Integration of Deep Learning and Sub-regional Radiomics Improves the Prediction of Pathological Complete Response to Neoadjuvant Chemoradiotherapy in Locally Advanced Rectal Cancer Patients

Academic Radiology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

The precise prediction of response to neoadjuvant chemoradiotherapy is crucial for tailoring perioperative treatment in patients diagnosed with locally advanced rectal cancer (LARC). This retrospective study aims develop and validate a model that integrates deep learning sub-regional radiomics from MRI imaging predict pathological complete (pCR) LARC. We retrospectively enrolled 768 eligible participants three independent hospitals who had received followed by radical surgery. Pretreatment pelvic scans (T2-weighted), were collected annotation feature extraction. K-means approach was used segment the tumor into sub-regions. Radiomics features extracted Pyradiomics 3D ResNet50, respectively. predictive models developed using radiomics, machine algorithm training cohort, then validated external tests. models' performance assessed various metrics, including area under curve (AUC), decision analysis, Kaplan-Meier survival analysis. constructed combined model, named SRADL, which includes signatures, enabling pCR LARC patients. SRADL satisfactory cohort (AUC 0.925 [95% CI 0.894 0.948]), test 1 0.915 0.869 0.949]) 2 0.902 0.846 0.945]). By employing optimal threshold 0.486, predicted group longer compared non-pCR across cohorts. also outperformed other single-modality models. novel showed high accuracy robustness predicting pretreatment images, making it promising tool personalized management

Language: Английский

---

Evaluation of EGFR-TKIs and ICIs treatment stratification in non-small cell lung cancer using an encrypted multidimensional radiomics approach

Cancer Imaging, Journal Year: 2025, Volume and Issue: 25(1)

Published: Jan. 20, 2025

Language: Английский

---

Optimized theory-guided convolutional neural network for lung cancer classification using CT images with advanced FPGA implementation

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107719 - 107719

Published: Feb. 22, 2025

Language: Английский

---

A novel deep learning radiopathomics model for predicting carcinogenesis promotor cyclooxygenase-2 expression in common bile duct in children with pancreaticobiliary maljunction: a multicenter study

Insights into Imaging, Journal Year: 2025, Volume and Issue: 16(1)

Published: March 27, 2025

Abstract Objectives To develop and validate a deep learning radiopathomics model (DLRPM) integrating radiological pathological imaging data to predict biliary cyclooxygenase-2 (COX-2) expression in children with pancreaticobiliary maljunction (PBM), compare its performance single-modality radiomics, radiomics (DLR), pathomics models. Methods This retrospective study included 219 PBM patients, divided into training set ( n = 104; median age, 2.8 years, 75.0% females) internal test 71; 2.2 83.1% from center I, an external 44; 3.4 65.9% II. Biliary COX-2 was detected using immunohistochemistry. Radiomics, DLR, features were extracted portal venous-phase CT images H&E-stained histopathological slides, respectively, build individual These then integrated the DLRPM, combining three predictive signatures. Model evaluated AUC, net reclassification index (NRI, for assessing improvement correct classification) discrimination (IDI). Results The DLRPM demonstrated highest performance, AUCs of 0.851 (95% CI, 0.759–0.942) 0.841 0.721–0.960) set. In comparison, models 0.532–0.602, 0.658–0.660, 0.787–0.805, respectively. significantly outperformed models, as by NRI IDI tests (all p < 0.05). Conclusion multimodal could accurately robustly expression, facilitating risk stratification personalized postoperative management PBM. However, prospective multicenter studies larger cohorts are needed further generalizability. Critical relevance statement Our proposed model, images, provides novel cost-effective approach potentially advancing individualized improving long-term outcomes pediatric patients maljunction. Key Points Predicting (PBM) is critical but challenging. A achieved high accuracy COX-2. supports patient Graphical

Language: Английский

---

Advancements in the Application of the Intersection of Medicine and Engineering in Cancer Research

Published: April 7, 2025

ABSTRACT Cancer research predominantly centers on diagnosis, treatment, and elucidation of underlying mechanisms. Nevertheless, the intricate nature tumor genesis development has rendered early diagnostic therapeutic outcomes less than optimal, making conquest a formidable challenge. The interdisciplinary fusion medicine engineering, termed “intersection engineering”, emerged as groundbreaking paradigm, offering novel avenues for advancing cancer studies. As this approach evolves, it yielded numerous breakthroughs in mechanistic exploration. In review, we summarize how intersection engineering propels progress by leveraging combined strengths medicine, bioinformatics, materials science, artificial intelligence. This addresses limitations traditional diagnostics therapies, such low sensitivity, poor efficacy, significant side effects, challenges associated with Moreover, highlight global cutting‐edge advancements potential future directions field.

Language: Английский

---

Habitat Radiomics and Deep Learning Features Based on CT for Predicting Lymphovascular Invasion in T1-stage Lung Adenocarcinoma: A Multicenter Study

Academic Radiology, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

Language: Английский

---

A Systematic Literature Review on Lung Cancer with Ensemble Learning

Lecture notes in networks and systems, Journal Year: 2025, Volume and Issue: unknown, P. 389 - 398

Published: Jan. 1, 2025

Language: Английский

---

Application of Chest CT Imaging Feature Model in Distinguishing Squamous Cell Carcinoma and Adenocarcinoma of the Lung

Cancer Management and Research, Journal Year: 2024, Volume and Issue: Volume 16, P. 547 - 557

Published: June 1, 2024

Purpose: In situations where pathological acquisition is difficult, there a lack of consensus on distinguishing between adenocarcinoma and squamous cell carcinoma from imaging images, each doctor can only make judgments based their own experience. This study aims to extract features chest CT, sensitive factors through logistic univariate multivariate analysis, model distinguish lung adenocarcinoma. Methods: We downloaded CT scans with clear diagnosis The Cancer Imaging Archive (TCIA), extracted 19 by radiologist thoracic surgeon, including location, spicule, lobulation, cavity, vacuolar sign, necrosis, pleural traction vascular bundle air bronchogram calcification, enhancement degree, distance pulmonary hilum, atelectasis, hilum bronchial lymph nodes, mediastinal interlobular septal thickening, metastasis, adjacent structures invasion, effusion. Firstly, we apply the glm function R language perform analysis all variables select P < 0.1. Then, selected obtain predictive model. Next, use roc in calculate AUC value draw ROC curve, val.prob Calibrat rmda package DCA curve clinical impact curve. At same time, 45 patients diagnosed surgery or biopsy Radiotherapy Department Thoracic Surgery our hospital 2023 2024 were included validation group. jointly determined recorded two doctors mentioned above image feature data are complete does not require preprocessing, so directly entering statistical calculations. Perform curves, calibration DCA, curves group further validate If performs well group, nomogram demonstrate. Results: 75 TCIA finally 18 for analysis. First, performed, total 5 obtained: Sign, nodes. After conducting modeling = 0.887, was established using cases hospital, Draw 0.865 evaluate accuracy Calibrate reliability practice practicality Conclusion: It possible influential ordinary determine carcinoma. have set up terms discrimination, accuracy, reliability, practicality. Keywords: cancer, LUAD, LSCC, features, predict

Language: Английский

---

Deep Learning and MRI Biomarkers for Precise Lung Cancer Cell Detection and Diagnosis

The Open Bioinformatics Journal, Journal Year: 2024, Volume and Issue: 17(1)

Published: Sept. 19, 2024

Aim This research work aimed to combine different AI methods create a modular diagnosis system for lung cancer, including Convolutional Neural Network (CNN), K-Nearest Neighbors (KNN), VGG16, and Recurrent (RNN) on MRI biomarkers. Models have then been evaluated compared in their effectiveness detecting using meticulously selected dataset containing 2045 images, with emphasis being put documenting the benefits of multimodal approach attacking complexities disease. Background Lung cancer remains most common cause death world, partly because challenges late stage presentation. Although Magnetic Resonance Imaging (MRI) has become critical modality identification staging too often, its is curtailed by interpretative variance among radiologists. Recent advances machine learning hold great promise augmenting analysis perhaps even increasing diagnostic accuracy start timely treatment. In this work, integration advanced models biomarkers solve these problems investigated. Objective The purpose present paper was assess integrating various machine-learning diagnostics, such as CNN, KNN, RNN. involved 2,045 performances were investigated comparing performance metrics determine best configuration interconnection while underpinning necessity accurate diagnoses and, consequently, better patient outcomes. Methods For study, we used 70% training 30% validation. We four photos: Systematic measures included study: accuracy, recall, precision, F1 score. confusion matrices study power every model comprehend pragmatic use real-world predictive capability. Results scores found be convolutional neural network terms tested, F1. rest models, RNN, performed decently but slightly lower than CNN. in-depth through thus established reliability revealing immense insight into capability identifying true positives minimizing false negatives enhancing detection. Conclusion findings obtained shown further support potential improve diagnosis. high sensitivity specificity KNN model, robustness results from VGG16 RNN pointed feasibility detection cancer. Our strong approach, which might impact future practice oncology treatment strategies outcomes medical imaging.

Language: Английский

---

Research advances in tumor diagnosis and early detection

Asia-Pacific Journal of Oncology, Journal Year: 2024, Volume and Issue: unknown, P. 55 - 65

Published: Sept. 13, 2024

This review explores recent advances in tumor diagnosis and early detection, focusing on cutting-edge developments molecular diagnostic technologies, imaging techniques, the integration of multi-omics data. Current methods have limitations terms sensitivity specificity, particularly for detection. However, with continuous progress research emerging especially advent liquid biopsy, which enables detection circulating DNA (ctDNA), exosomes, tumor-educated platelets (TEPs), accuracy cancer significantly improved. Moreover, combined application artificial intelligence high-resolution technology has enhanced precision diagnosis. Despite these advances, challenges, such as high cost difficulties data integration, continue to impede widespread clinical adoption. Therefore, I believe that future should prioritize innovation technologies improve their applicability across various types, ultimately contributing advancement personalized therapy.

Language: Английский

---