Abstract B059: Cancer Recurrence and Anxiety/Depression: High Accuracy Artificial Intelligence Predictive Modeling

1571 Background: Wearable activity trackers could provide useful data for managing cancer patients with respect to treatment selection, toxicity monitoring, and implementation of supportive care measures. Here, we seek to evaluate the association between daily step counts and hospitalizations in a cohort of patients with head and neck cancer (HNC). Methods: This analysis consists of patients enrolled in one of three prospective trials involving activity monitoring (NCT02649569, NCT03115398, NCT03102229) during chemoradiation. Study subjects were asked to wear a commercial fitness tracker continuously during the therapy. ECOG performance status (PS) was assessed at baseline, and quality of life (QoL) EORTC QLQ-C30 questionnaires were completed weekly. Multivariable Cox regression models with time-dependent covariates (average step count over the past 3 days, most recent QoL score) and time-fixed covariates (age, sex, baseline PS, study number, baseline tumor volume, and treatment setting [definitive versus postoperative]) were used to identify predictors of first hospital admission during the chemoradiotherapy course. In addition to the Cox regression models, linear mixed models were fitted with daily step count as the dependent variable to examine its relationship with certain independent variables including age, sex, weekend status, days after treatment initiation, and study number. Results: Sixty-six HNC patients who received chemoradiotherapy between 2015 and 2019 were included in the analysis. Median age was 60 (range: 27-88). 47% of patients had ECOG PS score 0, 47% ECOG score 1, and 6% ECOG score 2. 29% of patients had HPV-positive oropharyngeal tumors, and the most common other tumor subsites were larynx (27%), and nasopharynx (12%). The Cox regression survival model demonstrated a 26% reduction in the short-term hospitalization risk for every 1000 daily steps (averaged over the past 3 days, hazard ratio 0.74; 95% confidence interval (CI) 0.55-0.98, p = 0.0367). Hospitalizations were not significantly associated with most recent QoL or baseline ECOG PS. Additionally, according to the linear mixed model results, daily step count was not associated with age (p = 0.8048). Study subjects moved less on weekends (on average 245 fewer steps on weekends than weekdays, 95% CI 134-357, p < 0.0001). Also, an increase in most recently measured ECOG PS was associated with a decrease in daily step count (167 fewer steps for every increase in ECOG PS, 95% CI -289 to -45, p < 0.0072). Conclusions: Daily step counts are a dynamic predictor of hospitalizations in patients undergoing chemoradiotherapy for head and neck cancer. Interventional studies are needed to demonstrate feasibility of leveraging physical activity data to optimize supportive care during cancer therapy and enhance cancer care quality. Clinical trial information: NCT02649569, NCT03115398, NCT03102229.

Response

BackgroundFrailty is a strong predictor of adverse outcomes. However, longitudinal drivers of frailty are not well understood. This study aimed at investigating the longitudinal trajectories of a frailty index (FI) from adulthood to late life and identifying the factors associated with the level and rate of change in FI.MethodsAn age-based latent growth curve analysis was performed in the Swedish Adoption/Twin Study of Aging (N = 1,842; aged 29–102 years) using data from up to 15 measurement waves across 27 years. A 42-item FI was used to measure frailty at each wave.ResultsA bilinear, two-slope model with a turning point at age 65 best described the age-related change in FI, showing that the increase in frailty was more than twice as fast after age 65. Underweight, obesity, female sex, overweight, being separated from one’s co-twin during childhood, smoking, poor social support, and low physical activity were associated with a higher FI at age 65, with underweight having the largest effect size. When tested as time-varying covariates, underweight and higher social support were associated with a steeper increase in FI before age 65, whereas overweight and obesity were associated with less steep increase in FI after age 65.ConclusionsFactors associated with the level and rate of change in frailty are largely actionable and could provide targets for intervention. As deviations from normal weight showed the strongest associations with frailty, future public health programs could benefit from monitoring of individuals with abnormal BMI, especially those who are underweight.

Construction monitoring of cutter suction dredgers (CSD) is of great significance in ensuring dredging efficiency. However, existing models have not taken into account the physical constraints in the physical system of a CSD, which limits further improvements in prediction accuracy. To this end, this paper proposes a physics-assisted deep learning model for improved construction performance monitoring of CSD. The data-driven Lossu and the physics-driven Lossr are combined to form an improved physics-assisted loss function (PALF). And then, a physics-assisted deep learning (PADL) model incorporating PALF is developed to predict the construction productivity. In the case application, evaluation across three deep learning models confirms the feasibility and effectiveness of PALF for productivity prediction. The results show that the PALF-based PADL model achieves markedly improved prediction accuracy, reducing the mean absolute error by 20.33–54.33%. Across six training-set sizes (1000–11,000 samples), the improvement is larger in small-data scenarios, highlighting PADL’s strong low-sample robustness. The proposed model can effectively complement physical sensors in monitoring construction parameters and provide reliable decision support for assessing the operational state of CSDs.

Lung cancer is one of the most common malignant tumors with the highest mortality, with about 85% of cases of the disease being non-small cell lung cancer. To date, adjuvant chemotherapy based on platinum preparations remains the standard of treatment for patients with radically operated stage II or III non-small cell lung cancer. However, a large proportion of patients still have a risk of developing a recurrence of lung cancer even after complete resection of the tumor. That is why there is a need to search for new methods of treatment of early stages of NSCLC, which will minimize the likelihood of postoperative relapses and improve survival in a potentially curable group of patients. Already today, clinical trials are beginning to advance the treatment of non-small cell lung cancer in the early stages beyond the standard cytotoxic chemotherapy. Immunotherapy, represented by immune checkpoint inhibitors, is being investigated in an increasing number of clinical trials in patients with operable NSCLC at early stages, gradually enriching existing treatment methods. At the same time, some issues related to adjuvant immunotherapy have yet to be considered. The choice of the drug, the use of monotherapy or combined treatment regimens remain unclear. This review examines the progress of research aimed at improving adjuvant therapy through the inclusion of immune checkpoint inhibitors in the treatment of early stages of resectable non-small cell lung cancer. Adjuvant immunotherapy can improve relapse-free survival in individual patients with resectable lung cancer, and current or planned studies using biomarkers and immunotherapy may also ultimately lead to an improvement in the overall survival of this group of patients.

Area under the curve may hide poor generalisation to external datasets

Objectives: To assess the recurrence rate of breast cancer after modified radical mastectomy in cases of breast cancer at tertiary care hospital. Material and methods: Between the March 2020 to September 2020 (over the period of 6 months) total 110 women with breast cancer having age range 30-60 years were recruited from Department of Surgery, Bahawal Victoria Hospital Bahawalpur for this cross sectional study. Modified radical mastectomy was performed in all selected patients. At 6 months follow up, all the selected patients was again examined for recurrence of breast cancer. Results: Total 110 patients with breast cancer were recruited. Mean age of the patients was 43.56 ± 8.9 years. Recurrence of breast cancer was found in 25 (23%) cases. Total 11 (10%) patients belonged to age group <30 years followed by 13 (11.82%) patients to age group 30-40 years, 41 (37.27%) to age group 41-50 years and 45 (40.91%) patients to age group 51-60 years. Recurrence of breast cancer was noted in 2 (18.18%) patients, 3 (23.08%) patients, 10 (24.39%) patients and 10 (22.22%) patients respectively. Statistically insignificant association of recurrence with age group was noted with p value 0.9776. Conclusion: Results of this study showed a higher rate of recurrence of breast carcinoma after modified radical mastectomy. Most of the patients belonged to 5th decade of life. Parity, educational status and marital status showed no association with recurrence of breast cancer.

e12023 Background: The 21-gene recurrence score assay is a validated genomic predictor of cancer recurrence and the benefits of adjuvant chemotherapy in hormone receptor-positive, HER2-negative early breast cancer patients. However, the assay is rather expensive and complex, and many patients in our country can not afford it. We performed a retrospective study to evaluate the association between traditional clinicopathological features and 21-gene recurrence score in patients with node-negative, hormone receptor-positive, HER2-negative breast cancer. Methods: A total of 76 consecutive patients with node-negative, hormone receptor-positive, HER2-negative breast cancer who underwent 21-gene recurrence score testing in our hospital from 2015 to 2016 were enrolled. Data including age, tumor size, histological type, tumor grade were collected. Estrogen receptor (ER) expression, progesterone receptor (PR) expression, Ki-67 index, p53 and androgen receptor (AR) expression were detected by immunohistochemical assay. 21-gene recurrence score assay was performed by RT-PCR. The correlation between clinicopathological characteristics and 21-gene recurrence score assay was analyzed by using nonparametric tests. Results: Among the 76 patients, 43 (56.6%) had a low recurrence score of < 18, 13 (17.1%) had an intermediate recurrence score of 18-31, and 20 (26.3%) had a high recurrence score of ≥31. There was a significant difference in recurrence score in patients divided by PR expression ( P=0.027). Patients with low PR expression (<20%) had a higher recurrence score as compared to those with high PR expression. Tumor grade was observed to be significantly correlated with recurrence score ( P=0.004). No significant associations were found between recurrence score and age, tumor size, histological type, ER expression, Ki-67 index, p53 or AR expression in this study. Conclusions: Our study shows that low PR expression is associated with higher 21-gene recurrence score, and PR expression may be a promising predictor of 21-gene recurrence score assay in node-negative, hormone receptor-positive, HER2-negative breast cancer patients.

To develop a deep learning (DL) model for predicting disease-free survival (DFS) in clinical stage I lung cancer patients who underwent surgical resection using pre-treatment CT images, and further validate it in patients receiving stereotactic body radiation therapy (SBRT). A retrospective cohort of 2489 clinical stage I non-small cell lung cancer (NSCLC) patients treated with operation (2015-2017) was enrolled to develop a DL-based DFS prediction model. Tumor features were extracted from CT images using a three-dimensional convolutional neural network. External validation was performed on 248 clinical stage I patients receiving SBRT from two hospitals. A clinical model was constructed by multivariable Cox regression for comparison. Model performance was evaluated with Harrell's concordance index (C-index), which measures the model's ability to correctly rank survival times by comparing all possible pairs of subjects. In the surgical cohort, the DL model effectively predicted DFS with a C-index of 0.85 (95% CI: 0.80-0.89) in the internal testing set, significantly outperforming the clinical model (C-index: 0.76). Based on the DL model, 68 patients in the SBRT cohort identified as high-risk had significantly worse DFS compared to the low-risk group (p < 0.01, 5-year DFS rate: 34.7% vs 77.4%). The DL-score was demonstrated to be an independent predictor of DFS in both cohorts (p < 0.01). The CT-based DL model improved DFS prediction in clinical stage I lung cancer patients, identifying populations at high risk of recurrence and metastasis to guide clinical decision-making. Question The recurrence or metastasis rate of early-stage lung cancer remains high and varies among patients following radical treatments such as surgery or SBRT. Findings This CT-based DL model successfully predicted DFS and stratified varying disease risks in clinical stage I lung cancer patients undergoing surgery or SBRT. Clinical relevance The CT-based DL model is a reliable predictive tool for the prognosis of early-stage lung cancer. Its accurate risk stratification assists clinicians in identifying specific patients for personalized clinical decision making.

In medical imaging, the effective detection and classification of Breast Cancer (BC) is a current research important task because of the still existing difficulty to distinguish abnormalities from normal breast tissues due to their subtle appearance and ambiguous margins and distinguish abnormalities from the normal breast. Moreover, BC detection based on an automated detection model is needed, because manual diagnosis faces problems due to cost and shortage of skilled manpower, and also takes a very long time. Using deep learning and ensemble feature selection techniques, in this paper, a novel framework is introduced for classifying BC from histopathology images. The five primary steps of the suggested framework are as follows: 1) to make the largest original dataset and then deep learning model with data augmentation to improve the learning. 2) The best features are selected by an Ensemble Filter Feature selection Method (EFFM) which combines the best feature subsets to produce the final feature subsets. 3) Then the pruned Convolution Neural Network (CNN) model is utilized to extract the optimal features. 4) Finally, the classification is done through the Triplet Attention based Efficient Network (TAENet) classifier. The suggested model produces a 98% accuracy rate after being trained and tested on two different histopathology imaging datasets including images from four different data cohorts. Subsequently, the suggested strategy outperforms the conventional ones since the ensemble filter habitually acquires the best features, and experimental results demonstrate the importance of the proposed approach

Developing deep learning-based strategies to predict the risk of hepatocellular carcinoma among patients with nonalcoholic fatty liver disease from electronic health records

The long-term follow-up on local control of breast carcinoma treated by external beam irradiation plus iridium 192 implant

Purpose: The clinical data of HER-2 negative breast cancer patients were analyzed to investigate factors that influence bone metastasis. Methods: A total of 754 HER-2 negative breast cancer patients composed of the bone metastasis (237 patients) and non-bone metastasis (481 patients) group, were retrospectively and systematically evaluated for their clinicopathological characteristics, treatment modalities and their influencing factors. Both groups were administered a combined treatment of capecitabine and docetaxel and the statistical association between these factors and the development of bone metastases was investigated. Primary tumor location was identified through imaging while the hormone receptor status and tumour molecular classification were assessed by immunohistochemistry (IHC) on tumor samples and protein expression/genetic profiling, respectively. Results: A significant difference was seen between the groups in terms of T-stage, N-stage, hormone receptors, tumour molecular classification, axillary lymph node metastasis and quality of life scores (FACT-B) after combined therapy with capecitabine plus docetaxel (p &lt; 0.05), while statistical significance was absent in terms of age and location of the primary site (p &gt; 0.05). Furthermore, logistic regression analysis revealed that high T-stage, Luminal A molecular staging and the occurrence of axillary lymph node metastases were all risk factors for bone metastases in HER-2 negative breast cancer patients (p &lt; 0.001). Conclusion: Combined therapy with capecitabine and docetaxel is effective in the treatment of bone metastases in patients with HER-2 negative breast cancer. A more comprehensive study involving a larger population of diverse patients will be conducted in the future to provide more reliable clinical data.

Deep learning models for solar irradiance forecasting: A comprehensive review

The annual incidence of upper gastrointestinal hemorrhage (UGIB) is about 60 cases/100,000 people, and about 40% of UGIB patients have hemorrhagic ulcers. Ulcer formation is often associated with Helicobacter pylori (H. pylori) infection, non-steroidal anti-inflammatory drugs (NSAIDs) use and other factors, so ulcerative disease is the main cause of upper gastrointestinal bleeding. H. pylori induces chronic superficial gastritis with neutrophils infiltrating into the mucosa, so it is assumed that H. pylori infection is the basis of bleeding lesions. H. pylori infection is widespread worldwide, with about 50% of the population carrying the bacteria. Mortality during hospitalization is higher in patients with UGIB because rebleeding significantly increases the risk of death, especially if timely intervention is not provided. Rebleeding may also lead to severe complications such as shock and multiple organ failure. At present, the commonly used clinical scores for UGIB patients mainly include Rockall score (RS), AIMS65 score and Glasgow-Blatchford score (GBS). Because some hospitals are limited by local medical and health conditions, they lack timely and accurate endoscopic diagnosis and treatment equipment, and it is difficult to make accurate and timely judgments on patients. In this experiment, 254 patients with upper digestive tract hemorrhage from Shengjing Hospital affiliated to China Medical University were collected, and the clinical indicators and information of H. pylori infection, age, shock state, concomitant disease, H. pylori infection degree, systolic blood pressure, blood urea nitrogen, hemoglobin, pulse, black stool, syncope, liver disease and other patients were finally collected. We analyzed the correlation between various clinical indicators and rebleeding in hospitalized patients. Based on the collected clinical information and laboratory indicators, this study constructed a deep learning model, the data is divided into four categories (clinical information, vital signs, laboratory examination items, stool examination) as input, and Transformer is used as feature extractor. KAN as a classifier to predict the risk of rebleeding in patients with upper gastrointestinal bleeding. The model uses five-fold cross validation and calculates key metrics such as accuracy to evaluate its performance. In addition, the deep learning model was compared with a variety of machine learning methods (decision tree, random forest, logistic regression, K-nearest neighbor) and common clinical risk scores (Rockall score, AIMS65 score, Glasgow-Blatchford score) to verify its effectiveness and advantages. In order to highlight the importance of H. pylori infection degree to the model performance, we conducted a comparative experiment to observe the role of H. pylori infection degree in the model. In the correlation analysis between rebleeding and clinical data and related indicators, the risk of rebleeding in men (62.5%) was higher than that in women (43.47%), and the risk of rebleeding in patients with concurrent diseases (60.37%) was higher than that in patients without concurrent diseases. In the analysis of the correlation between the degree of infection and the laboratory test items, the hemoglobin level of patients will also change with the change of the degree of infection of patients (p < 0.05 in the above correlation analysis, all had statistical significance). The rebleeding detection rates of Rockall score, AIMS65 score and Glasgow Blatchford score were 16.14%, 0 and 77.17%, respectively. Of the four machine learning models, Random Forest (RF) had the highest accuracy on the test set at 0.68. The accuracy of the deep learning model on the verification set is the highest of 0.9750, and the accuracy of the test set is the highest of 0.9615. In addition, by exploring the influence of infection on the model prediction, it was found that the prediction accuracy of rebleeding in the non-H. pylori infection group (0.8989) was lower than that in the H. pylori infection group (0.9636), and other evaluation parameters were also lower than that in the infection group. In addition, by adding irrelevant random noise to mask the influence of infection degree on model output, it is found that the model prediction accuracy (0.7992) is significantly reduced. Based on the degree of H. pylori infection in patients with upper gastrointestinal bleeding, combined with a number of clinical laboratory tests and clinical data, we developed a clinical model for predicting the risk of rebleeding in patients with upper gastrointestinal bleeding. It provides an early prediction of rebleeding during a patient's hospitalization and optimizes early intervention for patients to a certain extent. It provides a more concise, convenient and effective guidance scheme for small and medium-sized hospitals to make clinical decisions for UGIB patients.