Fusing hand-crafted and deep-learning features in a convolutional neural network model to identify prostate cancer in pathology images

Frontiers in Oncology, Год журнала: 2022, Номер 12

Опубликована: Сен. 27, 2022

Prostate cancer can be diagnosed by prostate biopsy using transectal ultrasound guidance. The high number of pathology images from tissues is a burden on pathologists, and analysis subjective susceptible to inter-rater variability. use machine learning techniques could make histopathology diagnostics more precise, consistent, efficient overall. This paper presents new classification fusion network model that was created fusing eight advanced image features: seven hand-crafted features one deep-learning feature. These are the scale-invariant feature transform (SIFT), speeded up robust (SURF), oriented accelerated segment test (FAST) rotated binary independent elementary (BRIEF) (ORB) local features, shape texture cell nuclei, histogram gradients (HOG) cavities, color feature, convolution Matching, integrated, networks three essential components proposed network. integrated consists both backbone an additional When classifying 1100 this with different backbones (ResNet-18/50, VGG-11/16, DenseNet-121/201), we discovered ResNet-18 achieved best performance in terms accuracy (95.54%), specificity (93.64%), sensitivity (97.27%) as well area under receiver operating characteristic curve (98.34%). However, each assessment criteria for these separate had value lower than 90%, which demonstrates suggested combines differently derived characteristics effective manner. Moreover, Grad-CAM++ heatmap used observe differences between regions interest. map showed better at focusing cancerous cells ResNet-18. Hence, network, useful computer-aided diagnoses based cancer. Because similarities engineering deep types images, method other such those breast, thyroid

Язык: Английский

---

Dynamic radiomics for predicting the efficacy of antiangiogenic therapy in colorectal liver metastases

Frontiers in Oncology, Год журнала: 2023, Номер 13

Опубликована: Фев. 6, 2023

Background and objective For patients with advanced colorectal liver metastases (CRLMs) receiving first-line anti-angiogenic therapy, an accurate, rapid noninvasive indicator is urgently needed to predict its efficacy. In previous studies, dynamic radiomics predicted more accurately than conventional radiomics. Therefore, it necessary establish a efficacy prediction model for antiangiogenic therapy provide accurate guidance clinical diagnosis treatment decisions. Methods this study, we use feature extraction method that extracts static features using tomographic images of different sequences the same patient then quantifies them into new treatmentefficacy. retrospective collected 76 who were diagnosed unresectable CRLM between June 2016 2021 in First Hospital China Medical University. All received standard regimen bevacizumab combined chemotherapy treatment, contrast-enhanced abdominal CT (CECT) scans performed before treatment. Patients multiple primary lesions as well missing or imaging information excluded. Area Under Curve (AUC) accuracy used evaluate performance. Regions interest (ROIs) independently delineated by two radiologists extract features. Three machine learning algorithms construct scores based on best response progression-free survival (PFS). Results task will achieve after ROC curve analysis, can be seen relative change rate (RCR) among all linear discriminantanalysis (AUC: 0.945 accuracy: 0.855). terms predicting PFS, Kaplan–Meier plots suggested score constructed RCR could significantly distinguish good from those poor (Two-sided P<0.0001 analysis). Conclusions This study demonstrates application better compared It allows have assessment effect medical noninvasive, rapid, less expensive. Dynamic provides stronger selection options precision medicine.

Язык: Английский

---

A Robust [18F]-PSMA-1007 Radiomics Ensemble Model for Prostate Cancer Risk Stratification

Deleted Journal, Год журнала: 2024, Номер unknown

Опубликована: Сен. 30, 2024

Abstract The aim of this study is to investigate the role [ 18 F]-PSMA-1007 PET in differentiating high- and low-risk prostate cancer (PCa) through a robust radiomics ensemble model. This retrospective included 143 PCa patients who underwent PET/CT imaging. areas were manually contoured on images 1781 image biomarker standardization initiative (IBSI)-compliant features extracted. A 30 times iterated preliminary analysis pipeline, comprising least absolute shrinkage selection operator (LASSO) for feature fivefold cross-validation model optimization, was adopted identify most dataset variations, select candidate models modelling, optimize hyperparameters. Thirteen subsets selected features, 11 generated from plus two additional subsets, first based combination fine-tuning second only used train ensemble. Accuracy, area under curve (AUC), sensitivity, specificity, precision, f -score values calculated provide models’ performance. Friedman test, followed by post hoc tests corrected with Dunn-Sidak correction multiple comparisons, verify if statistically significant differences found different over iterations. trained obtained highest average accuracy (79.52%), AUC (85.75%), specificity (84.29%), precision (82.85%), (78.26%). Statistically ( p < 0.05) some performance metrics. These findings support improving risk stratification PCa, reducing dependence biopsies.

Язык: Английский

---

Comparison and fusion prediction model for lung adenocarcinoma with micropapillary and solid pattern using clinicoradiographic, radiomics and deep learning features

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Июнь 8, 2023

Abstract To investigate whether the combination scheme of deep learning score (DL-score) and radiomics can improve preoperative diagnosis in presence micropapillary/solid (MPP/SOL) patterns lung adenocarcinoma (ADC). A retrospective cohort 514 confirmed pathologically ADC 512 patients after surgery was enrolled. The clinicoradiographic model (model 1) 2) were developed with logistic regression. 3) constructed based on (DL-score). combine 4) DL-score R-score variables. performance these models evaluated area under receiver operating characteristic curve (AUC) compared using DeLong's test internally externally. prediction nomogram plotted, clinical utility depicted decision curve. 1, 2, 3 4 supported by AUCs 0.848, 0.896, 0.906, 0.921 Internal validation set, that 0.700, 0.801, 0.730, 0.827 external respectively. These existed statistical significance internal vs 3, P = 0.016; 0.009, respectively) 0.036; 0.047; 0.016, respectively). analysis (DCA) demonstrated predicting MPP/SOL structure would be more beneficial than 1and but comparable 2. combined pattern practice.

Язык: Английский

---

The Potential of Ultrasound Radiomics in Carpal Tunnel Syndrome Diagnosis: A Systematic Review and Meta-Analysis

Diagnostics, Год журнала: 2023, Номер 13(20), С. 3280 - 3280

Опубликована: Окт. 23, 2023

Background: Carpal tunnel syndrome (CTS) is the most common entrapment neuropathy for which ultrasound imaging has recently emerged as a valuable diagnostic tool. This meta-analysis aims to investigate role of radiomics in diagnosis CTS and compare it with other approaches. Methods: We conducted comprehensive search electronic databases from inception September 2023. The included studies were assessed quality using Quality Assessment Tool Diagnostic Accuracy Studies. primary outcome was performance compared radiologist evaluation diagnosing CTS. Results: Our five observational comprising 840 participants. In context evaluation, combined statistics sensitivity, specificity, odds ratio 0.78 (95% confidence interval (CI), 0.71 0.83), 0.72 CI, 0.59 0.81), 9 5 15), respectively. contrast, training mode yielded sensitivity 0.88 0.85 0.91), specificity 0.84 0.92), 58 38 87). Similarly, testing demonstrated an aggregated 0.89), 0.80 0.73 0.85), 22 12 41). Conclusions: contrast assessments by radiologists, exhibited superior detecting Furthermore, there minimal variability accuracy between sets radiomics, highlighting its potential robust tool

Язык: Английский

---

Radiomics Analyses to Predict Histopathology in Patients with Metastatic Testicular Germ Cell Tumors before Post-Chemotherapy Retroperitoneal Lymph Node Dissection

Journal of Imaging, Год журнала: 2023, Номер 9(10), С. 213 - 213

Опубликована: Окт. 7, 2023

Background: The identification of histopathology in metastatic non-seminomatous testicular germ cell tumors (TGCT) before post-chemotherapy retroperitoneal lymph node dissection (PC-RPLND) holds significant potential to reduce treatment-related morbidity young patients, addressing an important survivorship concern. Aim: To explore this possibility, we conducted a study investigating the role computed tomography (CT) radiomics models that integrate clinical predictors, enabling personalized prediction TGCT patients prior PC-RPLND. In retrospective study, included cohort 122 patients. Methods: Using dedicated software, segmented targets and extracted quantitative features from CT images. Subsequently, employed feature selection techniques developed radiomics-based machine learning predict histological subtypes. ensure robustness our procedure, implemented 5-fold cross-validation approach. When evaluating models’ performance, measured metrics such as area under receiver operating characteristic curve (AUC), sensitivity, specificity, precision, F-score. Result: Our model based on Support Vector Machine achieved optimal average AUC 0.945. Conclusions: presented CT-based can potentially serve non-invasive tool histopathological outcomes, differentiating among fibrosis/necrosis, teratoma, viable tumor It has be considered promising mitigate risk over- or under-treatment although multi-center validation is critical confirm utility proposed workflow.

Язык: Английский

---

Biodistribution Assessment of a Novel 68Ga-Labeled Radiopharmaceutical in a Cancer Overexpressing CCK2R Mouse Model: Conventional and Radiomics Methods for Analysis

Life, Год журнала: 2024, Номер 14(3), С. 409 - 409

Опубликована: Март 20, 2024

The aim of the present study consists evaluation biodistribution a novel 68Ga-labeled radiopharmaceutical, [68Ga]Ga-NODAGA-Z360, injected into Balb/c nude mice through histopathological analysis on bioptic samples and radiomics positron emission tomography/computed tomography (PET/CT) images. radiopharmaceutical was designed to specifically bind cholecystokinin receptor (CCK2R). This receptor, naturally in healthy tissues such as stomach, is biomarker for numerous tumors when overexpressed. In this experiment, were xenografted with human epidermoid carcinoma A431 cell line (A431 WT) overexpressing CCK2R CCK2R+), while controls received wild-type line. PET images processed, segmented after atlas-based co-registration and, consequently, 112 features extracted each investigated organ / tissue. To confirm histopathology at tissue level correlate it degree uptake, studies supported by digital pathology. As result analyses, differences different body districts confirmed correct targeting radiopharmaceutical. preclinical imaging, methodology confirms importance decision-support system based artificial intelligence algorithms assessment biodistribution.

Язык: Английский

---

Preclinical Implementation of matRadiomics: A Case Study for Early Malformation Prediction in Zebrafish Model

Journal of Imaging, Год журнала: 2024, Номер 10(11), С. 290 - 290

Опубликована: Ноя. 14, 2024

Radiomics provides a structured approach to support clinical decision-making through key steps; however, users often face difficulties when switching between various software platforms complete the workflow. To streamline this process, matRadiomics integrates entire radiomics workflow within single platform. This study extends

Язык: Английский

---

Implementation of Automatic Segmentation Framework as Preprocessing Step for Radiomics Analysis of Lung Anatomical Districts

BioMedInformatics, Год журнала: 2024, Номер 4(4), С. 2309 - 2320

Опубликована: Дек. 11, 2024

Background: The advent of artificial intelligence has significantly impacted radiology, with radiomics emerging as a transformative approach that extracts quantitative data from medical images to improve diagnostic and therapeutic accuracy. This study aimed enhance the radiomic workflow by applying deep learning, through transfer for automatic segmentation lung regions in computed tomography scans preprocessing step. Methods: Leveraging pipeline articulated (i) patient-based splitting, (ii) intensity normalization, (iii) voxel resampling, (iv) bed removal, (v) contrast enhancement (vi) model training, DeepLabV3+ convolutional neural network (CNN) was fine tuned perform whole-lung-region segmentation. Results: trained achieved high accuracy, Dice coefficient (0.97) BF (93.06%) scores, it effectively preserved region areas removed confounding anatomical such heart spine. Conclusions: introduces learning framework CT images, leveraging an demonstrating excellent performance model, isolating while excluding structures. Ultimately, this work paves way more efficient, automated tools cancer detection, potential clinical decision making patient outcomes.

Язык: Английский

---

CT Rendering and Radiomic Analysis in Post-Chemotherapy Retroperitoneal Lymph Node Dissection for Testicular Cancer to Anticipate Difficulties for Young Surgeons

Journal of Imaging, Год журнала: 2023, Номер 9(3), С. 71 - 71

Опубликована: Март 17, 2023

Post-chemotherapy retroperitoneal lymph node dissection (PC-RPLND) in non-seminomatous germ-cell tumor (NSTGCTs) is a complex procedure. We evaluated whether 3D computed tomography (CT) rendering and their radiomic analysis help predict resectability by junior surgeons. The ambispective was performed between 2016–2021. A prospective group (A) of 30 patients undergoing CT segmented using the Slicer software while retrospective (B) with conventional (without reconstruction). CatFisher’s exact test showed p-value 0.13 for 1.0 Group B. difference proportion 0.009149 (IC 0.1–0.63). correct classification 0.645 0.55–0.87) A, 0.275 0.11–0.43) Furthermore, 13 shape features were extracted: elongation, flatness, volume, sphericity, surface area, among others. Performing logistic regression entire dataset, n = 60, results were: Accuracy: 0.7 Precision: 0.65. Using randomly chosen, best result obtained 0.73 0.83, p-value: 0.025 Fisher’s test. In conclusion, significant prediction versus reconstruction surgeons experienced Radiomic used to elaborate an artificial intelligence model improve resectability. proposed could be great support university hospital, allowing it plan surgery anticipate complications.

Язык: Английский

---