage is the most severe type of stroke, and hematoma volume is the strongest prognostic factor. Developing a cost and time-effective method to compute volumes of intracranial hemorrhage in patients of a high volume, tertiary center would assist in rapid triaging and minimizing treatment delays. Description The volumes of epidural, subdural and intra-parenchymal hemorrhages were calculated using ABC/2 method and compared with planimetric software method. ICH score was calculated and correlated with in-hospital mortality in non-traumatic cases. Results The median IQR volume of the ABC/2 formula method versus planimetric method, suggested no statistically significant difference (p - 0.079) in hematoma volume calculation. No statistically significant difference existed w.r.t degree of irregularity (p-0.849) and heterogeneity (p-0.815) of hemorrhage. Excellent agreement was found between the two methods while considering all 3 types of hemorrhages together (ICC - 0.969); however, discrepancies arose in case of extensive subdural hemorrhages (p- 0.041). ICH score was found to be directly proportional to mortality. Conclusion Overall, the ABC/2 formula and software-assisted methods performed equally well in estimating volumes of intraparenchymal and epidural hemorrhages; however, ABC/2 method overestimates the volume in extensive subdural hemorrhages. Intraventricular and infratentorial extensions were other parameters of the ICH score with relatively higher significance.

ABSTRACT Early illness diagnosis and prediction are important goals in healthcare in order to offer timely preventive measures. The best, least invasive, and most reliable way for identifying any neurological disorder is EEG analysis. If neurological disorders could somehow be predicted in advance, patients could be saved from their detrimental consequences. With promising new advancements in machine learning-based algorithms, Early and precise prediction might induce a radical shift. Here, we present a thorough analysis of cutting-edge AI methods for exploiting EEG data for Parkinson’s disease early warning symptoms detection, sleep apnoea, drowsiness, schizophrenia, motor imagery classification, and emotion recognition, among other conditions. All of the EEG signal analysis procedures used by different authors, such as hardware software data sets, channel, frequency, epoch, preprocessing, decomposition method, features, and classification, have been compared and analysed in detail. We will point out the difficulties, gaps and limitations in the current research and suggest future avenues of research.