External Test Research Articles

Development of a Deep Learning Model for the Analysis of Dorsal Root Ganglion Chromatolysis in Rat Spinal Stenosis.

To create a deep learning (DL) model that can accurately detect and classify three distinct types of rat dorsal root ganglion neurons: normal, segmental chromatolysis, and central chromatolysis. The DL model has the potential to improve the efficiency and precision of neuron classification in research related to spinal injuries and diseases. H&E slide images were divided into an internal training set (80%) and a test set (20%). The training dataset was labeled by two pathologists using pre-defined grades. Using this dataset, a two-component DL model was developed with the first component being a convolutional neural network (CNN) that was trained to detect the region of interest (ROI) and the second component being another CNN used for classification. A total of 240 lumbar dorsal root ganglion (DRG) pathology slide images from rats were analyzed. The internal testing results showed an accuracy of 93.13%, and the external dataset testing demonstrated an accuracy of 93.44%. The DL model demonstrated a level of agreement comparable to that of pathologists in detecting and classifying normal and segmental chromatolysis neurons, although its agreement was slightly lower for central chromatolysis neurons. Significance: DL in improving the accuracy and efficiency of pathological analysis suggests that it may have a role in enhancing medical decision-making.

Development and validation of a nomogram for predicting lymph node metastasis in ductal carcinoma in situ with microinvasion: A SEER population-based study.

Ductal carcinoma in situ with microinvasion (DCIS-MI) is a special type of breast cancer. It is an invasive lesion less than 1.0 mm in size related to simple ductal carcinoma in situ (DCIS). Lymph node metastasis (LNM) in DCIS-MI often indicates a poor prognosis. Therefore, the management of lymph nodes plays a vital role in the treatment strategy of DCIS-MI. Since DCIS-MI is often diagnosed by postoperative paraffin section and immunohistochemical detection, to obtain the best clinical benefits for such patients, we aim to establish and verify a nomogram to predict the possibility of lymph node metastasis in DCIS-MI patients and help preoperative or intraoperative clinical decision-making. A retrospective analysis of patients with DCIS-MI in the Surveillance, Epidemiology, and End Results (SEER) database from 2010 to 2019 was performed. The study cohort was randomly divided into a training cohort and a validation cohort at a ratio of 7:3. The risk factors were determined by univariate and multivariate logistic regression analyses in the training cohort, and a nomogram was constructed. The receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA) were used to evaluate the nomogram in the training set and validation set. An independent data cohort was obtained from the Shanghai Jiao Tong University Breast Cancer Database (SJTU-BCDB) for external validation. This study included 3951 female patients from SEER with DCIS-MI, including 244 patients with regional lymph node metastasis, accounting for 6.18% of the total. An independent test set of 323 patients from SJTU-BCDB was used for external validation. According to the multifactorial logistic regression analysis results, age at diagnosis, ethnicity, grade, and surgical modality were included in the prediction model. The areas under the ROC curves (AUCs) were 0.739 (95% CI: 0.702~0.775), 0.732 (95% CI: 0.675~0.788), and 0.707 (95%CI: 0.607-0.807) in the training, validation and external test groups, suggesting that the column line graphs had excellent differentiation. The calibration curves slope was close to 1, and the model's predicted values were in good agreement with the actual values. The DCA curves showed good clinical utility. In this study, we constructed accurate and practical columnar maps with some clinical benefit to predict the likelihood of lymph node metastasis in patients with postoperatively diagnosed DCIS-MI and provide a reference value for specifying treatment strategies.

Predictors of Inadequate Serum Urate Response to Low-Dose Febuxostat in Male Patients with Gout.

This study aimed to understand predictors of inadequate response (IR) to low-dose febuxostat treatment based on clinical variables. We pooled data from 340 patients of an observational cohort and two clinical trials who received febuxostat 20 mg/day for at least 3 months. IR was defined as failure to reach the target serum urate level (sUA<6 mg/dL) at any time point during 3 months treatment. The potential predictors associated with short- or mid-term febuxostat IR after pooling the three cohorts were explored using mixed-effect logistic analysis. Machine learning models were performed to evaluate the predictors for IR using the pooled data as the discovery set and validated in an external test set. Of the 340 patients, 68.9% and 51.8% were non-responders to low-dose febuxostat during short- and mid-term follow-up, respectively. Serum urate and triglyceride (TG) levels were significantly associated with febuxostat IR, but were also selected as significant features by LASSO analysis combined with age, BMI, and C-reactive protein (CRP). These five features in combination, using the best-performing stochastic gradient descent classifier, achieved an area under the receiver operating characteristic curve of 0.873 (95% CI [0.763, 0.942]) and 0.706 (95% CI [0.636, 0.727]) in the internal and external test sets, respectively, to predict febuxostat IR. Response to low-dose febuxostat is associated with early sUA improvement in individual patients, as well as patient age, BMI, and levels of TG and CRP.

US-based Sequential Algorithm Integrating an AI Model for Advanced Liver Fibrosis Screening.

Background Noninvasive tests can be used to screen patients with chronic liver disease for advanced liver fibrosis; however, the use of single tests may not be adequate. Purpose To construct sequential clinical algorithms that include a US deep learning (DL) model and compare their ability to predict advanced liver fibrosis with that of other noninvasive tests. Materials and Methods This retrospective study included adult patients with a history of chronic liver disease or unexplained abnormal liver function test results who underwent B-mode US of the liver between January 2014 and September 2022 at three health care facilities. A US-based DL network (FIB-Net) was trained on US images to predict whether the shear-wave elastography (SWE) value was 8.7 kPa or higher, indicative of advanced fibrosis. In the internal and external test sets, a two-step algorithm (Two-step#1) using the Fibrosis-4 Index (FIB-4) followed by FIB-Net and a three-step algorithm (Three-step#1) using FIB-4 followed by FIB-Net and SWE were used to simulate screening scenarios where liver stiffness measurements were not or were available, respectively. Measures of diagnostic accuracy were calculated using liver biopsy as the reference standard and compared between FIB-4, SWE, FIB-Net, and European Association for the Study of the Liver guidelines (ie, FIB-4 followed by SWE), along with sequential algorithms. Results The training, validation, and test data sets included 3067 (median age, 42 years [IQR, 33-53 years]; 2083 male), 1599 (median age, 41 years [IQR, 33-51 years]; 1124 male), and 1228 (median age, 44 years [IQR, 33-55 years]; 741 male) patients, respectively. FIB-Net obtained a noninferior specificity with a margin of 5% (P < .001) compared with SWE (80% vs 82%). The Two-step#1 algorithm showed higher specificity and positive predictive value (PPV) than FIB-4 (specificity, 79% vs 57%; PPV, 44% vs 32%) while reducing unnecessary referrals by 42%. The Three-step#1 algorithm had higher specificity and PPV compared with European Association for the Study of the Liver guidelines (specificity, 94% vs 88%; PPV, 73% vs 64%) while reducing unnecessary referrals by 35%. Conclusion A sequential algorithm combining FIB-4 and a US DL model showed higher diagnostic accuracy and improved referral management for all-cause advanced liver fibrosis compared with FIB-4 or the DL model alone. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Ghosh in this issue.

Lower Extremity Growth according to AI Automated Femorotibial Length Measurement on Slot-Scanning Radiographs in Pediatric Patients.

Background Commonly used pediatric lower extremity growth standards are based on small, dated data sets. Artificial intelligence (AI) enables creation of updated growth standards. Purpose To train an AI model using standing slot-scanning radiographs in a racially diverse data set of pediatric patients to measure lower extremity length and to compare expected growth curves derived using AI measurements to those of the conventional Anderson-Green method. Materials and Methods This retrospective study included pediatric patients aged 0-21 years who underwent at least two slot-scanning radiographs in routine clinical care between August 2015 and February 2022. A Mask Region-based Convolutional Neural Network was trained to segment the femur and tibia on radiographs and measure total leg, femoral, and tibial length; accuracy was assessed with mean absolute error. AI measurements were used to create quantile polynomial regression femoral and tibial growth curves, which were compared with the growth curves of the Anderson-Green method for coverage based on the central 90% of the estimated growth distribution. Results In total, 1874 examinations in 523 patients (mean age, 12.7 years ± 2.8 [SD]; 349 female patients) were included; 40% of patients self-identified as White and not Hispanic or Latino, and the remaining 60% self-identified as belonging to a different racial or ethnic group. The AI measurement training, validation, and internal test sets included 114, 25, and 64 examinations, respectively. The mean absolute errors of AI measurements of the femur, tibia, and lower extremity in the test data set were 0.25, 0.27, and 0.33 cm, respectively. All 1874 examinations were used to generate growth curves. AI growth curves more accurately represented lower extremity growth in an external test set (n = 154 examinations) than the Anderson-Green method (90% coverage probability: 86.7% [95% CI: 82.9, 90.5] for AI model vs 73.4% [95% CI: 68.4, 78.3] for Anderson-Green method; χ2 test, P < .001). Conclusion Lower extremity growth curves derived from AI measurements on standing slot-scanning radiographs from a diverse pediatric data set enabled more accurate prediction of pediatric growth. © RSNA, 2024 Supplemental material is available for this article.

Hologram quantitative structure-activity relationship on the gas chromatographic retention index of plant essential oil constituents

The gas chromatography retention index (RI) is an important parameter for the identification of different types of compounds in the field of chromatographic analysis; however, the experimental collection of RI values is a extremely cumbersome process. Thus, there is an urgent need for the establishment of a simple, efficient, and accurate model for the prediction of the RI values of compounds. In this study, first, the experimental RI values for 60 plant essential oil constituents were obtained. Next, a model describing the hologram quantitative structure-activity relationship (HQSAR) between the structural properties of the essential oil constituents and their RI values was investigated and constructed. The optimal HQSAR model was established by setting the model parameters "fragment size", "fragment distinction", "hologram length" and "principal components" to "1-4", "C, Ch", "199", and "4", respectively. Finally, the predictive ability of the model was verified using external test set validation and leave-one-out cross-validation (LOO-CV). The experimental results were as follows, the root mean square error of prediction (RMSEP), predictive determination coefficient ([Formula: see text]), concordance correlation coefficient (CCC), and mean relative error (MRE) for external test set validation were 40.45, 0.984, 0.968, and 2.20%, respectively. Meanwhile, the root mean square error of cross validation (RMSECV) and MRE for LOO-CV were 72.56 and 4.17%, respectively. These findings demonstrate that the established HQSAR model has a good predictive ability and can accurately predict the RI values of plant essential oil constituents. In addition, the molecular contribution maps of the HQSAR model revealed that the RI values of aromatic compounds increase when hydroxyl groups are connected to their alkyl chains. Aliphatic compounds feature long chain alkyl groups, which can lead to an increase in RI values. The above phenomena highlight the promising application prospects of HQSAR for studying the RI values of plant essential oil constituents. Therefore, this study provides a reliable theoretical basis for predicting the RI values of other essential oil constituents.

Development and evaluation of a deep learning-based model for simultaneous detection and localization of rib and clavicle fractures in trauma patients’ chest radiographs

PurposeTo develop a rib and clavicle fracture detection model for chest radiographs in trauma patients using a deep learning (DL) algorithm.Materials and methodsWe retrospectively collected 56 145 chest X-rays (CXRs)...

Novel peripheral blood mononuclear cell mRNA signature for IFN-beta therapy responsiveness prediction in multiple sclerosis

Interferon-beta (IFN- β ) is one of the classical drugs for immunomodulatory therapy in relapsing-remitting multiple sclerosis (RRMS) patients, but the drug responsiveness of different patients varies. Currently, there is no valid model to predict IFN- β responsiveness. This research attempted to develop an IFN- β responsiveness prediction model based on mRNA expression in RRMS patient peripheral blood mononuclear cells. Peripheral blood mononuclear cell mRNA expression datasets including 50 RRMS patients receiving IFN- β treatment were obtained from GEO. Among the datasets, 24 cases from GSE24427 were included in a training set, and 18 and 9 cases from GSE19285 and GSE33464, respectively, were adopted as two independent test sets. In the training set, blood samples were collected immediately before first, second, month 1, 12, and 24 IFN- β injection, and the mRNA expression data at four time points, namely, two days, one month, one year and two years after the onset of IFN- β treatment, were compared with pre-treatment data to identify IFN-stimulated genes (ISGs). The ISGs at the one-month time point were used to construct the drug responsiveness prediction model. Next, the drug responsiveness model was verified in the two independent test sets to examine the performance of the model in predicting drug responsiveness. Finally, we used CIBERSORTx to estimate the content of cell subtypes in samples and evaluated whether differences in the proportions of cell subtypes were related to differences in IFN- β responsiveness. Among the four time points, one month was the time point when the training set GSE24427 and test set GSE33464 had the highest number of ISGs. Functional analysis showed that these one-month ISGs were enriched in biological functions such as the innate immune response, type-I interferon signalling pathway, and other IFN- β -associated functions. Based on these ISGs, we obtained a four-factor prediction model for IFN- β responsiveness including MX1, MX2, XAF1, and LAMP3. In addition, the model demonstrated favourable predictive performance within the training set and two external test sets. A higher proportion of activated NK cells and lower naive CD4/total CD4 ratio might indicate better drug responsiveness. This research developed a polygene-based biomarker model that could predict RRMS patient IFN- β responsiveness in the early treatment period. This model could probably help doctors screen out patients who would not benefit from IFN- β treatment early and determine whether a current treatment plan should be continued.

SurvConvMixer: robust and interpretable cancer survival prediction based on ConvMixer using pathway-level gene expression images

Cancer is one of the leading causes of deaths worldwide. Survival analysis and prediction of cancer patients is of great significance for their precision medicine. The robustness and interpretability of the survival prediction models are important, where robustness tells whether a model has learned the knowledge, and interpretability means if a model can show human what it has learned. In this paper, we propose a robust and interpretable model SurvConvMixer, which uses pathways customized gene expression images and ConvMixer for cancer short-term, mid-term and long-term overall survival prediction. With ConvMixer, the representation of each pathway can be learned respectively. We show the robustness of our model by testing the trained model on absolutely untrained external datasets. The interpretability of SurvConvMixer depends on gradient-weighted class activation mapping (Grad-Cam), by which we can obtain the pathway-level activation heat map. Then wilcoxon rank-sum tests are conducted to obtain the statistically significant pathways, thereby revealing which pathways the model focuses on more. SurvConvMixer achieves remarkable performance on the short-term, mid-term and long-term overall survival of lung adenocarcinoma, lung squamous cell carcinoma and skin cutaneous melanoma, and the external validation tests show that SurvConvMixer can generalize to external datasets so that it is robust. Finally, we investigate the activation maps generated by Grad-Cam, after wilcoxon rank-sum test and Kaplan–Meier estimation, we find that some survival-related pathways play important role in SurvConvMixer.

Development and validation of an interpretable model integrating multimodal information for improving ovarian cancer diagnosis

Ovarian cancer, a group of heterogeneous diseases, presents with extensive characteristics with the highest mortality among gynecological malignancies. Accurate and early diagnosis of ovarian cancer is of great significance. Here, we present OvcaFinder, an interpretable model constructed from ultrasound images-based deep learning (DL) predictions, Ovarian–Adnexal Reporting and Data System scores from radiologists, and routine clinical variables. OvcaFinder outperforms the clinical model and the DL model with area under the curves (AUCs) of 0.978, and 0.947 in the internal and external test datasets, respectively. OvcaFinder assistance led to improved AUCs of radiologists and inter-reader agreement. The average AUCs were improved from 0.927 to 0.977 and from 0.904 to 0.941, and the false positive rates were decreased by 13.4% and 8.3% in the internal and external test datasets, respectively. This highlights the potential of OvcaFinder to improve the diagnostic accuracy, and consistency of radiologists in identifying ovarian cancer.

Screening and staging of chronic obstructive pulmonary disease with deep learning based on chest X-ray images and clinical parameters

BackgroundChronic obstructive pulmonary disease (COPD) is underdiagnosed with the current gold standard measure pulmonary function test (PFT). A more sensitive and simple option for early detection and severity evaluation of COPD could benefit practitioners and patients.MethodsIn this multicenter retrospective study, frontal chest X-ray (CXR) images and related clinical information of 1055 participants were collected and processed. Different deep learning algorithms and transfer learning models were trained to classify COPD based on clinical data and CXR images from 666 subjects, and validated in internal test set based on 284 participants. External test including 105 participants was also performed to verify the generalization ability of the learning algorithms in diagnosing COPD. Meanwhile, the model was further used to evaluate disease severity of COPD by predicting different grads.ResultsThe Ensemble model showed an AUC of 0.969 in distinguishing COPD by simultaneously extracting fusion features of clinical parameters and CXR images in internal test, better than models that used clinical parameters (AUC = 0.963) or images (AUC = 0.946) only. For the external test set, the AUC slightly declined to 0.934 in predicting COPD based on clinical parameters and CXR images. When applying the Ensemble model to determine disease severity of COPD, the AUC reached 0.894 for three-classification and 0.852 for five-classification respectively.ConclusionThe present study used DL algorithms to screen COPD and predict disease severity based on CXR imaging and clinical parameters. The models showed good performance and the approach might be an effective case-finding tool with low radiation dose for COPD diagnosis and staging.

Automatic segmentation of fat metaplasia on sacroiliac joint MRI using deep learning

ObjectiveTo develop a deep learning (DL) model for segmenting fat metaplasia (FM) on sacroiliac joint (SIJ) MRI and further develop a DL model for classifying axial spondyloarthritis (axSpA) and non-axSpA.Materials and methodsThis study retrospectively collected 706 patients with FM who underwent SIJ MRI from center 1 (462 axSpA and 186 non-axSpA) and center 2 (37 axSpA and 21 non-axSpA). Patients from center 1 were divided into the training, validation, and internal test sets (n = 455, 64, and 129). Patients from center 2 were used as the external test set. We developed a UNet-based model to segment FM. Based on segmentation results, a classification model was built to distinguish axSpA and non-axSpA. Dice Similarity Coefficients (DSC) and area under the curve (AUC) were used for model evaluation. Radiologists’ performance without and with model assistance was compared to assess the clinical utility of the models.ResultsOur segmentation model achieved satisfactory DSC of 81.86% ± 1.55% and 85.44% ± 6.09% on the internal cross-validation and external test sets. The classification model yielded AUCs of 0.876 (95% CI: 0.811–0.942) and 0.799 (95% CI: 0.696–0.902) on the internal and external test sets, respectively. With model assistance, segmentation performance was improved for the radiological resident (DSC, 75.70% vs. 82.87%, p < 0.05) and expert radiologist (DSC, 85.03% vs. 85.74%, p > 0.05).ConclusionsDL is a novel method for automatic and accurate segmentation of FM on SIJ MRI and can effectively increase radiologist’s performance, which might assist in improving diagnosis and progression of axSpA.Critical relevance statementDL models allowed automatic and accurate segmentation of FM on sacroiliac joint MRI, which might facilitate quantitative analysis of FM and have the potential to improve diagnosis and prognosis of axSpA.Key points• Deep learning was used for automatic segmentation of fat metaplasia on MRI.• UNet-based models achieved automatic and accurate segmentation of fat metaplasia.• Automatic segmentation facilitates quantitative analysis of fat metaplasia to improve diagnosis and prognosis of axial spondyloarthritis.Graphical

Federated learning with knowledge distillation for multi-organ segmentation with partially labeled datasets

The state-of-the-art multi-organ CT segmentation relies on deep learning models, which only generalize when trained on large samples of carefully curated data. However, it is challenging to train a single model that can segment all organs and types of tumors since most large datasets are partially labeled or are acquired across multiple institutes that may differ in their acquisitions. A possible solution is Federated learning, which is often used to train models on multi-institutional datasets where the data is not shared across sites. However, predictions of federated learning can be unreliable after the model is locally updated at sites due to ‘catastrophic forgetting’. Here, we address this issue by using knowledge distillation (KD) so that the local training is regularized with the knowledge of a global model and pre-trained organ-specific segmentation models. We implement the models in a multi-head U-Net architecture that learns a shared embedding space for different organ segmentation, thereby obtaining multi-organ predictions without repeated processes. We evaluate the proposed method using 8 publicly available abdominal CT datasets of 7 different organs. Of those datasets, 889 CTs were used for training, 233 for internal testing, and 30 volumes for external testing. Experimental results verified that our proposed method substantially outperforms other state-of-the-art methods in terms of accuracy, inference time, and the number of parameters.

Heart disease prediction utilizing machine learning techniques

Heart disease is an international public health issue and a significant health risk for many people. The World Health Organization (WHO) reported that heart disease has been identified as one of the primary causes of death. Owing to its rapid development and application in many areas, Machine Learning has become an effective technique for predicting heart disease. Machine learning, along with large-scale medical data and advanced algorithms, can assist healthcare professionals in accurately predicting the risk of heart disease, thus providing early intervention and treatment for patients. This research paper uses the “heart_2020_cleaned.csv” dataset, containing 319795 instances and 18 attributes, of which 70% of instances were randomly selected for the training set and 30% for testing. Applying machine learning algorithms in data mining such as Decision Tree (DT), LightGBM, Random Forest (RF) and Logistic Regression (LR) to forecast heart disease. Before constructing models, data cleaning, feature selection and hyperparameter tuning processes were done, aiming to explore the potential patterns among data. Comparative Analysis was conducted on the external test set to compare the prediction performance of different models at the same level. The result reported that the highest accuracy achieved with LightGBM was 76.9%, followed by Logistic Regression and Random Forest, with Decision Tree being the worst.

A Deep Learning Based on Automatic Cerebral Aneurysm Detection in Brain Computed Tomography Angiography Scan Images

Cerebral aneurysms are an important disease that threatens human life. Rupture of these aneurysms causes hemorrhages in the cerebral arteries. Computed Tomography Angiography (CTA) is widely used in the clinical diagnosis of cerebral aneurysms. Interpretation errors by radiologists in examining these Computed Tomography (CT) images are vital for patients. Based on this importance, deep learning-based studies aim to help keep these errors to a minimum. For this purpose, CTA images were used to detect cerebral aneurysms in this study. For CTA image analysis, deep learning methodology was preferred through Convolutional Neural Network (CNN). The validation accuracy of the training obtained as a result of deep learningg has a high rate of validation with 99.54% accuracy, 100% sensitivity, 98.89% specificity and 100% precision. As a training dataset, it yielded 127 true positives and 1 false positive for patient images with aneurysm, 89 true negatives and 0 false negative for images of patients with non-aneurysms. In this trained network, results were obtained with a high accuracy of 86.6% on 75 CTA images for external test. Regional dimension analysis was also performed for an image with an aneurysm detected in the test process.

Graph to Activation Energy Models Easily Reach Irreducible Errors but Show Limited Transferability.

Activation energy characterization of competing reactions is a costly but crucial step for understanding the kinetic relevance of distinct reaction pathways, product yields, and myriad other properties of reacting systems. The standard methodology for activation energy characterization has historically been a transition state search using the highest level of theory that can be afforded. However, recently, several groups have popularized the idea of predicting activation energies directly based on nothing more than the reactant and product graphs, a sufficiently complex neural network, and a broad enough data set. Here, we have revisited this task using the recently developed Reaction Graph Depth 1 (RGD1) transition state data set and several newly developed graph attention architectures. All of these new architectures achieve similar state-of-the-art results of ∼4 kcal/mol mean absolute error on withheld testing sets of reactions but poor performance on external testing sets composed of reactions with differing mechanisms, reaction molecularity, or reactant size distribution. Limited transferability is also shown to be shared by other contemporary graph to activation energy architectures through a series of case studies. We conclude that an array of standard graph architectures can already achieve results comparable to the irreducible error of available reaction data sets but that out-of-distribution performance remains poor.

Abstract 4796: Improving cancer screening performance for women with dense breast tissue

Abstract Breast cancer is the most common cancer in women in the United States. To reduce mortality, screening by mammography in females is recommended by the United States Preventive Services Task Force, successfully reducing mortality by 15 - 32% in women aged 40 - 69 years. This is partially due to the high compliance rate (60 - 80%) among the candidate population. Mammography increases early-stage disease detection, which is correlated with better outcomes. While mammography has been largely successful, this screening method has limitations. Overall, mammography has false-negative and false-positive rates estimated at 12 - 20% and 7 - 12%, respectively. Consequently, annual mammograms result in 50 - 60% of women having a false-positive result within a 10-year span. Additionally, high breast density, found in approximately 50% of females over the age of 40, increases susceptibility to false positive and negative results. For this reason, in 2023 the FDA required that dense breast tissue status, an independent risk factor for cancer, be disclosed to mammogram recipients. The subsequent diagnostic pathway for women with dense breast tissue is not well defined for the care provider. Alternative imaging options have performance issues or are less accessible. Here we describe a liquid biopsy test to complement mammography in women with dense breast tissue to help resolve ambiguity in results. Blood from early- and late-stage breast cancer patients was prospectively collected in Streck cfDNA BCT Devices (n=100) or retrospectively acquired (n=100); and blood from presumed normal samples was collected (n=200). cfDNA was extracted and converted into libraries for low-pass whole genome sequencing (LP-WGS). Sequenced reads were analyzed to generate fragment end-motif and size (FEMS) and fragment coverage dataframes for classification using Genece Health’s proprietary analysis pipeline and machine-learning (ML) algorithm. 150 breast cancer samples distributed across stages and 150 presumed normal samples were selected to create a training cohort. Performance was tested using a 5-fold cross validation strategy. The best performance was observed using a convolution neural network (CNN) ML model with &amp;gt;80% sensitivity when specificity was set at 85%. The remaining 50 breast cancer samples and 50 presumed normal samples were used as an external test set. The performance in this cohort was comparable to the training cohort, with both specificity and sensitivity &amp;gt;80%. Genece presents an early proof of concept breast cancer screening liquid biopsy assay that uses LP-WGS fragmentomics and ML to detect cancer from a single blood sample. Continuing development with additional cohorts and optimization will yield even greater performance. The Genece liquid biopsy test has the potential to be paired with mammography, especially in women with dense breast tissue, to improve screening sensitivity, specificity, and overall outcomes. Citation Format: Molly Smith, Mengchi Wang, Andrew Carson, Byung In Lee, Michael Salmans, Bryan Leatham. Improving cancer screening performance for women with dense breast tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4796.

Abstract 886: Integrated deep learning model for predicting DNA methylation and tumor types from histopathology in central nervous system tumors

Abstract Precision in diagnosis of diverse central nervous system (CNS) tumor types is crucial for optimal patient treatment. DNA methylation profiles, which capture the methylation status of thousands of individual CpG sites, are data-driven means to enhance diagnostic accuracy, but this technique is expensive, time-consuming, and not yet routinely available. To address this, we developed DEPLOY, a deep-learning model that predicts 10 major categories of CNS tumors from histopathology. DEPLOY integrates three distinct components: the first classifies CNS tumors directly from histopathology slide images (‘direct model’); the second initially generates predictions for DNA methylation beta values, which are subsequently used for tumor classification (‘indirect model’); and the third classifies tumor types directly from routinely available patient demographics. First, we find that DEPLOY accurately predicted beta values from histopathology images, suggesting that application of inferred methylation data is a promising approach for deep learning histopathology-based classifiers. Using a 10-class model trained on an internal dataset, we apply this model on two independent external test datasets of 1,522 and 348 cases, achieving top-1 accuracies of 96% and 94%, respectively on samples which are predicted with high confidence. Taken together, DEPLOY could assist pathologists in diagnosing CNS tumors, in an equitable very low-cost manner, within a clinically relevant short time frame. Citation Format: Eldad David Shulman, Danh-Tai Hoang, Rust Turakulov, Zied Abdullaev, Omkar Singh, Emma M. Campagnolo, Eric A. Stone, MacLean P. Nasrallah, Eytan Ruppin, Kenneth Aldape. Integrated deep learning model for predicting DNA methylation and tumor types from histopathology in central nervous system tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 886.

Lightweight preprocessing and template matching facilitate streamlined ischemic myocardial scar classification.

Ischemic myocardial scarring (IMS) is a common outcome of coronary artery disease that potentially leads to lethal arrythmias and heart failure. Late-gadolinium-enhanced cardiac magnetic resonance (CMR) imaging scans have served as the diagnostic bedrock for IMS, with recent advancements in machine learning enabling enhanced scar classification. However, the trade-off for these improvements is intensive computational and time demands. As a solution, we propose a combination of lightweight preprocessing (LWP) and template matching (TM) to streamline IMS classification. CMR images from 279 patients (151 IMS, 128 control) were classified for IMS presence using two convolutional neural networks (CNNs) and TM, both with and without LWP. Evaluation metrics included accuracy, sensitivity, specificity, F1-score, area under the receiver operating characteristic curve (AUROC), and processing time. External testing dataset analysis encompassed patient-level classifications (PLCs) and a CNN versus TM classification comparison (CVTCC). LWP enhanced the speed of both CNNs (4.9x) and TM (21.9x). Furthermore, in the absence of LWP, TM outpaced CNNs by over 10x, while with LWP, TM was more than 100x faster. Additionally, TM performed similarly to the CNNs in accuracy, sensitivity, specificity, F1-score, and AUROC, with PLCs demonstrating improvements across all five metrics. Moreover, the CVTCC revealed a substantial 90.9% agreement. Our results highlight the effectiveness of LWP and TM in streamlining IMS classification. Anticipated enhancements to LWP's region of interest (ROI) isolation and TM's ROI targeting are expected to boost accuracy, positioning them as a potential alternative to CNNs for IMS classification, supporting the need for further research.

Semi-supervised learning towards automated segmentation of PET images with limited annotations: application to lymphoma patients.

Manual segmentation poses a time-consuming challenge for disease quantification, therapy evaluation, treatment planning, and outcome prediction. Convolutional neural networks (CNNs) hold promise in accurately identifying tumor locations and boundaries in PET scans. However, a major hurdle is the extensive amount of supervised and annotated data necessary for training. To overcome this limitation, this study explores semi-supervised approaches utilizing unlabeled data, specifically focusing on PET images of diffuse large B-cell lymphoma (DLBCL) and primary mediastinal large B-cell lymphoma (PMBCL) obtained from two centers. We considered 2-[18F]FDG PET images of 292 patients PMBCL (n = 104) and DLBCL (n = 188) (n = 232 for training and validation, and n = 60 for external testing). We harnessed classical wisdom embedded in traditional segmentation methods, such as the fuzzy clustering loss function (FCM), to tailor the training strategy for a 3D U-Net model, incorporating both supervised and unsupervised learning approaches. Various supervision levels were explored, including fully supervised methods with labeled FCM and unified focal/Dice loss, unsupervised methods with robust FCM (RFCM) and Mumford-Shah (MS) loss, and semi-supervised methods combining FCM with supervised Dice loss (MS + Dice) or labeled FCM (RFCM + FCM). The unified loss function yielded higher Dice scores (0.73 ± 0.11; 95% CI 0.67-0.8) than Dice loss (p value < 0.01). Among the semi-supervised approaches, RFCM + αFCM (α = 0.3) showed the best performance, with Dice score of 0.68 ± 0.10 (95% CI 0.45-0.77), outperforming MS + αDice for any supervision level (any α) (p < 0.01). Another semi-supervised approach with MS + αDice (α = 0.2) achieved Dice score of 0.59 ± 0.09 (95% CI 0.44-0.76) surpassing other supervision levels (p < 0.01). Given the time-consuming nature of manual delineations and the inconsistencies they may introduce, semi-supervised approaches hold promise for automating medical imaging segmentation workflows.