Intelligence-based medicine最新文献

Human movement patterns reflect central nervous system function. Small deficits in repetitive fast movements, such as slightly slowed finger-tapping or mildly irregular rhythm of stepping, are often an early sign of a neurological disorder. Accessible tools that precisely measure the individual components of fast movements would thus enhance disease detection, monitoring and research. Deep learning-based computer vision methods applied to digital video-recordings hold promise but current state-of-the-art tools, including DeepLabCut (DLC) and other advanced models, fail to accurately track the fastest range of human movements, primarily due to image blur. To solve this, we developed a new end-to-end, Rapid-Motion-Track (RMT) computer vision tool. This study aimed to evaluate the accuracy of RMT compared to DLC and other advanced computer vision tools. 220 finger-tapping tests were performed at frequencies between 0.5Hz and 6Hz and recorded simultaneously with a standard 30 frames/sec 2D laptop camera and a high-speed 250 frames/sec 3D motion tracking system (ground-truth). Bland-Altman plots and paired Welch's t-test were used to quantify the validity of movement features extracted by computer vision methods with the ground-truth. The movement features extracted by RMT (e.g. frequency, speed, variance) exhibited high concurrent validity across all tapping-frequencies. RMT outperformed other computer vision methods for very fast movements >4Hz. RMT also robustly tracked other fast motions including sit-to-stand, head-turning, foot-tapping, and leg agility. This new tool provides an accurate method to precisely and automatically measure even the fastest and finest human movements. It holds potential of wide reach as digital cameras are ubiquitous in homes, clinics and research centres.

In this retrospective single-centre study spanning five years (2016–2021) and involving 2368 adult Intensive Care Unit (ICU) patients requiring over 4 h of mechanical ventilation (MV) in a tertiary care hospital, we investigated the feasibility and accuracy of using machine learning (ML) models in predicting outcomes post-ICU discharge compared to conventional statistical methods (CSM). The study aimed to identify associated risk factors impacting these outcomes. Poor outcomes, defined as ICU readmission, mortality, and prolonged hospital stays, affected 40.2 % of the discharged MV patients. The Extreme Gradient Boost (XGBoost) ML model showed superior performance compared to CSM (Area under the receiver operating characteristic curve: 0.693 vs. 0.667; p-value = 0.03). At 95 % specificity, XGBoost displayed enhanced sensitivity (30.6 % vs. 23.8 %) compared to CSM. Risk factors such as Glasgow Coma Score (GCS) and GCS best motor score at ICU discharge, MV duration, ICU length of stay, and Charlson Comorbidity Index were identified. While both ML and CSM exhibited moderate accuracy, the study suggests ML algorithms have the potential for better predictive capabilities and individual risk factor identification, potentially aiding in the improvement of patient outcomes by identifying high-risk patients requiring closer monitoring. Further validation in larger studies is necessary, but the study underscores the potential for real-time application of ML algorithms developed from the increasing availability of electronic medical records (EMR).

Background

Methods

Results

Conclusion

A growing number of new cases and fatalities occur each year due to breast cancer, making it the most frequent malignancy globally. Utilizing a multioutput classifier technique with algorithms such as CatBoost, XGBoost, NN, and NN Binary, this work presents a new model for predicting breast cancer treatments: surgery, radiotherapy, and chemotherapy. We tackle the pressing need for accurate medical treatments by developing a model to enhance the predicted accuracy of breast cancer treatment outcomes. The model accomplishes impressive results in predicting surgical outcomes; in particular, Neural Networks (NN and NN Binary) perform exceptionally well in terms of recall and precision, reaching 97 % accuracy and 98 % F1-scores. While the model's accuracy is only about 63 % for radiotherapy, it shows a promising recall of up to 84 %. Accuracy and precision in chemotherapy predictions remain stable at 82 %, with AUC-ROC values of up to 89 %, suggesting excellent discrimination ability. By combining multioutput classifiers with sophisticated algorithms, we hope to make treatment prediction models more tailored to individual breast cancer patient profiles, which might usher in a new era of tailored treatment plans and meet the rising demand for precision medicine in cancer care.

Context

Skin cancer is the most incident neoplasia in Brazil, and their invasiveness can be impacted by various factors, including geographical aspects. Identifying these factors is important for improving diagnosis and treatment.

Objective

The research focused on analyzing the impact of region on the invasiveness of skin cancer in Brazil, through the identification of regional predictive patterns.

Methods

An analysis and processing of data from the Hospital Cancer Registries (RHC) of Brazil's National Cancer Institute (INCA) were conducted, followed by the application of machine learning algorithms. The SHapley Additive exPlanations (SHAP) approach was employed to provide explanations for the developed artificial intelligence models.

Results

It was revealed that geography plays a significant role in predicting the invasiveness of skin cancer, reinforcing the need to consider regional specificities in future studies.

Conclusions

The study identified that regional characteristics of Brazil impacts the prediction of the invasiveness of skin cancer. Despite limitations, such as the issue of data imbalance, the findings are important for developing more effective policies in the fight against skin cancer in the Brazil.

This work introduces the “MedTransCluster” framework, a novel approach to medical image clustering in chest radiography through the application of transfer learning, leveraging the capabilities of pre-trained deep learning models. Our evaluation encompassed a variety of neural networks, considering their adaptability to the nuances of medical imaging data. The study incorporated four renowned clustering algorithms and an expanded set of evaluation metrics, offering a comprehensive comparison and a refined analysis of these models’ ability to cluster complex diagnostic features. Notably, EfficientNetB0 coupled with DBSCAN clustering algorithm achieved a silhouette score of 0.924131, and ResNet152 with KMeans displayed a Calinski Harabasz score of 9655.213964, indicating their superior proficiency in capturing the intricacies of medical features. These results emphasize the critical importance of model refinement within the healthcare imaging sphere and underscore the potential of methodologies like MedTransCluster in enhancing diagnostic accuracy and patient outcomes.