Cancer Detection Model Research Articles

Stochastic Dilated Residual Ghost Model for Breast Cancer Detection.

One of the most contentious issues in modern medicine is how to effectively standardise breast cancer screening. Deep learning models are already saving lives in the medical field due to their capacity to distinguish between benign and malignant tumours. Histopathology imaging poses difficulties due to the possibility of large colour variations caused by the staining technique and the biopsy material used to make the image; this problem leads to inaccurate breast cancer diagnoses. Our primary focus in this assessment is on the four main research concerns listed in the following: Overfitting and colour divergence must be rectified before moving on to other aspects of breast cancer categorisation. To overcome this issue, strain normalisation is utilised, and adding extra components is used to cope with overfitting; both techniques yielded positive results. The multiscale stochastic and dilation unit was then created to extract and enhance fine-grained characteristics such as edges, contours, and colour accuracy. To achieve this, the image is scaled to various different levels. The last challenge is to overcome the stochastic dilated residual ghost model's unreliability when used to recognise very tiny objects. The stochastic pooling block in this model makes effective use of downsampling to simplify the process without compromising the capacity to retrieve deep information. This upgrade was done as part of a bigger endeavour to eliminate unneeded or redundant components. In this case, we use convolution and identity mapping to create and maintain accurate mappings of the object's inherent characteristics. Upsampling is frequently used in conjunction with stochastic pooling to reduce feature dimensionality. The results of the experiments show that the suggested method is better than some of the current methods, with a network performance measurement area under the curve of 96.15 and percentages of 98.50 and 97.36.

Optimized Stacking Ensemble Learning Model for Breast Cancer Detection and Classification Using Machine Learning

Breast cancer is the most frequently encountered medical hazard for women in their forties, affecting one in every eight women. It is the greatest cause of death worldwide, and early detection and diagnosis of the disease are extremely challenging. Breast cancer currently exceeds all other female cancers, including ovarian cancer. Researchers can use access to healthcare records to find previously unknown healthcare trends. According to the National Cancer Institute (NCI), breast cancer mortality rates can be lowered if the disease is detected early. The novelty of our work is to develop an optimized stacking ensemble learning (OSEL) model capable of early breast cancer prediction. A dataset from the University of California, Irvine repository was used, and comparisons to modern classifier models were undertaken. The implementation analyses reveal the unique approach’s efficacy and superiority when compared to existing contemporary categorization models (AdaBoostM1, gradient boosting, stochastic gradient boosting, CatBoost, and XGBoost). In every classification task, predictive models may be used to predict the class level, and the current research explores a range of predictive models. It is better to integrate multiple classification algorithms to generate a set of prediction models capable of predicting each class level with 91–99% accuracy. On the breast cancer Wisconsin dataset, the suggested OSEL model attained a maximum accuracy of 99.45%, much higher than any single classifier. Thus, the study helps healthcare professionals find breast cancer and prevent it from happening.

Breast cancer detection model using fuzzy entropy segmentation and ensemble classification

According to studies, breast cancer is the deadliest disease and the main cause of the elevated mortality rates among women. The main method for detecting breast cancer is mammography. Even today, using a mammography image to make earlier detection of breast cancer remains a complex undertaking. In this work, the proposed system includes the following processes: “(i) pre-processing (ii) segmentation (iii) feature extraction (iv) optimal feature selection and (v) classification”. Initially, RGB-Grey scale conversion is done during pre-processing, and segmentation is carried out using the proposed fuzzy entropy segmentation model. Then the features including “fractal features like lacunarity, fractal dimension and texture features such as Grey Level Co-occurrence Matrix (GLCM) and Proposed Local Binary Patterns (LBP)” are determined. Then, in the prediction phase, an optimized ensemble classifier is introduced. The proposed ensemble classifier is constructed by amalgamating the “Random Forest (RF), Support Vector Machine (SVM) Neural Network (NN) and Optimized CNN” respectively. The extracted features are subjected to SVM, NN and RF, simultaneously. Here, optimized CNN determines the final results. In particular, Self Improved Cat Swarm Optimization (SI-CSO) is used to fine-tune CNN's weights. Finally, many metrics demonstrate the proposed model's efficiency.

Optimization driven model and segmentation network for skin cancer detection

The cancers in skins are considered as most risky disease. Recently, occurrence of skin cancer is considerably noticed among people globally. Earlier detection of skin cancer can result in reduced death rate. Dermoscopy is an effective way to detect and classify skin cancer. Since the visual assessment of dermoscopic images is a dull and cumbersome process, automated tools using computer aided diagnosis (CAD) model becomes essential. The current advances in machine learning (ML) such as deep learning (DL) has been considerably developed in the healthcare sector. The contemporary technical designs and techniques make it liable to discover this type of cancer effectually, but automated localization and skin lesion segmentation at prior phases is a complex task because of less contrast. This paper present an optimization based model to discover skin cancer using set of images. At primary, the input image is generated from a database and it is pre-processed using Gaussian filter and Region of Interest (ROI) extraction, which removes noises and mines interesting regions. The segmentation is performed with proposed U-RP-Net. Here, the proposed U-RP-Net model is obtained by integrating U-Net and RP-Net. Meanwhile, the output obtained from both U-Net and RP-Net model is integrated using Jaccard similarity-based fusion model. The data augmentation is processed for improving the detection performance. Finally, skin cancer finding is done using SqueezeNet. Moreover, the SqueezeNet is trained by proposed Aquila Whale Optimization (AWO) algorithm. The developed AWO approach is newly devised by integrating Aquila Optimizer (AO) and Whale Optimization Algorithm (WOA). The developed AWO-based SqueezeNet outperformed by highest testing accuracy of 92.5%, sensitivity of 92.1% and specificity of 91.7%.

Aquila Optimizer with Bayesian Neural Network for Breast Cancer Detection on Ultrasound Images

Breast cancer is the second most dominant kind of cancer among women. Breast Ultrasound images (BUI) are commonly employed for the detection and classification of abnormalities that exist in the breast. The ultrasound images are necessary to develop artificial intelligence (AI) enabled diagnostic support technologies. For improving the detection performance, Computer Aided Diagnosis (CAD) models are useful for breast cancer detection and classification. The current advancement of the deep learning (DL) model enables the detection and classification of breast cancer with the use of biomedical images. With this motivation, this article presents an Aquila Optimizer with Bayesian Neural Network for Breast Cancer Detection (AOBNN-BDNN) model on BUI. The presented AOBNN-BDNN model follows a series of processes to detect and classify breast cancer on BUI. To accomplish this, the AOBNN-BDNN model initially employs Wiener filtering (WF) related noise removal and U-Net segmentation as a pre-processing step. Besides, the SqueezeNet model derives a collection of feature vectors from the pre-processed image. Next, the BNN algorithm will be utilized to allocate appropriate class labels to the input images. Finally, the AO technique was exploited to fine-tune the parameters related to the BNN method so that the classification performance is improved. To validate the enhanced performance of the AOBNN-BDNN method, a wide experimental study is executed on benchmark datasets. A wide-ranging experimental analysis specified the enhancements of the AOBNN-BDNN method in recent techniques.

An Efficient Deep Learning Approach for Colon Cancer Detection

Colon cancer is the second most common cause of cancer death in women and the third most common cause of cancer death in men. Therefore, early detection of this cancer can lead to lower infection and death rates. In this research, we propose a new lightweight deep learning approach based on a Convolutional Neural Network (CNN) for efficient colon cancer detection. In our method, the input histopathological images are normalized before feeding them into our CNN model, and then colon cancer detection is performed. The efficiency of the proposed system is analyzed with publicly available histopathological images database and compared with the state-of-the-art existing methods for colon cancer detection. The result analysis demonstrates that the proposed deep model for colon cancer detection provides a higher accuracy of 99.50%, which is considered the best accuracy compared with the majority of other deep learning approaches. Because of this high result, the proposed approach is computationally efficient.

Optimal deep neural network-driven computer aided diagnosis model for skin cancer

Image-guided intervention is a medical procedure that leverages computerized systems to deliver virtual image overlays to help physicians in visualization and targeting the surgical site in an accurate manner. Computer Aided Diagnosis (CAD) models that use Deep Learning (DL) techniques are useful in achieving accurate skin cancer classification. In this background, the current research paper concentrates on the design of Optimal Deep Neural Network Driven Computer Aided Diagnosis Model for Skin Cancer Detection and Classification (ODNNCADSCC) model. The presented ODNNCADSCC model primarily applies Wiener Filtering (WF)-based pre-processing step followed by U-Net segmentation approach. In addition, SqueezeNet model is also exploited to generate a collection of feature vectors. Finally, Improved Whale Optimization Algorithm (IWOA) with DNN model is utilized for effectual skin cancer detection and classification. In this procedure, IWOA is applied to select the DNN parameters in a proficient manner. The comparative analysis results established the promising performance of the proposed ODNNCADSCC model over recent approaches with a maximum accuracy of 99.90%.

Detection and segmentation of melanoma skin cancer in dermoscopy images using modified Alexnet convolutional neural network‐morphological methodology

SummaryMelanoma is most vicious and dangerous type of skin cancer that its early detection can save patients' lives. Computer‐aided methods can be used for this early detection with acceptable performance. In this article, the melanoma skin cancer dermoscopy images are detected and cancer regions are segmented using the proposed modified AlexNet convolutional neural network‐morphological segmentation (CNN‐MS) method. This proposed work consists of enhancement stage, feature matrix construction stage, CNN classification stage, and segmentation stage. In enhancement stage, the skin pixels are contrast enhanced by histogram equalization technique and then non‐sub sampled contourlet transform (NSCT) is applied on the enhanced skin image to obtain the decomposed coefficients. Further, local ternary pattern (LTP) features are computed from the enhanced image and the feature matrix (FM) is constructed by integrating the NSCT coefficients and the LTP features. This constructed FM is fed to the CNN model for the classification of Melanoma skin cancer images from the normal skin images. Finally, morphological functions are applied on the classified melanoma skin image to segment the cancer regions. The performance of this developed model for skin cancer detection is analyzed on international skin imaging collaboration (ISIC) dataset images and achieves 98.11% of Se, 98.5% of Sp, 99.11% of Acc, and 0.96 of MCC. Also, the performance of this developed model for skin cancer detection is analyzed on SIIM‐ISIC dataset images and achieves 98.04% Se, 98.46% Sp, 99.12% Acc, and 0.97 MCC.

Transfer Learning-Based Multi-Scale Denoising Convolutional Neural Network for Prostate Cancer Detection.

Simple SummaryTo enhance the automatic diagnosis of the prostate cancer using machine learning algorithm, we modify the design of convolutional neural network to support multi-scale denoising of cancer images. Transfer learning is employed to leverage the detection accuracy of the prostate cancer detection model by taking advantages from more unseen data from a source dataset. Compared to existing methodologies, our work improves the accuracy by more than 10%. Ablation studies have conducted to evaluate the contributions of the components of the proposed algorithm, with 2.80%, 3.30%, and 3.13% for image denoising, multi-scale scheme, and transfer learning, respectively. The results reveal the effectiveness of the algorithm and provide insights for five future research directions.Background: Prostate cancer is the 4th most common type of cancer. To reduce the workload of medical personnel in the medical diagnosis of prostate cancer and increase the diagnostic accuracy in noisy images, a deep learning model is desired for prostate cancer detection. Methods: A multi-scale denoising convolutional neural network (MSDCNN) model was designed for prostate cancer detection (PCD) that is capable of noise suppression in images. The model was further optimized by transfer learning, which contributes domain knowledge from the same domain (prostate cancer data) but heterogeneous datasets. Particularly, Gaussian noise was introduced in the source datasets before knowledge transfer to the target dataset. Results: Four benchmark datasets were chosen as representative prostate cancer datasets. Ablation study and performance comparison between the proposed work and existing works were performed. Our model improved the accuracy by more than 10% compared with the existing works. Ablation studies also showed average improvements in accuracy using denoising, multi-scale scheme, and transfer learning, by 2.80%, 3.30%, and 3.13%, respectively. Conclusions: The performance evaluation and comparison of the proposed model confirm the importance and benefits of image noise suppression and transfer of knowledge from heterogeneous datasets of the same domain.

SAFNet: A deep spatial attention network with classifier fusion for breast cancer detection

Breast cancer is a top dangerous killer for women. An accurate early diagnosis of breast cancer is the primary step for treatment. A novel breast cancer detection model called SAFNet is proposed based on ultrasound images and deep learning. We employ a pre-trained ResNet-18 embedded with the spatial attention mechanism as the backbone model. Three randomized network models are trained for prediction in the SAFNet, which are fused by majority voting to produce more accurate results. A public ultrasound image dataset is utilized to evaluate the generalization ability of our SAFNet using 5-fold cross-validation. The simulation experiments reveal that the SAFNet can produce higher classification results compared with four existing breast cancer classification methods. Therefore, our SAFNet is an accurate tool to detect breast cancer that can be applied in clinical diagnosis.

An Improved VGG Model for Skin Cancer Detection

An Improved VGG Model for Skin Cancer Detection

Dilated residual grooming kernel model for breast cancer detection

• Solves the overfitting issue in the Convolution Neural Networks . • The proposed fully connected dilation unit enhances small and low-level features. • Accurate cancerous cell Detection using dilated model. • The proposed model solves the issue of desegregation. • Efficiently removal of redundant and similar features through proposed model. Breast cancer is the most significant cause of mortality among women. When detected and treated early, it saves lives. Breast cancer detection is becoming more accessible and accurate thanks to machine learning and deep learning models. This research aims to enhance medical science and technology by employing a deep learning model to detect small cancer cells with pinpoint accuracy. The proposed model uses datasets from the Breast Cancer Histopathological Image Classification (BreakHis) and Breast Cancer Histopathological Annotation and Diagnosis (BreCaHAD). Next, the image processing procedure employs strain normalization to rectify color divergence caused by using different slide scanners, staining processes, and biopsy materials. The data augmentation with nineteen parameters such as scaling, rotation, flip, resize, and gamma value tackles the overfitting problems. The augmented images are then processed using the dilated residual (DR) model; the DR model combines the proposed dilated spatial convolution unit, fully connected dilation unit, and dilated channel convolution. The first unit is the dilated spatial convolution, which handles all channels equally and amplifies the valuable aspects. The second unit is a fully connected dilation unit; it displays low-level properties such as edges, contours, and color. The third unit is dilated channel convolution, which detects tiny objects and thin boundaries without adding complexity. The proposed dilated residual grooming kernel (DRGK) model is a 14-layer deep learning model that stretches the receptive field while retaining feature information, using the proposed DR unit and ghost model, as well as convolution, pooling, downsampling, and dilated convolution. Dilated convolutions are extensively used in the proposed model to extract features. Accuracy, the area under the curve, average precision score, precision, sensitivity, and f1 score all improve with a learning rate of 0.001. With 96.33% and 93.35% marks, the proposed approach surpasses several state-of-the-art methods.

Ada-GridRF: A Fast and Automated Adaptive Boost Based Grid Search Optimized Random Forest Ensemble model for Lung Cancer Detection.

Lung cancer is considered one of the leading causes of death all across the world. Various radiology-related fields increasingly have used Computer-aided diagnosis (CAD) systems. It just has already become a part of clinical work for lung cancer detection. In this article, we proposed an Adaptive Boost-based Grid Search Optimized Random Forest (Ada-GridRF) classifier that best optimized the hyperparameters of the base random forest model to identify the malignant and non-malignant nodules from the trained CT images. Improved performance speed and reduced computational complexity were the advantages of the proposed method. The proposed methodology was compared with other hyperparameter optimization techniques and also with different conventional approaches. It even outperformed the popular state-of-the-art deep learning techniques such as transfer learning and convolutional neural network. The experimental results proved that the proposed method yielded the best performance metrics of 97.97% accuracy, 100% sensitivity, 96% specificity, 96.08% precision, 98% F1-score, 4% False positives rate, and 99.8% Area under the ROC curve (AUC). It took only 8 msec to train the model. Thus, the proposed Ada-GridRF model can aid radiologists in fast lung cancer detection.

A Two-Stage Model for Breast Cancer Detection and Classification from Mammograms

Breast cancer is a type of disease which is common to women. It is an abnormal behaviour which hinders the functionality of the normal breast cells which leads to forming a tumor in the form of lump. Early detection of breast cancers tends to reduce death risk and provides a chance of a suitable and better treatment. A two-stage classification model was proposed to detect and classify breast cancer using two publicly available datasets namely MIAS and the DDSM dataset. The first classification stage uses Convolution Neural Network (CNN) to classify the mammogram image as either normal or abnormal while K-Nearest Neighbour (KNN) was used for the second classification stage to classify the abnormal image into benign or malignant. The features used for the second stage classification were extracted using Gray Level Co-occurrence Matrix (GLCM), which was later used for classification. The performance was evaluated based on accuracy, sensitivity, specificity, precision, False Positive Rate (FPR), F1 score and Matthews Correlation Coefficient (MCC). The results show that CNN obtained the highest accuracy, sensitivity, specificity, precision, FPR, F1 score and MCC of 99.03%, 0.9831, 1.0000, 1.0000, 0.0000, 0.9915 and 0.9804 respectively, while KNN has an accuracy, sensitivity, specificity, precision, FPR, F1 score and MCC of 76.27%, 0.7667, 0.7586, 0.7667, 0.2414, 0.7667 and 0.5253 respectively.

Hybrid Rider Optimization with Deep Learning Driven Biomedical Liver Cancer Detection and Classification.

Biomedical engineering is the application of the principles and problem-solving methods of engineering to biology along with medicine. Computation intelligence is the study of design of intelligent agents which are systems acting perceptively. The computation intelligence paradigm offers more advantages to the enhancement and maintenance of the field of biomedical engineering. Liver cancer is the major reason of mortality worldwide. Earlier-stage diagnosis and treatment might increase the survival rate of liver cancer patients. Manual recognition of the cancer tissue is a time-consuming and difficult task. Hence, a computer-aided diagnosis (CAD) is employed in decision making procedures for accurate diagnosis and effective treatment. In contrast to classical image-dependent “semantic” feature evaluation from human expertise, deep learning techniques could learn feature representation automatically from sample images using convolutional neural network (CNN). This study introduces a Hybrid Rider Optimization with Deep Learning Driven Biomedical Liver Cancer Detection and Classification (HRO-DLBLCC) model. The proposed HRO-DLBLCC model majorly focuses on the identification of liver cancer in the medical images. To do so, the proposed HRO-DLBLCC model employs preprocessing in two stages, namely, Gabor filtering (GF) based noise removal and watershed transform based segmentation. In addition, the proposed HRO-DLBLCC model involves NAdam optimizer with DenseNet-201 based feature extractor to generate an optimal set of feature vectors. Finally, the HRO algorithm with recurrent neural network–long short-term memory (RNN-LSTM) model is applied for liver cancer classification, in which the hyperparameters of the RNN-LSTM model are tuned by the use of HRO algorithm. The HRO-DLBLCC model is experimentally validated and compared with existing models. The experimental results assured the promising performance of the HRO-DLBLCC model over recent approaches.

A Novel Hybrid Deep Learning Model for Metastatic Cancer Detection.

Cancer has been found as a heterogeneous disease with various subtypes and aims to destroy the body's normal cells abruptly. As a result, it is essential to detect and prognosis the distinct type of cancer since they may help cancer survivors with treatment in the early stage. It must also divide cancer patients into high- and low-risk groups. While realizing efficient detection of cancer is frequently a time-taking and exhausting task with the high possibility of pathologist errors and previous studies employed data mining and machine learning (ML) techniques to identify cancer, these strategies rely on handcrafted feature extraction techniques that result in incorrect classification. On the contrary, deep learning (DL) is robust in feature extraction and has recently been widely used for classification and detection purposes. This research implemented a novel hybrid AlexNet-gated recurrent unit (AlexNet-GRU) model for the lymph node (LN) breast cancer detection and classification. We have used a well-known Kaggle (PCam) data set to classify LN cancer samples. This study is tested and compared among three models: convolutional neural network GRU (CNN-GRU), CNN long short-term memory (CNN-LSTM), and the proposed AlexNet-GRU. The experimental results indicated that the performance metrics accuracy, precision, sensitivity, and specificity (99.50%, 98.10%, 98.90%, and 97.50) of the proposed model can reduce the pathologist errors that occur during the diagnosis process of incorrect classification and significantly better performance than CNN-GRU and CNN-LSTM models. The proposed model is compared with other recent ML/DL algorithms to analyze the model's efficiency, which reveals that the proposed AlexNet-GRU model is computationally efficient. Also, the proposed model presents its superiority over state-of-the-art methods for LN breast cancer detection and classification.

A Smart Solution for Cancer Patient Monitoring Based on Internet of Medical Things Using Machine Learning Approach.

The Internet of Medical Things (IoMT) is a huge, exciting new phenomenon that is changing the world of technology and innovating various industries, including healthcare. It has specific applications and changes in the medical world based on what can be done for clinical workflow models. The first and most fundamental thing that IoMT does in healthcare is to bring a flood of new data into medical processes. In this study, an efficient Internet of Medical Things based cancer detection model was proposed. In fact, for many, new fitness monitors and watches are one of the best examples on the Internet; these mobile, portable, wearable devices can record real-time heart rate, blood pressure, and eye movement of cancer patients. These details are sent to doctors or anywhere else. The proposed method leads to a kind of big data renaissance in the health service. The proposed model gets more accuracy while comparing with the existing models. This will help the doctors to analyze the patients' health report and provides better treatment.

Dense Convolutional Neural Network for Detection of Cancer from CT Images.

In this paper, we develop a detection module with strong training testing to develop a dense convolutional neural network model. The model is designed in such a way that it is trained with necessary features for optimal modelling of the cancer detection. The method involves preprocessing of computerized tomography (CT) images for optimal classification at the testing stages. A 10-fold cross-validation is conducted to test the reliability of the model for cancer detection. The experimental validation is conducted in python to validate the effectiveness of the model. The result shows that the model offers robust detection of cancer instances that novel approaches on large image datasets. The simulation result shows that the proposed method provides analyzes with 94% accuracy than other methods. Also, it helps to reduce the detection errors while classifying the cancer instances than other methods the several existing methods.

Abstract 6371: Deep learning algorithm for multi-cancer detection and classification using cf-WGS

Abstract Purpose: Several cell-free DNA (cf-DNA) features, such as genome-wide coverage, fragment size, and fragment end motif frequency, have shown their potentials for cancer detection. In this study, we developed two independent models, GC (gross chromatin), and FEMS (fragment end motif frequency and size). Each model uses images generated from genome-wide normalized sequencing coverage and cf-DNA fragment end motif frequencies according to the different cf-DNA size profiles. Then we integrated them into a single ensemble model to improve cancer detection and multi-cancer type classification accuracy. Methods: Low depth cf-WGS data was generated from 1,396 patients (stage I: 14.9%, stage II: 35.6%, stage III: 24.9%, stage IV: 24.2%, unknown: 0.4%) with breast (n=702), liver (n=213), esophageal (n=155), ovarian (n=151), pancreatic (n=85), lung (n=53), head and neck (n=16), biliary tract (n=15), and colon cancer (n=6) and 417 healthy individuals. Samples were randomly split into training, validation, and test set stratifying cancer type and stages. Cancer types with a small number of samples (<20) were excluded for multi-cancer type classification. Each model was trained using a convolutional neural network, then integrated into a single ensemble model by averaging the predicted probabilities calculated from each model. Results: For cancer detection, the ensemble model achieved sensitivities of 85.2% [95% confidence interval (CI): 71.8% to 94.5%], 74.9% (CI: 68.0% to 88.0%), 73.2% (CI: 66.7% to 85.9%) at a specificity of 95%, 98% and 99% and the AUC value of 0.97(CI: 0.95-0.99) in the test dataset. By the cancer stages, sensitivity was 62.8% (CI: 48.8% to 83.7%) in stage I, 66.3% (CI: 57.7% to 82.7%) in stage II, 85.9% (CI: 77.5% to 94.4%) in stage III, and 76.1% (CI: 63.4% to 87.3%) in stage IV at 99% specificity. For multi-cancer classification, the overall accuracy of 85.1% (CI: 80.4% to 89.3%) was achieved including 6 cancer types. Conclusions: Highly sensitive and accurate deep learning model for cancer detection and multi-cancer classification was generated by combining different types of cf-DNA features. This result provides the opportunity for general population multi-cancer screening using various cf-DNA features. Citation Format: Tae-Rim Lee, Jin Mo Ahn, Joo Hyuk Sohn, Sook Ryun Park, Min Hwan Kim, Gun Min Kim, Ki-Byung Song, Eunsung Jun, Dongryul Oh, Jeong-Won Lee, Joseph J Noh, Young Sik Park, Sun-Young Kong, Sang Myung Woo, Bo Hyun Kim, Eui Kyu Chie, Hyun-Cheol Kang, Youn Jin Choi, Ki-Won Song, Jeong-Sik Byeon, Junnam Lee, Dasom Kim, Chang-Seok Ki, Eunhae Cho. Deep learning algorithm for multi-cancer detection and classification using cf-WGS [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6371.

Abstract 5150: Multimodal analysis of plasma cell-free DNA methylome for sensitive multi-cancer detection

Abstract Introduction: Genomic scale copy number, methylation, and fragmentation aberrations are proven genetic and epigenetic biomarkers of circulating tumor DNA (ctDNA). We developed a novel cell-free DNA (cfDNA) methylome sequencing assay that allows for integrative analysis of these genomic features for sensitive detection of multiple types of cancer. Methods: Whole methylome sequencing (WMS) libraries were generated from enzymatically converted cfDNA. Low-pass (~2X) paired-end NGS sequencing was performed on WMS libraries and paired whole genome sequencing (WGS) libraries of unconverted cfDNA for technical comparison and analytical validation. For development of cancer detection models, we profiled the genome-wide methylation density (MD), fragment size index (FSI), fragment end motif (motif) and chromosome instability (CIN) based on WMS data from a discovery cohort of 352 healthy controls and 559 newly diagnosed cancer patients (45 breast, 105 colorectal, 44 esophageal, 79 gastric, 79 liver, 110 lung, 83 pancreatic, and 14 others), 34.5% of which were at stage I or II. Machine learning models, including KNN, SVM, LR, GBDT, and random forest were trained and tested for individual biomarker types, with a final ensemble classifier to integrate all biomarkers. Performance of the predictive model was confirmed on an independent validation cohort consisting of 145 healthy controls and 236 cancer patients (21 breast, 45 colorectal, 18 esophageal, 35 gastric, 34 liver, 47 lung, and 36 pancreatic), among which 31.8% were at early stages (I or II). Results: WMS and WGS data from 512 cfDNA samples showed high concordance in CIN (R=0.988, 95% CI: 0.986-0.990) and FSI (R=0.961, 0.954-0.967) profiles. On the independent validation cohort, the optimal model selected for each of individual genomic features achieved following area under the ROC curve (AUC) values for cancer detection: MD-KNN, 0.830 (0.789-0.870); FSI-SVM, 0.904 (0.874-0.933); motif-SVM, 0.943 (0.920-0.966); and CIN-PAscore, 0.812 (0.770-0.854). The ensemble classifier based on linear SVM outperformed individual biomarkers, with an AUC value of 0.952 (0.934-0.971), which translated to, at 95% specificity, detection sensitivity of 66.7% for breast, 77.8% for colorectal, 83.3% for esophageal, 62.9% for gastric, 82.4% for liver, 66.0% for lung, and 77.8% for pancreatic cancers. Noteworthily, the overall sensitivity on early-stage cancer was 74.7%. Conclusions: These results demonstrate the first proof of principle on the feasibility of integrating multiple genomic cancer markers on the same WMS technical platform. Low-pass WMS on plasma cfDNA from 10ml of blood with integrative multimodal analysis of methylation, fragmentation, and CNV profiles yields in satisfactory sensitivity and specificity for detection of multiple types of cancer, warranting a forthcoming prospective study to further assess its clinical performance in a larger cohort. Citation Format: Yulong Li, Fenglong Bie, Fengwei Tan, Tiancheng Han, Shunli Yang, Fang Lv, Peiyao Nie, Qi Zhang, Yuanyuan Hong, Zhijie Wang, Ji He, Weizhi Chen, Liang Zhao, Shugeng Gao. Multimodal analysis of plasma cell-free DNA methylome for sensitive multi-cancer detection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5150.