Studies in Health Technology and Informatics

Development and Evaluation of an Electronic Shared Decision-Making Tool to Reduce Decision Anxiety in Premenopausal Hormone Receptor-Positive Breast Cancer Patients

This study describes the development, implementation, and evaluation of a digital shared decision-making (SDM) tool designed for premenopausal women with hormone receptor-positive breast cancer.

A Deep-Learning Framework for Ovarian Cancer Subtype Classification Using Whole Slide Images

Ovarian cancer, a leading cause of cancer-related deaths among women, comprises distinct subtypes each requiring different treatment approaches. This paper presents a deep-learning framework for classifying ovarian cancer subtypes using Whole Slide Imaging (WSI). Our method contains three stages: image tiling, feature extraction, and multi-instance learning. Our approach is trained and validated on a public dataset from 80 distinct patients, achieving up to 89,8% accuracy with a notable improvement in computational efficiency. The results demonstrate the potential of our framework to augment diagnostic precision in clinical settings, offering a scalable solution for the accurate classification of ovarian cancer subtypes.

HELPeR: Interface Design Decision and Evaluation

Inequities in health information access contribute to disparities in health outcomes. Health recommender systems have emerged as a promising solution to help users find the right information. Despite their various applications, it remains understudied how these systems can aid cancer patients. In this paper, we introduce HELPeR, a recommender system designed to assist ovarian cancer patients with their information needs. The design addresses cold-start challenges, drawing input from health experts and ovarian cancer forum posts. We evaluated HELPeR with nurse practitioners in a cold-start scenario, highlighting its benefits and areas for future improvement.

Towards Building an E-Librarian System: Exploring Recommendation System Preferences Among Ovarian Cancer Patients and Caregivers

Ovarian cancer (OvCa) patients encounter complex treatment decisions, and often have difficulties in searching and integrating online health information to guide their treatment decision-making. The objective of this study was to explore the preference of online health information among OvCa patients and caregivers, by exploring their preferred content, format, and function features for the design of a personalized recommender system. This study used qualitative research methods to collect data through in-depth interviews with 18 OvCa patients and 2 caregivers. A total of (N=20) face-to-face interviews were conducted, and subsequently analyzed by audio recordings, verbatim transcription, and theory-driven approach with thematic analysis. A total of 5 themes were identified for content-related design, 4 themes identified for system function and one theme identified for frequency format. The results of this study inform the preference and therefore OvCa specific features can be tailor-made in a recommendation system.

From Normal to Optimal: Investigating Metabolic and Inflammatory Parameters as Predictors of Survival in Locally Advanced Cervical Cancer

Cervical cancer is the third most common cancer in women, and recent studies have highlighted the importance of body composition markers in predicting patient outcomes. We build upon the data of 83 patients from the Uterus-11 study, to explore the relation of pairwise feature combinations to long-term progression-free survival. We propose a framework to identify the parameter combinations with pre-defined thresholds of “normal range” which provide good separation of the survival group. Further, we optimize the pair-wise thresholds to further improve the separation measured by F1 scores. This approach allowed us to improve the statistical significance of hazard ratios in comparison to the previous studies. The optimization results suggest that the normal ranges of well-established biomarkers such as body mass index could be shifted in the context of specific diseases to achieve optimal outcome.

Bias Detection in Histology Images Using Explainable AI and Image Darkness Assessment

The study underscores the importance of addressing biases in medical AI models to improve fairness, generalizability, and clinical utility. In this paper, we present a novel framework that combines Explainable AI (XAI) with image darkness assessment to detect and mitigate bias in cervical histology image classification. Four deep learning architectures were employed—AlexNet, ResNet-50, EfficientNet-B0, and DenseNet-121—with EfficientNet-B0 demonstrating the highest accuracy post-mitigation. Grad-CAM and saliency maps were used to identify biases in the models’ predictions. After applying brightness normalisation and synthetic data augmentation, the models shifted focus toward clinically relevant features, improving both accuracy and fairness. Statistical analysis using ANOVA confirmed a reduction in the influence of image darkness on model predictions after mitigation, as evidenced by a decrease in the F-statistic from 120.79 to 14.05, indicating improved alignment of the models with clinically relevant features.

Clinical Factors Associated with Delayed Diagnosis of Ovarian Cancer: Retrospective Cohort Study Using Linked Medical Records

Using linked primary care, hospital and cancer registry data, we show that presenting symptom and existence comorbid long-term conditions offering alternative explanations for symptoms, both influence the diagnostic delay for ovarian cancer.

Utilizing Digital Twin to Create Personas Representing Ovarian Cancer Patients and Their Families

OvCa patients and caregivers perceived challenges in online health information seeking. The HELPeR recommendation system utilized digital twins to create personas reflecting real-world OvCa patients and caregivers. The aim of this study was to describe the creation of digital twins and demonstrate their use cases in the study. Digital twins of OvCa patients and caregivers were created by triangulating multiple sources, including online cancer forums, direct interviews with patients and caregivers, domain expert input, and clinical notes. 10 personas were created for both OvCa patients and caregivers who had a variety of cancer trajectories and information interests. These digital twins present a potential solution for training artificial intelligence models at the initial phase when there is a scarcity of user information.

Development, Implementation and Initial Results of CDSS Recommendations for Patients at Risk of Hereditary Breast Cancer

Breast cancer represents 23% of all cancers diagnosed among women each year. BRCA1 and BRCA2 are tumor suppressor genes related to the most frequent form of hereditary breast and ovarian cancer, as well as other types of cancer. The aim of this work is to describe the development of Clinical Decision Support Systems (CDSS) for referral to genetic counseling in patients at increased risk of pathogenic variants in BRCA1 and BRCA2, and to describe results during the pilot study implementation (from January 5, 2021 to March 5, 2021). To achieve integration and system interoperability, we used FHIR and CDS-Hooks within the CDSS development. A total of 142 alerts were triggered by the system for 72 physicians in 98 patients. Results showed an acceptance rate for the recommendation of 2.1%, which could improve using intrusive alerts in all of the hooks.

Process Mining to Explore Variations in Endometrial Cancer Pathways from GP Referral to First Treatment

The main challenge in the pathway analysis of cancer treatments is the complexity of the process. Process mining is one of the approaches that can be used to visualize and analyze these complex pathways. In this study, our purpose was to use process mining to explore variations in the treatment pathways of endometrial cancer. We extracted patient data from a hospital information system, created the process model, and analyzed the variations of the 62-day pathway from a General Practitioner referral to the first treatment in the hospital. We also analyzed the variations based on three different criteria: the type of the first treatment, the age at diagnosis, and the year of diagnosis. This approach should be of interest to others dealing with complex medical and healthcare processes.

Lifestyle Cancer Survival Predictors: Influence of Vegetarian Diet on the Relapse of Endometrial Cancer

Endometrial cancer (EC) is the most common gynecological tumor in high-income countries, and its incidence has increased over time. The most critical risk factor for EC is the long-term unopposed exposure to increased estrogens both exogenous and endogenous. Machine learning can be used as a promising tool to resolve longstanding challenges and support identification of the risk factors and their correlations before the clinical trials and make them more focused. In this paper we present the results of the research of the correlation analysis of Endometrial cancer risk factors. The study was performed with EC patients of the Almazov center in Saint-Petersburg, Russia. All women involved in the current study underwent radical surgical intervention due to EC. After initial cancer treatment, they were referred to the Almazov center outpatient specialists for follow-up visits. Many of them were readmitted of the inpatient clinic due to relapse. We extracted a variety of parameters related to lifestyle, dietary habits, socioeconomic, and reproductive features from the inpatient and outpatient databases of Almazov center. The medical records of the women with enough data were included in the study. Prediction of Progression-free survival (PFS) and overall survival (OS) were analyzed respectively. The AUC of ROC was calculated for PFS = 0.93 and for OS = 0.94.

Cervical Intraepithelial Neoplasia Grading from Prepared Digital Histology Images

The paper proposes an integrated approach to the automated diagnosis of cervical intraepithelial neoplasia (CIN) in epithelial patches extracted from digital histology images. Experiments were conducted to determine the most suitable deep learning model for the dataset and fuse patch predictions to decide the final CIN grade of the histology samples. Seven candidate CNN architectures were assessed in this study. Three fusion methods were applied to the best CNN classifier. The model ensemble, combined CNN classifier and highest performing fusion method achieved an accuracy of 94.57%. This result shows significant improvement over the state-of-the-art classifiers for cervical cancer histopathology images. It is hoped that this work will contribute towards further research to automate diagnosis of CIN from digital histopathology images.

An Integrated Approach to Automated Diagnosis of Cervical Intraepithelial Neoplasia in Digital Histology Images

The study proposes an integrated approach to automated cervical intraepithelial neoplasia (CIN) diagnosis in epithelial patches extracted from digital histology images. The model ensemble, combined CNN classifier, and highest-performing fusion approach achieved an accuracy of 94.57%. This result demonstrates significant improvement over the state-of-the-art classifiers for cervical cancer histopathology images and promises further improvement in the automated diagnosis of CIN.

Development of a Graph-Based Database for Ovarian Cancer Symptoms

Uncontrolled cancer-related symptoms severely affect the quality of life of cancer patients. This study leveraged symptom data written in the online health communities (OHCs) and developed a symptom knowledge database for cancer patients. Further study is needed to apply this pilot study to the larger OHCs data.

Unraveling Endometrial Cancer Survival Predictors Through Advanced Machine Learning Techniques

This study explores endometrial cancer (EC) within the broader context of oncogynecology, focusing on 3,845 EC patients at the Almazov National Research Center. The research analyzes clinical data, employing machine learning techniques like random forest regression and decision tree analysis. Key findings include age-dependent impacts on EC outcomes, unexpected correlations between dietary habits and recurrence risk (e.g., higher risk for vegans), and intriguing associations like soft drink consumption influencing relapse. Despite limitations like a retrospective design and self-reported data, the study’s extended eight-year follow-up and robust database enhance its credibility. The nuanced insights into EC risk factors, influenced by factors like physical activity and diet, open avenues for targeted diagnostics and prevention strategies, showcasing the potential of machine learning in predicting outcomes.

Quality Assessment of Meta-Analyses on Cervical Cancer

Meta-analysis (MA) quantitatively summarizes the findings of independent studies and is considered the highest quality of evidence for evidence-based medicine. However, issues in reporting and methodological rigor of MA hamper reproducibility and create the potential for bias. By applying PRISMA reporting guideline and AMSTAR2 execution guidelines on 40 cervical cancer MA samples covering topics such as interventions and risk factors, we determined the extent to which MA execution adhered to best practice guidelines. The results show that the elements with least adherence include “review methods established before MA” and “principal summary measures defined” (each 32.5% per PRISMA) and “characteristics of included studies” (31.3% per AMSTAR2) which undermine reproducibility and increase the risk of bias. This initial work presents common pitfalls in MA and is intended to improve awareness of these issues for clinicians who are interested in conducting MA and to pave the way toward quality improvement via informatics approaches.

Using Statistical Learning to Examine Variables that Contribute to Longer Hospital Stays After Endometrial Cancer Surgery

Endometrial cancer, primarily arising from the lining of the uterus (endometrium), is the most common gynecological malignancy in many developed countries. Risk factors include obesity, age, and a family history of related cancers. The prognosis is generally favorable when detected early. One important measure of clinical results and resource use is the duration of hospital stay (LOS) after surgery. Understanding the causes of prolonged hospital stays is crucial for improving patient care because both organizational and patient-related factors can have a substantial impact on LOS. This study uses a statistical learning technique to investigate the LOS following endometrial cancer surgery at the Antonio Cardarelli Hospital in Naples, Italy. It expands on earlier research on the reasons for prolonged LOS in surgical oncology. Illness severity showed a potential influence on LOS but was not statistically significant in the model.

Natural Language Processing to Identify Abnormal Breast, Lung, and Cervical Cancer Screening Test Results from Unstructured Reports to Support Timely Follow-up

Cancer screening and timely follow-up of abnormal results can reduce mortality. One barrier to follow-up is the failure to identify abnormal results. While EHRs have coded results for certain tests, cancer screening results are often stored in free-text reports, which limit capabilities for automated decision support. As part of the multilevel Follow-up of Cancer Screening (mFOCUS) trial, we developed and implemented a natural language processing (NLP) tool to assist with real-time detection of abnormal cancer screening test results (including mammograms, low-dose chest CT scans, and Pap smears) and identification of gynecological follow-up for higher risk abnormalities (i.e. colposcopy) from free-text reports. We demonstrate the integration and implementation of NLP, within the mFOCUS system, to improve the follow-up of abnormal cancer screening results in a large integrated healthcare system. The NLP pipelines have detected scenarios when guideline-recommended care was not delivered, in part because the provider mis-identified the text-based result reports.

Bridging AI, NLP, and Digital Health: A Bi-Directional Learning Experience at Emory University

Digital health, natural language processing, and artificial intelligence collectively possess considerable promise, with the potential to transform healthcare particularly within resource-limited settings. This manuscript discusses a research fellowship at the Department of Biomedical Informatics at Emory University, focusing on the outcomes of applying AI and NLP to address healthcare challenges in low- and middle-income countries (LMICs) and the bi-directional learning between academia and real-world field application. The fellowship explored multilingual generative language models, gender bias in NLP, and conversational AI, demonstrating its application through the enhanced development of AvivaAI, a mobile conversational chatbot for cervical cancer awareness in Nigeria.