International Journal of Medical Informatics

Artificial intelligence generated visual communication improves comprehension and adherence in cervical cancer screening: a randomized controlled study

Cervical cancer is preventable, yet poor comprehension of Pap smear results and non-adherence to follow-up major barriers, particularly in low health literacy settings. In Georgia, where screening coverage is below 20%, innovative communication strategies are needed. Artificial intelligence (AI) offers opportunities to strengthen patient communication through adaptive, emotionally expressive visual tools. To evaluate whether AI-generated visual explanations, paired with simplified text, improve comprehension, satisfaction, and follow-up adherence after cervical cancer screening compared with conventional text reporting. A randomized controlled trial enrolled 3,000 women aged 21-65 who underwent Pap smear testing between March and October 2024. Participants were randomized to three groups: Control (standard text), Text-only (enhanced plain-language text), and Intervention (AI-generated visuals plus text). Visuals were created with Craiyon, refined through expert and patient feedback, and aligned with Bethesda categories. Surveys assessed comprehension, satisfaction, and follow-up intent, while electronic records verified adherence. Analyses included chi-square tests, Kruskal-Wallis conformation for ordinal outcomes, and logistic regression for demographics and health literacy. The Intervention group achieved superior outcomes across all metrics. Comprehension reached 90 % versus 78 % in Text-only and 65 % in Control (χ AI-generated visual communication significantly improved comprehension, satisfaction, and follow-up adherence in cervical cancer screening. This study demonstrates a scalable informatics solution for patient engagement, though challenges remain regarding long-term behavioral impact, cross-cultural adaptation, and integration into routine health information systems.

Using machine learning to predict ovarian cancer

Ovarian cancer (OC) is one of the most common types of cancer in women. Accurately prediction of benign ovarian tumors (BOT) and OC has important practical value. Our dataset consists of 349 Chinese patients with 49 variables including demographics, blood routine test, general chemistry, and tumor markers. Machine learning Minimum Redundancy - Maximum Relevance (MRMR) feature selection method was applied on the 235 patients' data (89 BOT and 146 OC) to select the most relevant features, with which a simple decision tree model was constructed. The model was tested on the rest of 114 patients (89 BOT and 25 OC). The results were compared with the predictions produced by using the risk of ovarian malignancy algorithm (ROMA) and logistic regression model. Eight notable features were selected by MRMR, among which two were identified as the top features by the decision tree model: human epididymis protein 4 (HE4) and carcinoembryonic antigen (CEA). Particularly, CEA is a valuable marker for OC prediction in patients with low HE4. The model also yields better prediction result than ROMA. Machine learning approaches were able to accurately classify BOT and OC. Our goal is to derive a simple predictive model which also carries a good performance. Using our approach, we obtained a model that consists of just two biomarkers, HE4 and CEA. The model is simple to interpret and outperforms the existing OC prediction methods. It demonstrates that the machine learning approach has good potential in predictive modeling for the complex diseases.

MSCI: A multistate dataset for colposcopy image classification of cervical cancer screening

Cervical cancer is the second most common female cancer globally, and it is vital to detect cervical cancer with low cost at an early stage using automated screening methods of high accuracy, especially in areas with insufficient medical resources. Automatic detection of cervical intraepithelial neoplasia (CIN) can effectively prevent cervical cancer. Due to the deficiency of standard and accessible colposcopy image datasets, we present a dataset containing 4753 colposcopy images acquired from 679 patients in three states (acetic acid reaction, green filter, and iodine test) for detection of cervical intraepithelial neoplasia. Based on this dataset, a new computer-aided method for cervical cancer screening was proposed. We employed a wide range of methods to comprehensively evaluate our proposed dataset. Hand-crafted feature extraction methods and deep learning methods were used for the performance verification of the multistate colposcopy image (MSCI) dataset. Importantly, we propose a gated recurrent convolutional neural network (C-GCNN) for colposcopy image analysis that considers time series and combined multistate cervical images for CIN grading. The experimental results showed that the proposed C-GCNN model achieves the best classification performance in CIN grading compared with hand-crafted feature extraction methods and classic deep learning methods. The results showed an accuracy of 96.87 %, a sensitivity of 95.68 %, and a specificity of 98.72 %. A multistate colposcopy image dataset (MSCI) is proposed. A CIN grading model (C-GCNN) based on the MSCI dataset is established, which provides a potential method for automated cervical cancer screening.

Deep learning based cervical screening by the cross-modal integration of colposcopy, cytology, and HPV test

To develop and evaluate the colposcopy based deep learning model using all kinds of cervical images for cervical screening, and investigate the synergetic benefits of the colposcopy, the cytology test, and the HPV test for improving cervical screening performance. This study consisted of 2160 women who underwent cervical screening, there were 442 cases with the histopathological confirmed high-grade squamous intraepithelial lesion (HSIL) or cancer, and the remained 1718 women were controls. Three kinds of cervical images were acquired from colposcopy including the saline image of cervix after saline irrigation, the acetic acid image of cervix after applying acetic acid solution, and the iodine image of cervix after applying Lugol's iodine solution. Each kind of image was used to build a single-image based deep learning model by the VGG-16 convolutional neural network, respectively. A multiple-images based deep learning model was built using multivariable logistic regression (MLR) by combining the single-image based models. The performance of the visual inspection was also obtained. The results of the cytology test and HPV test were used to build a Cytology-HPV joint diagnostic model by MLR. Finally, a cross-modal integrated model was built using MLR by combining the multiple-images based deep learning model, the cytology test results, and the HPV test results. The performances of models were tested in an independent test set using the area under the receiver operating characteristic curve (AUC). The saline image, acetic acid image, and iodine image based deep learning models had AUC of 0.760, 0.791, and 0.840. The multiple-images based deep learning model achieved an improved AUC of 0.845. The AUC of the visual inspection was 0.751. The Cytology-HPV joint diagnostic model had an AUC of 0.837, which was higher than the cytology test (AUC = 0.749) and the HPV test (AUC = 0.742). The cross-modal integrated model achieved the best performance with AUC of 0.921. Combining all kinds of cervical images were benefit for improving the performance of the colposcopy based deep learning model, and more accurate cervical screening could be achieved by incorporating the colposcopy based deep learning model, the cytology test results, and the HPV test results.

Can natural language processing be effectively applied for audit data analysis in gynaecological oncology at a UK cancer centre?

The British Gynaecological Cancer Society (BGCS) has highlighted the disparity of ovarian cancer outcomes in the UK compared to other European countries. Therefore, cancer quality assurance audits and subspecialty training are important in improving the UK standard of care for these patients. The current workforce crisis afflicting the NHS creates difficulty in dedicating teams of clinicians to these audits. We present a single institution study to evaluate if NLP-generated code can improve the efficiency of ovarian cancer and subspeciality reaccreditations audits. We used the chat bot Google Bard to write Visual Basic Applications algorithms that utilise Excel files from electronic health records. Primary ovarian cancer data from 2019 to 2022 was retrospectively collected from the Cambridge University Hospital electronic health records. The surgical subspecialty reaccreditation audit analysed the 2022 surgical database. A modular coding approach with Google Bard was applied to generate audit algorithms. The time to complete these current audits was compared against the 2016 ovarian cancer and 2020 subspeciality reaccreditation audits. The previous ovarian cancer audit conducted in 2016 required 3 clinicians for the 135 cases and data collection required 1800 min. Data analysis was completed in 300 min. The current ovarian cancer audit allocated 2 clinicians to the 600 surgical cases. Data collection was completed in 3120 min, 3360 min for code development and 720 min for testing. The 2020 subspecialty reaccreditation audit was completed in 360 min. The 2022 subspecialty reaccreditation audit was completed in 1680 min, with 960 min for code development, 240 for debugging and 480 min for testing. We have demonstrated that NLP-generated code can significantly increase the efficiency of surgical quality assurance audits by eliminating the need for manual data analysis. With the current trajectory of NLP development, increasingly complex algorithms can be developed with minimal programming knowledge.

Cross-population evaluation of cervical cancer risk prediction algorithms

Cervical cancer is a preventable disease, despite being one of the most common types of female cancers worldwide. Integrating existing programs for cervical cancer screening with personalized risk prediction algorithms can improve population-level cancer prevention by enabling more targeted screening and contrive preventive healthcare innovations. While algorithms developed for cervical cancer risk prediction have shown promising performance in internal validation on more homogeneous populations, their ability to generalize to external populations remains to be assessed. To address this gap, we perform a cross-population comparative study of personalized prediction algorithms for more personalized cervical cancer screening. Using data from the Norwegian and Estonian populations, the algorithms are validated on internal and external datasets to study their potential biases and limitations when applied to different populations. We evaluate the algorithms in predicting progression from low-grade precancerous cervical lesions, simulating a clinically relevant application of more personalized risk stratification. As expected, our numerical experiments show that algorithm performance varies depending on the population. However, some algorithms show strong generalization capacity across different data sources. Using Kaplan-Meier estimates, we demonstrate the strengths and limitations of the algorithms in detecting cancer progression over time by comparing to the trends observed from data. We assess their overall discrimination performance in personalized risk predictions by analyzing the accuracy and confidence in individual risk estimates. This study examines the effectiveness of personalized prediction algorithms across different populations. Our results demonstrate the potential for generalizing risk prediction algorithms to external populations. These findings highlight the importance of considering population diversity when developing risk prediction algorithms.

Using sexual orientation and gender identity data in electronic health records to assess for disparities in preventive health screening services

Lesbian, gay, bisexual, transgender, and queer (LGBTQ) populations have an increased risk of multiple adverse health outcomes. Capturing patient data on sexual orientation and gender identity (SOGI) in the electronic health record (EHR) can enable healthcare organizations to identify inequities in the provision of preventive health screenings and other quality of care services to their LGBTQ patients. However, organizations may not be familiar with methods for analyzing and interpreting SOGI data to detect health disparities. To assess an approach for using SOGI EHR data to identify potential screening disparities of LGBTQ patients within distinct healthcare organizations. Five US federally qualified health centers (FQHCs) retrospectively extracted three consecutive months of EHR patient data on SOGI and routine screening for cervical cancer, tobacco use, and clinical depression. The screening data were stratified across SOGI categories. Chi-Square and Fisher's Exact test were used to identify statistically significant differences in screening compliance across SOGI categories within each FQHC. In all FQHCs, cervical cancer screening percentages were lower among lesbian/gay patients than among bisexual and straight/heterosexual patients. In three FQHCs, cervical cancer screening percentages were lower for transgender men than for cisgender (i.e., not transgender) women. Within each FQHC, we observed statistically significant associations (P < 0.05) between SOGI categories and at least one screening measure. The small number of transgender patients, and limitations in EHR functionality, created challenges in interpretation of SOGI data. To our knowledge, this is the first published report of using SOGI data from EHRs to detect potential disparities in healthcare services to LGBTQ patients. Our finding that lesbian/gay and transgender male patients had lower cervical cancer screening rates compared to heterosexual, bisexual, and cisgender women, is consistent with the research literature and suggests that using SOGI EHR data to detect preventive screening disparities has value. EHR functionality should allow for cross-checking gender identity with sex assigned at birth to reduce errors in data interpretation. Additional functionality, like clinical decision support based on anatomical inventories rather than gender identity, is needed to more accurately identify services that transgender patients need.