Using similar patients to predict complication in patients with diabetes, hypertension, and lipid disorder: a domain knowledge-infused convolutional neural network approach.

OBJECTIVE: This study aims to develop a convolutional neural network-based learning framework called domain knowledge-infused convolutional neural network (DK-CNN) for retrieving clinically similar patient and to personalize the prediction of macrovascular complication using the retrieved patients. MATERIALS AND METHODS: We use the electronic health records of 169 434 patients with diabetes, hypertension, and/or lipid disorder. Patients are partitioned into 7 subcohorts based on their comorbidities. DK-CNN integrates both domain knowledge and disease trajectory of patients over multiple visits to retrieve similar patients. We use normalized discounted cumulative gain (nDCG) and macrovascular complication prediction performance to evaluate the effectiveness of DK-CNN compared to state-of-the-art models. Ablation studies are conducted to compare DK-CNN with reduced models that do not use domain knowledge as well as models that do not consider short-term, medium-term, and long-term trajectory over multiple visits. RESULTS: Key findings from this study are: (1) DK-CNN is able to retrieve clinically similar patients and achieves the highest nDCG values in all 7 subcohorts; (2) DK-CNN outperforms other state-of-the-art approaches in terms of complication prediction performance in all 7 subcohorts; and (3) the ablation studies show that the full model achieves the highest nDCG compared with other 2 reduced models. DISCUSSION AND CONCLUSIONS: DK-CNN is a deep learning-based approach which incorporates domain knowledge and patient trajectory data to retrieve clinically similar patients. It can be used to assist physicians who may refer to the outcomes and past treatments of similar patients as a guide for choosing an effective treatment for patients.

Diabetes medication recommendation system using patient similarity analytics.

Type-2 diabetes mellitus (T2DM) is a medical condition in which oral medications avail to patients to curb their hyperglycaemia after failed dietary therapy. However, individual responses to the prescribed pharmacotherapy may differ due to their clinical profiles, comorbidities, lifestyles and medical adherence. One approach is to identify similar patients within the same community to predict their likely response to the prescribed diabetes medications. This study aims to present an evidence-based diabetes medication recommendation system (DMRS) underpinned by patient similarity analytics. The DMRS was developed using 10-year electronic health records of 54,933 adult patients with T2DM from six primary care clinics in Singapore. Multiple clinical variables including patient demographics, comorbidities, laboratory test results, existing medications, and trajectory patterns of haemoglobin A1c (HbA1c) were used to identify similar patients. The DMRS was evaluated on four groups of patients with comorbidities such as hyperlipidaemia (HLD) and hypertension (HTN). Recommendations were assessed using hit ratio which represents the percentage of patients with at least one recommended sets of medication matches exactly the diabetes prescriptions in both the type and dosage. Recall, precision, and mean reciprocal ranking of the recommendation against the diabetes prescriptions in the EHR records were also computed. Evaluation against the EHR prescriptions revealed that the DMRS recommendations can achieve hit ratio of 81% for diabetes patients with no comorbidity, 84% for those with HLD, 78% for those with HTN, and 75% for those with both HLD and HTN. By considering patients' clinical profiles and their trajectory patterns of HbA1c, the DMRS can provide an individualized recommendation that resembles the actual prescribed medication and dosage. Such a system is useful as a shared decision-making tool to assist clinicians in selecting the appropriate medications for patients with T2DM.

Using Domain Knowledge and Data-Driven Insights for Patient Similarity Analytics.

Patient similarity analytics has emerged as an essential tool to identify cohorts of patients who have similar clinical characteristics to some specific patient of interest. In this study, we propose a patient similarity measure called D3K that incorporates domain knowledge and data-driven insights. Using the electronic health records (EHRs) of 169,434 patients with either diabetes, hypertension or dyslipidaemia (DHL), we construct patient feature vectors containing demographics, vital signs, laboratory test results, and prescribed medications. We discretize the variables of interest into various bins based on domain knowledge and make the patient similarity computation to be aligned with clinical guidelines. Key findings from this study are: (1) D3K outperforms baseline approaches in all seven sub-cohorts; (2) our domain knowledge-based binning strategy outperformed the traditional percentile-based binning in all seven sub-cohorts; (3) there is substantial agreement between D3K and physicians (κ = 0.746), indicating that D3K can be applied to facilitate shared decision making. This is the first study to use patient similarity analytics on a cardiometabolic syndrome-related dataset sourced from medical institutions in Singapore. We consider patient similarity among patient cohorts with the same medical conditions to develop localized models for personalized decision support to improve the outcomes of a target patient.

Progression-Free Survival Prediction in Patients with Nasopharyngeal Carcinoma after Intensity-Modulated Radiotherapy: Machine Learning vs. Traditional Statistics.

BACKGROUND: The Cox proportional hazards (CPH) model is the most commonly used statistical method for nasopharyngeal carcinoma (NPC) prognostication. Recently, machine learning (ML) models are increasingly adopted for this purpose. However, only a few studies have compared the performances between CPH and ML models. This study aimed at comparing CPH with two state-of-the-art ML algorithms, namely, conditional survival forest (CSF) and DeepSurv for disease progression prediction in NPC. METHODS: From January 2010 to March 2013, 412 eligible NPC patients were reviewed. The entire dataset was split into training cohort and testing cohort in a ratio of 90%:10%. Ten features from patient-related, disease-related, and treatment-related data were used to train the models for progression-free survival (PFS) prediction. The model performance was compared using the concordance index (c-index), Brier score, and log-rank test based on the risk stratification results. RESULTS: DeepSurv (c-index = 0.68, Brier score = 0.13, log-rank test p = 0.02) achieved the best performance compared to CSF (c-index = 0.63, Brier score = 0.14, log-rank test p = 0.38) and CPH (c-index = 0.57, Brier score = 0.15, log-rank test p = 0.81). CONCLUSIONS: Both CSF and DeepSurv outperformed CPH in our relatively small dataset. ML-based survival prediction may guide physicians in choosing the most suitable treatment strategy for NPC patients.

Patient similarity analytics for explainable clinical risk prediction.

BACKGROUND: Clinical risk prediction models (CRPMs) use patient characteristics to estimate the probability of having or developing a particular disease and/or outcome. While CRPMs are gaining in popularity, they have yet to be widely adopted in clinical practice. The lack of explainability and interpretability has limited their utility. Explainability is the extent of which a model's prediction process can be described. Interpretability is the degree to which a user can understand the predictions made by a model. METHODS: The study aimed to demonstrate utility of patient similarity analytics in developing an explainable and interpretable CRPM. Data was extracted from the electronic medical records of patients with type-2 diabetes mellitus, hypertension and dyslipidaemia in a Singapore public primary care clinic. We used modified K-nearest neighbour which incorporated expert input, to develop a patient similarity model on this real-world training dataset (n = 7,041) and validated it on a testing dataset (n = 3,018). The results were compared using logistic regression, random forest (RF) and support vector machine (SVM) models from the same dataset. The patient similarity model was then implemented in a prototype system to demonstrate the identification, explainability and interpretability of similar patients and the prediction process. RESULTS: The patient similarity model (AUROC = 0.718) was comparable to the logistic regression (AUROC = 0.695), RF (AUROC = 0.764) and SVM models (AUROC = 0.766). We packaged the patient similarity model in a prototype web application. A proof of concept demonstrated how the application provided both quantitative and qualitative information, in the form of patient narratives. This information was used to better inform and influence clinical decision-making, such as getting a patient to agree to start insulin therapy. CONCLUSIONS: Patient similarity analytics is a feasible approach to develop an explainable and interpretable CRPM. While the approach is generalizable, it can be used to develop locally relevant information, based on the database it searches. Ultimately, such an approach can generate a more informative CRPMs which can be deployed as part of clinical decision support tools to better facilitate shared decision-making in clinical practice.

MRI-based radiomics signature is a quantitative prognostic biomarker for nasopharyngeal carcinoma.

This study aimed to develop prognosis signatures through a radiomics analysis for patients with nasopharyngeal carcinoma (NPC) by their pretreatment diagnosis magnetic resonance imaging (MRI). A total of 208 radiomics features were extracted for each patient from a database of 303 patients. The patients were split into the training and validation cohorts according to their pretreatment diagnosis date. The radiomics feature analysis consisted of cluster analysis and prognosis model analysis for disease free-survival (DFS), overall survival (OS), distant metastasis-free survival (DMFS) and locoregional recurrence-free survival (LRFS). Additionally, two prognosis models using clinical features only and combined radiomics and clinical features were generated to estimate the incremental prognostic value of radiomics features. Patients were clustered by non-negative matrix factorization (NMF) into two groups. It showed high correspondence with patients' T stage (p < 0.00001) and overall stage information (p < 0.00001) by chi-squared tests. There were significant differences in DFS (p = 0.0052), OS (p = 0.033), and LRFS (p = 0.037) between the two clustered groups but not in DMFS (p = 0.11) by log-rank tests. Radiomics nomograms that incorporated radiomics and clinical features could estimate DFS with the C-index of 0.751 [0.639, 0.863] and OS with the C-index of 0.845 [0.752, 0.939] in the validation cohort. The nomograms improved the prediction accuracy with the C-index value of 0.029 for DFS and 0.107 for OS compared with clinical features only. The DFS and OS radiomics nomograms developed in our study demonstrated the excellent prognostic estimation for NPC patients with a noninvasive way of MRI. The combination of clinical and radiomics features can provide more information for precise treatment decision.

Convolutional neural network for cell classification using microscope images of intracellular actin networks.

Automated cell classification is an important yet a challenging computer vision task with significant benefits to biomedicine. In recent years, there have been several studies attempted to build an artificial intelligence-based cell classifier using label-free cellular images obtained from an optical microscope. Although these studies showed promising results, such classifiers were not able to reflect the biological diversity of different types of cell. While in terms of malignant cell, it is well-known that intracellular actin filaments are altered substantially. This is thought to be closely related to the abnormal growth features of tumor cells, their ability to invade surrounding tissues and also to metastasize. Therefore, being able to classify different types of cell based on their biological behaviors using automated technique is more advantageous. This article reveals the difference in the actin cytoskeleton structures between breast normal and cancer cells, which may provide new information regarding malignant changes and be used as additional diagnostic marker. Since the features cannot be well detected by human eyes, we proposed the application of convolutional neural network (CNN) in cell classification based on actin-labeled fluorescence microscopy images. The CNN was evaluated on a large number of actin-labeled fluorescence microscopy images of one human normal breast epithelial cell line and two types of human breast cancer cell line with different levels of aggressiveness. The study revealed that the CNN performed better in the cell classification task compared to a human expert.

Dosimetric predictors of temporal lobe injury after intensity-modulated radiotherapy for T4 nasopharyngeal carcinoma: a competing risk study.

BACKGROUND: In patients with T4 nasopharyngeal carcinoma (NPC), death may occur prior to the occurrence of temporal lobe injury (TLI). Because such competing risk death precludes the occurrence of TLI and thus the competing risk analysis should be applied to TLI research. The aim was to investigate the incidence and predictive factors of TLI after intensity-modulated radiotherapy (IMRT) among T4 NPC patients. METHODS: From March 2008 to December 2014, T4 NPC patients treated with full-course radical IMRT at our center were reviewed retrospectively. A nested case-control study was designed for this cohort of patients. The cases were patients with TLI diagnosed by MRI during the follow-up period, and the controls were patients without TLI after IMRT matched 1:1 to each case by gender, age at diagnosis, intercranial involvement, and follow-up time. The end point was time to TLI or death without prior TLI. We analyzed the cumulative incidence function (CIF) and performed a competing risk regression model to identify the predictors of TLI. RESULTS: With a median follow-up of 40.1 months, 63 patients (63/506, 12.5%) developed TLI as diagnosed by MRI, and 136 deaths occurred during the period. The cumulative incidence of TLI at 5 years was 13.2%, while 26.7% died without prior TLI. The univariate analysis showed that all selected dosimetric parameters were associated with the occurrence of TLI. On multivariate analysis, D1cc and V20 remained statistically significant. Based on the area-under-the-curve (AUC) values, D1cc was considered the most predictive. The patients with D1cc > 71.14 Gy had a 7.920-fold increased risk of TLI compared with those with D1cc ≤71.14 Gy (P < 0.05). Similarly, V20 > 42.22 cc was found to result in a statistically significant higher risk of TLI (subdistribution hazard ratio [sHR] =3.123, P < 0.05). CONCLUSIONS: TL D1cc and V20 were predictive of TLI after IMRT for T4 NPC. They should be considered as first and second priorities of dose constraints of the TL. D1cc ≤71.14 Gy and V20 ≤ 42.22 cc could be useful dose-volume constraints for reducing the occurrence of TLI during IMRT treatment planning without obviously compromising the tumor coverage.

Prognostic value of nutritional markers in nasopharyngeal carcinoma patients receiving intensity-modulated radiotherapy: a propensity score matching study.

PURPOSE: To investigate the prognostic value of nutritional markers for survival in nasopharyngeal carcinoma (NPC) patients receiving intensity-modulated radiotherapy (IMRT), with or without chemotherapy. PATIENTS AND METHODS: This retrospective study included 412 NPC patients who received IMRT-based treatment. Weight loss (WL) during treatment, hemoglobin level (Hb) and serum albumin level (Alb) before treatment were measured. The prognostic values of these markers for overall survival (OS), locoregional recurrence-free survival (LRFS) and distant metastasis-free survival (DMFS) were analyzed using Kaplan-Meier method and Cox proportional hazards regression analysis. Propensity score matching was performed to reduce the effect of confounders. RESULTS: WL, Hb and Alb were significantly correlated with each other and inflammatory markers. Adjusted Cox regression analysis showed that critical weight loss (CWL) (WL≥5%) was an independent prognostic factor for OS (HR: 2.399, 95% CI: 1.267-4.540, P=0.007) and LRFS (HR: 2.041, 95% CI: 1.052-3.960, P=0.035), while low pretreatment Hb was independently associated with poor DMFS (HR: 2.031, 95% CI: 1.144-3.606, P=0.016). However, no significant correlation was found between Alb and survival in our study cohort. The prognostic value of these markers was further confirmed in the propensity-matched analysis. CONCLUSION: CWL, Hb and Alb have a significant impact on survival in NPC patients undergoing IMRT. They can be utilized in combination with conventional staging system to predict the prognosis of NPC patients treated with IMRT.

Prognostic value of inflammation-based prognostic index in patients with nasopharyngeal carcinoma: a propensity score matching study.

PURPOSE: The aim of this article is to investigate the significance of pretreatment prognostic nutritional index (PNI), systemic immune-inflammation index (SII), and their combination in nasopharyngeal carcinoma (NPC) patients receiving intensity-modulated radiotherapy (IMRT). MATERIALS AND METHODS: A total of 585 patients were included. PNI and SII were calculated within 2 weeks prior to treatment. The optimal cutoff points were determined based on receiver operating characteristics curve analysis. The correlation between variables was analyzed. Kaplan-Meier method and Cox proportional hazards model were performed to evaluate the impact of both indices on overall survival (OS), progression-free survival (PFS) and distant metastasis-free survival (DMFS). Further propensity score matching (PSM) was carried out to minimize the effects of confounders. RESULTS: The optimal cutoff point of 53.0 for PNI and 527.20 for SII were selected. Pearson correlation coefficient showed an inverse correlation between PNI and SII (r = -0.232, P < 0.001). Multivariate analysis demonstrated that pretreatment PNI was an independent prognostic factor for OS (P = 0.047) and DMFS (P = 0.002) while pretreatment SII was an independent prognostic factor for OS (P = 0.003), PFS (P = 0.002), and DMFS (P = 0.002). After PSM, both parameters remained as independent prognosticators of survival. Additional prognostic value was observed in the combined use of PNI and SII. CONCLUSION: Pretreatment PNI and SII are promising indicators of survival in NPC patients undergoing IMRT. They can be utilized to refine current TNM staging system in predicting prognosis and developing an individualized treatment in these patients.

Prognostic values of hematological biomarkers in nasopharyngeal carcinoma patients treated with intensity-modulated radiotherapy.

PURPOSE: In this study, we evaluated the prognostic values of hematological biomarkers in primary nasopharyngeal carcinoma (NPC) patients receiving definitive intensity-modulated radiotherapy (IMRT). METHODS: There were 427 NPC patients enrolled between January 2010 and March 2013 at Fudan University Shanghai Cancer Center. Pre-treatment absolute neutrophil count (ANC), platelet count (APC), lymphocyte count (ALC), neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) were collected as prognostic biomarkers. The Kaplan-Meier method and log-rank test were utilized to calculate progression-free survival (PFS) and overall survival (OS). The Cox proportional hazard models were applied to assess variables. RESULTS: ANC, APC and ALC were declined, while NLR and PLR were elevated significantly after therapy (P < 0.001 each). On multivariate analysis, pre-treatment NLR ≥ 2.32 was associated with shortened OS (P = 0.048) and PFS (P = 0.008), whereas PLR ≥ 123.0 was related with inferior OS (P = 0.032), yet it was not correlated with PFS (P = 0.161). CONCLUSIONS: High pre-treatment NLR and PLR indicated poor survival in NPC patients treated with IMRT-based therapy. As easily accessible and economically feasible biomarkers, NLR and PLR can be applied into clinical practice, in combination with current TNM staging, to design a more personalized treatment in these patients.

The prognostic value of preoperative prognostic nutritional index in patients with hypopharyngeal squamous cell carcinoma: a retrospective study.

BACKGROUND: To analyze the prognostic value of preoperative prognostic nutritional index (PNI) in predicting the survival outcome of hypopharyngeal squamous cell carcinoma (HPSCC) patients receiving radical surgery. METHODS: From March 2006 to August 2016, 123 eligible HPSCC patients were reviewed. The preoperative PNI was calculated as serum albumin (g/dL) × 10 + total lymphocyte count (mm-3) × 0.005. These biomarkers were measured within 2 weeks prior to surgery. The impact of preoperative PNI on overall survival (OS), progression-free survival (PFS), locoregional recurrence-free survival (LRFS) and distant metastasis-free survival (DMFS) were analyzed using Kaplan-Meier method and Cox proportional hazards model. RESULTS: Median value of 52.0 for the PNI was selected as the cutoff point. PNI value was then classified into two groups: high PNI (> 52.0) versus low PNI (≤ 52.0). Multivariate analysis showed that high preoperative PNI was an independent prognostic factor for better OS (P = 0.000), PFS (P = 0.001), LRFS (P = 0.005) and DMFS (P = 0.016). CONCLUSIONS: High PNI predicts superior survival in HPSCC patients treated with radical surgery. As easily accessible biomarkers, preoperative PNI together with the conventional TNM staging system can be utilized to enhance the accuracy in predicting survival and determining therapy strategies in these patients.

Pre-treatment Serum Lactate Dehydrogenase is Predictive of Survival in Patients with Nasopharyngeal Carcinoma Undergoing Intensity-Modulated Radiotherapy.

Objective: To analyze the prognostic value of pre-treatment serum lactate dehydrogenase (SLDH) level in patients with nasopharyngeal carcinoma (NPC) receiving intensity-modulated radiotherapy (IMRT) with or without chemotherapy. Methods: From January 2010 to March 2013, 427 eligible patients were reviewed. Pre-treatment SLDH level was measured within 2 weeks prior to treatment. Receiver operating characteristic (ROC) curve analysis was performed to select the optimal cutoff point. The impact of pre-treatment SLDH on overall survival (OS), progression-free survival (PFS) and distant metastasis-free survival (DMFS) were analyzed using Kaplan-Meier method and Cox proportional hazards model. Further propensity score matching was carried out to adjust bias. Results: The optimal cutoff point of 168.5 IU/L was selected based on ROC curve analysis. Multivariate analysis showed that high pre-treatment SLDH level was an independent prognostic factor for OS (P=0.001), PFS (P=0.004) and DMFS (P=0.001). After propensity score matching was performed, it remained to be significantly associated with poor OS (P=0.009), PFS (P=0.015) and DMFS (P=0.008) in the adjusted model. Conclusion: High pre-treatment SLDH level predicts poor survival in patients with NPC treated with IMRT-based therapy. As a routinely performed biomarker, pre-treatment SLDH can be utilized in combination with current Tumor-Node-Metastasis staging to predict survival and to plan a personalized treatment in these patients.