12 Plagues of AI in Healthcare: A Practical Guide to Current Issues With Using Machine Learning in a Medical Context

Evaluating gender bias in ML-based clinical risk prediction models: A study on multiple use cases at different hospitals

BACKGROUND

An inherent difference exists between male and female bodies, the historical under-representation of females in clinical trials widened this gap in existing healthcare data. The fairness of clinical decision-support tools is at risk when developed based on biased data. This paper aims to quantitatively assess the gender bias in risk prediction models. We aim to generalize our findings by performing this investigation on multiple use cases at different hospitals.

METHODS

First, we conduct a thorough analysis of the source data to find gender-based disparities. Secondly, we assess the model performance on different gender groups at different hospitals and on different use cases. Performance evaluation is quantified using the area under the receiver-operating characteristic curve (AUROC). Lastly, we investigate the clinical implications of these biases by analyzing the underdiagnosis and overdiagnosis rate, and the decision curve analysis (DCA). We also investigate the influence of model calibration on mitigating gender-related disparities in decision-making processes.

RESULTS

Our data analysis reveals notable variations in incidence rates, AUROC, and over-diagnosis rates across different genders, hospitals and clinical use cases. However, it is also observed the underdiagnosis rate is consistently higher in the female population. In general, the female population exhibits lower incidence rates and the models perform worse when applied to this group. Furthermore, the decision curve analysis demonstrates there is no statistically significant difference between the model's clinical utility across gender groups within the interested range of thresholds.

CONCLUSION

The presence of gender bias within risk prediction models varies across different clinical use cases and healthcare institutions. Although inherent difference is observed between male and female populations at the data source level, this variance does not affect the parity of clinical utility. In conclusion, the evaluations conducted in this study highlight the significance of continuous monitoring of gender-based disparities in various perspectives for clinical risk prediction models.

Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary

Over the past few years, artificial intelligence (AI) has emerged as a powerful tool used to efficiently automate several tasks across multiple domains. Sleep medicine is perfectly positioned to leverage this tool due to the wealth of physiological signals obtained through sleep studies or sleep tracking devices and abundance of accessible clinical data through electronic medical records. However, caution must be applied when utilizing AI, due to intrinsic challenges associated with novel technology. The Artificial Intelligence in Sleep Medicine Committee of the American Academy of Sleep Medicine reviews advancements in AI within the sleep medicine field. In this article, the Artificial Intelligence in Sleep Medicine committee members provide a commentary on the scope of AI technology in sleep medicine. The commentary identifies 3 pivotal areas in sleep medicine that can benefit from AI technologies: clinical care, lifestyle management, and population health management. This article provides a detailed analysis of the strengths, weaknesses, opportunities, and threats associated with using AI-enabled technologies in each pivotal area. Finally, the article broadly reviews barriers and challenges associated with using AI-enabled technologies and offers possible solutions.

CITATION

Bandyopadhyay A, Oks M, Sun H, et al. Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary. J Clin Sleep Med. 2024;20(7):1183-1191.

Disability risk prediction model based on machine learning among Chinese healthy older adults: results from the China Health and Retirement Longitudinal Study

Background

Predicting disability risk in healthy older adults in China is essential for timely preventive interventions, improving their quality of life, and providing scientific evidence for disability prevention. Therefore, developing a machine learning model capable of evaluating disability risk based on longitudinal research data is crucial.

Methods

We conducted a prospective cohort study of 2,175 older adults enrolled in the China Health and Retirement Longitudinal Study (CHARLS) between 2015 and 2018 to develop and validate this prediction model. Several machine learning algorithms (logistic regression, k-nearest neighbors, naive Bayes, multilayer perceptron, random forest, and XGBoost) were used to assess the 3-year risk of developing disability. The optimal cutoff points and adjustment parameters are explored in the training set, the prediction accuracy of the models is compared in the testing set, and the best-performing models are further interpreted.

Results

During a 3-year follow-up period, a total of 505 (23.22%) healthy older adult individuals developed disabilities. Among the 43 features examined, the LASSO regression identified 11 features as significant for model establishment. When comparing six different machine learning models on the testing set, the XGBoost model demonstrated the best performance across various evaluation metrics, including the highest area under the ROC curve (0.803), accuracy (0.757), sensitivity (0.790), and F1 score (0.789), while its specificity was 0.712. The decision curve analysis (DCA) indicated showed that XGBoost had the highest net benefit in most of the threshold ranges. Based on the importance of features determined by SHAP (model interpretation method), the top five important features were identified as right-hand grip strength, depressive symptoms, marital status, respiratory function, and age. Moreover, the SHAP summary plot was used to illustrate the positive or negative effects attributed to the features influenced by XGBoost. The SHAP dependence plot explained how individual features affected the output of the predictive model.

Conclusion

Machine learning-based prediction models can accurately evaluate the likelihood of disability in healthy older adults over a period of 3 years. A combination of XGBoost and SHAP can provide clear explanations for personalized risk prediction and offer a more intuitive understanding of the effect of key features in the model.

Prediction of orthognathic surgery plan from 3D cephalometric analysis via deep learning

BACKGROUND

Preoperative planning of orthognathic surgery is indispensable for achieving ideal surgical outcome regarding the occlusion and jaws' position. However, orthognathic surgery planning is sophisticated and highly experience-dependent, which requires comprehensive consideration of facial morphology and occlusal function. This study aimed to investigate a robust and automatic method based on deep learning to predict reposition vectors of jawbones in orthognathic surgery plan.

METHODS

A regression neural network named VSP transformer was developed based on Transformer architecture. Firstly, 3D cephalometric analysis was employed to quantify skeletal-facial morphology as input features. Next, input features were weighted using pretrained results to minimize bias resulted from multicollinearity. Through encoder-decoder blocks, ten landmark-based reposition vectors of jawbones were predicted. Permutation importance (PI) method was used to calculate contributions of each feature to final prediction to reveal interpretability of the proposed model.

RESULTS

VSP transformer model was developed with 383 samples and clinically tested with 49 prospectively collected samples. Our proposed model outperformed other four classic regression models in prediction accuracy. Mean absolute errors (MAE) of prediction were 1.41 mm in validation set and 1.34 mm in clinical test set. The interpretability results of the model were highly consistent with clinical knowledge and experience.

CONCLUSIONS

The developed model can predict reposition vectors of orthognathic surgery plan with high accuracy and good clinically practical-effectiveness. Moreover, the model was proved reliable because of its good interpretability.

Graph Theory Measures and Their Application to Neurosurgical Eloquence

Improving patient safety and preserving eloquent brain are crucial in neurosurgery. Since there is significant clinical variability in post-operative lesions suffered by patients who undergo surgery in the same areas deemed compensable, there is an unknown degree of inter-individual variability in brain 'eloquence'. Advances in connectomic mapping efforts through diffusion tractography allow for utilization of non-invasive imaging and statistical modeling to graphically represent the brain. Extending the definition of brain eloquence to graph theory measures of hubness and centrality may help to improve our understanding of individual variability in brain eloquence and lesion responses. While functional deficits cannot be immediately determined intra-operatively, there has been potential shown by emerging technologies in mapping of hub nodes as an add-on to existing surgical navigation modalities to improve individual surgical outcomes. This review aims to outline and review current research surrounding novel graph theoretical concepts of hubness, centrality, and eloquence and specifically its relevance to brain mapping for pre-operative planning and intra-operative navigation in neurosurgery.

Parcellation-based tractographic modeling of the salience network through meta-analysis

BACKGROUND

The salience network (SN) is a transitory mediator between active and passive states of mind. Multiple cortical areas, including the opercular, insular, and cingulate cortices have been linked in this processing, though knowledge of network connectivity has been devoid of structural specificity.

OBJECTIVE

The current study sought to create an anatomically specific connectivity model of the neural substrates involved in the salience network.

METHODS

A literature search of PubMed and BrainMap Sleuth was conducted for resting-state and task-based fMRI studies relevant to the salience network according to PRISMA guidelines. Publicly available meta-analytic software was utilized to extract relevant fMRI data for the creation of an activation likelihood estimation (ALE) map and relevant parcellations from the human connectome project overlapping with the ALE data were identified for inclusion in our SN model. DSI-based fiber tractography was then performed on publicaly available data from healthy subjects to determine the structural connections between cortical parcellations comprising the network.

RESULTS

Nine cortical regions were found to comprise the salience network: areas AVI (anterior ventral insula), MI (middle insula), FOP4 (frontal operculum 4), FOP5 (frontal operculum 5), a24pr (anterior 24 prime), a32pr (anterior 32 prime), p32pr (posterior 32 prime), and SCEF (supplementary and cingulate eye field), and 46. The frontal aslant tract was found to connect the opercular-insular cluster to the middle cingulate clusters of the network, while mostly short U-fibers connected adjacent nodes of the network.

CONCLUSION

Here we provide an anatomically specific connectivity model of the neural substrates involved in the salience network. These results may serve as an empiric basis for clinical translation in this region and for future study which seeks to expand our understanding of how specific neural substrates are involved in salience processing and guide subsequent human behavior.