Application of artificial intelligence radiomics in the diagnosis, treatment, and prognosis of hepatocellular carcinoma

Computers in Biology and Medicine, Год журнала: 2024, Номер 173, С. 108337 - 108337

Опубликована: Март 24, 2024

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer, with an increasing incidence and poor prognosis. In past decade, artificial intelligence (AI) technology has undergone rapid development in field clinical medicine, bringing advantages efficient data processing accurate model construction. Promisingly, AI-based radiomics played increasingly important role decision-making HCC patients, providing new technical guarantees for prediction, diagnosis, prognostication. this review, we evaluated current landscape AI management HCC, including its individual treatment, survival Furthermore, discussed remaining challenges future perspectives regarding application HCC.

Язык: Английский

---

MRI‐Based Radiomics and Deep Learning in Biological Characteristics and Prognosis of Hepatocellular Carcinoma: Opportunities and Challenges

Journal of Magnetic Resonance Imaging, Год журнала: 2023, Номер 59(3), С. 767 - 783

Опубликована: Авг. 30, 2023

Hepatocellular carcinoma (HCC) is the fifth most common malignancy and third leading cause of cancer‐related death worldwide. HCC exhibits strong inter‐tumor heterogeneity, with different biological characteristics closely associated prognosis. In addition, patients often distribute at stages require diverse treatment options each stage. Due to variability in tumor sensitivity therapies, determining optimal approach can be challenging for clinicians prior treatment. Artificial intelligence (AI) technology, including radiomics deep learning approaches, has emerged as a unique opportunity improve spectrum clinical care by predicting prognosis medical imaging field. The utilizes handcrafted features derived from specific mathematical formulas construct various machine‐learning models applications. terms approach, convolutional neural network are developed achieve high classification performance based on automatic feature extraction images. Magnetic resonance offers advantage superior tissue resolution functional information. This comprehensive evaluation plays vital role accurate assessment effective planning patients. Recent studies have applied approaches develop AI‐enabled accuracy prognosis, such microvascular invasion recurrence. Although demonstrated promising potential prediction performance, one biggest challenges, interpretability, hindered their implementation practice. future, continued research needed interpretability models, aspects domain knowledge, novel algorithms, multi‐dimension data sources. Overcoming these challenges would allow significantly impact provided patients, ultimately deployment use. Level Evidence 5 Technical Efficacy Stage 2

Язык: Английский

---

An Informative Review of Radiomics Studies on Cancer Imaging: The Main Findings, Challenges and Limitations of the Methodologies

Current Oncology, Год журнала: 2024, Номер 31(1), С. 403 - 424

Опубликована: Янв. 10, 2024

The aim of this informative review was to investigate the application radiomics in cancer imaging and summarize results recent studies support oncological with particular attention breast cancer, rectal primitive secondary liver cancer. This also aims provide main findings, challenges limitations current methodologies. Clinical published last four years (2019–2022) were included review. Among 19 analyzed, none assessed differences between scanners vendor-dependent characteristics, collected images individuals at additional points time, performed calibration statistics, represented a prospective study registered database, conducted cost-effectiveness analysis, reported on clinical application, or multivariable analysis non-radiomics features. Seven reached high radiomic quality score (RQS), seventeen earned by using validation steps considering two datasets from distinct institutes open science data domains (radiomics features calculated set representative ROIs are source). potential is increasingly establishing itself, even if there still several aspects be evaluated before passage into routine practice. There challenges, including need for standardization across all stages workflow cross-site real-world heterogeneous datasets. Moreover, multiple centers more samples that add inter-scanner characteristics will needed future, as well collecting time points, reporting statistics performing database.

Язык: Английский

---

Update on the Applications of Radiomics in Diagnosis, Staging, and Recurrence of Intrahepatic Cholangiocarcinoma

Diagnostics, Год журнала: 2023, Номер 13(8), С. 1488 - 1488

Опубликована: Апрель 20, 2023

Background: This paper offers an assessment of radiomics tools in the evaluation intrahepatic cholangiocarcinoma. Methods: The PubMed database was searched for papers published English language no earlier than October 2022. Results: We found 236 studies, and 37 satisfied our research criteria. Several studies addressed multidisciplinary topics, especially diagnosis, prognosis, response to therapy, prediction staging (TNM) or pathomorphological patterns. In this review, we have covered diagnostic developed through machine learning, deep neural network recurrence biological characteristics. majority were retrospective. Conclusions: It is possible conclude that many performing models been make differential diagnosis easier radiologists predict genomic However, all retrospective, lacking further external validation prospective multicentric cohorts. Furthermore, expression results should be standardized automatized applicable clinical practice.

Язык: Английский

---

Development and validation of an ultrasound-based deep learning radiomics nomogram for predicting the malignant risk of ovarian tumours

BioMedical Engineering OnLine, Год журнала: 2024, Номер 23(1)

Опубликована: Апрель 9, 2024

Abstract Background The timely identification and management of ovarian cancer are critical determinants patient prognosis. In this study, we developed validated a deep learning radiomics nomogram (DLR_Nomogram) based on ultrasound (US) imaging to accurately predict the malignant risk tumours compared diagnostic performance DLR_Nomogram that ovarian-adnexal reporting data system (O-RADS). Methods This study encompasses two research tasks. Patients were randomly divided into training testing sets in an 8:2 ratio for both task 1, assessed malignancy 849 patients with tumours. 2, evaluated 391 O-RADS 4 5 neoplasms. Three models predicted outcomes each sample merged form new feature set was utilised as input logistic regression (LR) model constructing combined model, visualised DLR_Nomogram. Then, these by receiver operating characteristic curve (ROC). Results demonstrated superior predictive predicting tumours, evidenced area under ROC (AUC) values 0.985 0.928 respectively. AUC value its lower than O-RADS; however, difference not statistically significant. exhibited highest 0.955 0.869 showed satisfactory fitting tasks Hosmer–Lemeshow testing. Decision analysis yielded greater net clinical benefits within specific range threshold values. Conclusions US-based has shown capability exhibiting efficacy comparable O-RADS.

Язык: Английский

---

Ultrasound-based deep learning radiomics model for differentiating benign, borderline, and malignant ovarian tumours: a multi-class classification exploratory study

BMC Medical Imaging, Год журнала: 2024, Номер 24(1)

Опубликована: Апрель 15, 2024

Abstract Background Accurate preoperative identification of ovarian tumour subtypes is imperative for patients as it enables physicians to custom-tailor precise and individualized management strategies. So, we have developed an ultrasound (US)-based multiclass prediction algorithm differentiating between benign, borderline, malignant tumours. Methods We randomised data from 849 with tumours into training testing sets in a ratio 8:2. The regions interest on the US images were segmented handcrafted radiomics features extracted screened. applied one-versus-rest method classification. inputted best machine learning (ML) models constructed radiomic signature (Rad_Sig). maximum trimmed sections pre-trained convolutional neural network (CNN) model. After internal enhancement complex algorithms, each sample’s predicted probability, known deep transfer (DTL_Sig), was generated. Clinical baseline analysed. Statistically significant clinical parameters semantic set used construct signatures (Clinic_Sig). results Rad_Sig, DTL_Sig, Clinic_Sig sample fused new feature sets, build combined model, namely, (DLR_Sig). receiver operating characteristic (ROC) curve area under ROC (AUC) estimate performance classification Results included 440 44 196 109 11 49 DLR_Sig three-class model had overall class-specific performance, micro- macro-average AUC 0.90 0.84, respectively, set. Categories 0.85, 0.83 tumours, respectively. In confusion matrix, classifier Rad_Sig could not recognise borderline However, proportions identified by highest at 54.55% 63.27%, Conclusions US-based can discriminate Therefore, may guide clinicians determining differential

Язык: Английский

---

Machine Learning Approaches with Textural Features to Calculate Breast Density on Mammography

Current Oncology, Год журнала: 2023, Номер 30(1), С. 839 - 853

Опубликована: Янв. 7, 2023

breast cancer (BC) is the world's most prevalent in female population, with 2.3 million new cases diagnosed worldwide 2020. The great efforts made to set screening campaigns, early detection programs, and increasingly targeted treatments led significant improvement patients' survival. Full-Field Digital Mammograph (FFDM) considered gold standard method for diagnosis of BC. From several previous studies, it has emerged that density (BD) a risk factor development BC, affecting periodicity plans present today at an international level.in this study, focus mammographic image processing techniques allow extraction indicators derived from textural patterns mammary parenchyma indicative BD factors.a total 168 patients were enrolled internal training test while 51 compose external validation cohort. Different Machine Learning (ML) have been employed classify breasts based on values tissue density. Textural features extracted only which train classifiers, thanks aid ML algorithms.the accuracy different tested classifiers varied between 74.15% 93.55%. best results reached by Support Vector (accuracy 93.55% percentage true positives negatives equal TPP = 94.44% TNP 92.31%). was not influenced choice selection approach. Considering cohort, SVM, as classifier 7 selected wrapper method, showed 0.95, sensitivity 0.96, specificity 0.90.our preliminary Radiomics analysis approach us objectively identify BD.

Язык: Английский

---

Radiomics-based Machine Learning to Predict the Recurrence of Hepatocellular Carcinoma: A Systematic Review and Meta-analysis

Academic Radiology, Год журнала: 2023, Номер 31(2), С. 467 - 479

Опубликована: Окт. 20, 2023

Язык: Английский

---

Deep Learning Model Based on Contrast-Enhanced Computed Tomography Imaging to Predict Postoperative Early Recurrence after the Curative Resection of a Solitary Hepatocellular Carcinoma

Cancers, Год журнала: 2023, Номер 15(7), С. 2140 - 2140

Опубликована: Апрель 4, 2023

We aimed to develop the deep learning (DL) predictive model for postoperative early recurrence (within 2 years) of hepatocellular carcinoma (HCC) based on contrast-enhanced computed tomography (CECT) imaging. This study included 543 patients who underwent initial hepatectomy HCC and were randomly classified into training, validation, test datasets at a ratio 8:1:1. Several clinical variables arterial CECT images used create models recurrence. Artificial intelligence implemented using convolutional neural networks multilayer perceptron as classifier. Furthermore, Youden index was discriminate between high- low-risk groups. The importance values each explanatory variable calculated permutation importance. DL developed with area under curve 0.71 (test datasets) 0.73 (validation datasets). Postoperative incidences in groups 73% 30%, respectively (p = 0.0057). Permutation demonstrated that among variables, highest value imaging analysis. predict DL-based analysis is effective determining treatment strategies HCC.

Язык: Английский

---

CEMRI-Based Quantification of Intratumoral Heterogeneity for Predicting Aggressive Characteristics of Hepatocellular Carcinoma Using Habitat Analysis: Comparison and Combination of Deep Learning

Academic Radiology, Год журнала: 2023, Номер 31(6), С. 2346 - 2355

Опубликована: Дек. 6, 2023

Язык: Английский

---