Keywords: Diagnosis/Prediction, Radiomics, Functional liver reserve; Delta radiomics; Gd-EOB-DTPA;

Motivation: The current indocyanine green clearance test is not yet widespread. Gd-EOB-DTPA enhanced MRI imaging can assess the overall liver function, but it is challenging to quantitatively and accurately evaluate it.

Goal(s): Our goal is to assess the ability of Gd-EOB-DTPA enhanced MRI radiomics and Delta radiomics for predicting functional liver reserve.

Approach: We obtained MRI images from 117 patients, trained models on 70% of them, and tested models on the remaining patients.

Results: The prediction models based on Gd-EOB-DTPA enhanced MRI radiomics and Delta radiomics are effective diagnostic tools for the quantitative assessment of functional liver reserve.

Impact: Gd-EOB-DTPA enhanced MRI radiomics and Delta radiomics models can quantitatively assess functional liver reserve, aiding clinicians in selecting the right treatment, monitoring treatment progress, and predicting patient survival.

Keywords: Diagnosis/Prediction, Machine Learning/Artificial Intelligence, Adherent perinephric fat

Motivation: Sticky perinephric fat (SPF) poses a surgical challenge for patients with renal cell carcinoma and the pre-operative identification of SPF is of clinical interest.

Goal(s): The aim of this study was to investigate the effectiveness of using MRI-based radiomics features in predicting the presence of SPF.

Approach: Machine learning algorithms were trained using radiomics features from T1-weighted contrast-enhanced MRI images and clinical factors (gender and BMI).

Results: The promising results on internal and external test sets pave the way to validate the current approach in a larger data set.

Impact: Machine learning models trained with MRI-derived radiomics features can provide a tool for preoperative prediction of sticky perinephric fat. The results from this study suggest that this approach may assist in improving surgical prognosis and outcomes.

Keywords: Diagnosis/Prediction, Breast

Motivation: Contrast-free sequences is receiving increasing attention. As a novel technology, radiomics analysis of Sythetic MRI(SyMRI) in breat treatment has not been widely explored.

Goal(s): To analyse the radiomics features extracted from SyMRI and its complementary value to conventional ADC sequence.

Approach: Recursive feature elimination (RFE) was used to select features and support vector machine (SVM) was used to build models. The model performance was assessed by receiver operating characteristic (ROC) analysis and compared by DeLong test.

Results: Delta-radiomics models based on SyMRI sequences outperformed the single time-point models. The SyMRI sequence that complements the conventional ADC is delta-T2 mapping.

Impact: Radiomics modle generated from delta-T2 mapping showed stable performance and complementary value to ADC sequence. Once scan-multiparameter and contrast-free SyMRI can obtain the comparative or even elevated value as ADC in early prediction of NAC response in breast cancer. 

Keywords: Diagnosis/Prediction, Liver

Motivation: MRI-derived texture analysis could assess liver fibrosis effectively, yet lacking large-scale datasets for validation or comparisons with  transient elastography-based liver stiffness measurement (TE-LSM). 

Goal(s): develop and validate the combined radiomics-clinic  model (CoRC) on MRI for triaging clinically significant liver fibrosis (≥ F2),  comparing or combining with TE-LSM.

Approach: This retrospective multi-center study recruited 595 patients with biopsy proven liver fibrosis. CoRC model integrated Radiomics features extracted from the ResUNet-based automated entire liver segmentation and clinical variables with multivariate logistic regression.

Results: Additive value of CoRC model to TE-LSM was explored with combined AUC of 0.86, and 0.81 in the internal, and temporal sets.

Impact: Complementary information provided by Radiomics features could be in combination with clinical risk factors in order to assist clinicians in assessing liver fibrosis comprehensively. CoRC models exhibited promising diagnostic performances for clinically significant liver fibrosis, complementary to TE-LSM. 

Keywords: Diagnosis/Prediction, Biliary, Biomarkers; Intrahepatic cholangiocarcinoma; Magnetic resonance imaging; Tertiary lymphoid structures; Radiomics

Motivation: Tertiary lymphoid structures (TLSs) can only be assessed by postoperative specimen and a non-invasive tool to preoperatively evaluate TLSs is still lacking.

Goal(s): To explore the association between TLSs status of patients with ICC and preoperative magnetic resonance imaging (MRI) radiomics analysis.

Approach:  Radiomics features were subjected to LASSO regression to select the most associated features of TLSs and construct the radiomics model.  

Results: The AUCs of Rad-score were 0.85, 0.81, and 0.84 in the T, V1, and V2 cohorts, respectively. Low-risk group showed significantly better median RFS than that of the high-risk group, which was also confirmed in cohort V1 and V2.

Impact: TLSs have been reported to have prognostic value and guiding significance to immunotherapy in ICC patients. The MRI radiomics signature could serve as a non-invasive tool to preoperatively predict intra-tumoural TLSs status of ICC patients and correlate significantly with prognosis.

Keywords: Diagnosis/Prediction, Skeletal

Motivation: Increasing the identification accuracy of patients with high risk of progression could help guide treatment decision in soft tissue sarcoma (STS).

Goal(s): To establish a radiomics nomogram that incorporated tumor habitat and peritumor features and validate its performance to predict tumor progression in patients with STS.

Approach: A nomogram combining radiomics based on intratumoral habitat and peritumorwith clinical information was established.

Results: This nomogram predicts tumor progression in STS patients and stratifies them according to the risk of progression.

Impact: Combining radiomics features derived from the intratumoral habitat and peritumoral region resulted in superior performance for predicting progression-free-survival in patients with STS, which is helpful for clinical decision making.

Keywords: Diagnosis/Prediction, Bioeffects & Magnetic Fields

Motivation: Enhancing Placenta accreta spectrum (PAS) prediction through advanced MR T2WI radiomics, improving outcomes for both mothers and babies.

Goal(s): Improve PAS detection, enhance prenatal care, and reduce maternal and fetal risks.

Approach: Employing radiomics analysis on MR T2-weighted imaging, we establish a predictive model for PAS, augmenting prenatal diagnosis.

Results: The radiomics model exhibited exceptional accuracy and reliability, showcasing its potential for significantly enhancing PAS prediction in clinical practice.

Impact: The successful implementation of this study stands to significantly enhance the early identification capacity for PAS and offer robust support for clinical decision-making. Such an achievement carries immense practical significance in addressing the potential rise in PAS cases. 

Keywords: Diagnosis/Prediction, Radiomics, Breast neoplasms; Peritumoral; Neoplasm recurrence; Image interpretation

Motivation: 21-gene assay is recommend to guide decision on the use of adjuvant chemotherapy. However, this test is expensive and time-consuming, so it is not widely used in clinic. 

Goal(s): To develop a radiomics model for predicting 21-gene recurrence score based on MRI intratumoral and peritumoral features.

Approach: Prediction models in tumor and different peritumoral regions (2 mm, 4 mm, 6 mm, 8 mm, 10 mm) were established using machine learning method. Feature-fusion and logistics-regression methods were used to fuse information.

Results: Combining 4-mm peritumoral model on T2WI (AUC = 0.66) with intratumoral and clinical-imaging features, the fusion model performed best (AUC = 0.75).

Impact: To provide an alternative for patients who cannot afford Oncotype 21-gene assay and to reduce the medical costs for those who could afford it.

Keywords: Diagnosis/Prediction, Machine Learning/Artificial Intelligence

Motivation: The assessment of thyroid-associated ophthalmopathy (TAO) activity is crucial for determining the appropriate treatment. However, it was based on clinical activity score(CAS) which relies on subjective symptoms and judgment heavily.

Goal(s): The aim of this study was to establish an objective evaluating model for TAO’s activity based on MR Imaging-Derived Radiomics.

Approach: MR Imaging was performed for different activity status of TAO patients, after that radiomics features were extracted and selected. Radiomics models were constructed. 

Results: The radiomics model of T2 weighted SPIR-based extraocular muscles can achieve satisfactory performance than those of optic nerve or interorbital tissue in evaluating the activity of TAO.

Impact: This study constructed the T2 -weighted MR imaging-derived radiomics model to evaluate the activity of thyroid-associated ophthalmopathy, which could be a more objective and consistent evaluating approach than clinical activity score.

Keywords: Diagnosis/Prediction, Radiomics, Placenta

Motivation: Placenta accrete spectrum disorder (PAS) is a dangerous pregnancy complication that posed a threat to the safety of pregnant women, and its incidence is still on the rise.

Goal(s): To develop a machine learning model for effective diagnosis of PAS.

Approach: We developed machine learning models based on T2WI radiomics, clinical features, and clinical-radiomics features in the diagnosis of PAS.

Results: Radiomics models have a great diagnostic performance for PAS, with sagittal-based model shows better performance. The clinical-radiomics model exhibits the highest diagnostic performance in this study.

Impact: Machine learning models that combined with radiomics and clinical features can improve the diagnosis of PAS, and benefit PAS patients. Furthermore, our results provide new insights for future research.

Keywords: Diagnosis/Prediction, Breast

Motivation: As manual slice-by-slice analysis of breast MR images is both time-consuming and error-prone.

Goal(s): To develop a deep learning-based system for the detection and classification of breast lesions in DCE-MRI.

Approach: DCE-MRI images were fed into the developed cascade feature pyramid network system(CFPN), feature pyramid network, and faster region-based convolutional neural network for breast lesion detection and classification. 

Results: CFPN achieved the highest sensitivities in detection at the lowest FPs at both the slice level and the patient level. 

Impact: DL-based systems can automatically detect and classify breast lesions on DCE-MRI. These results illustrate the potential use of this technique in a clinically relevant setting.

Keywords: Diagnosis/Prediction, Breast, Multiparametric magnetic resonance imaging

Motivation: Multiparametric magnetic resonance imaging (mpMRI) offers valuable insights for predicting HER2 expression. However, when fusing mpMRI features, redundancy or wastage of information may impact model performance.

Goal(s): Our aim was to construct an effective deep learning model by incorporating the interrelated and complementary features of different MRI sequences.

Approach: Leveraging a contrastive learning approach, we aligned features across sequences and within each sequence separately to obtain sequence-shared and sequence-specific features. Subsequently, these two features were fused by utilizing an adaptive weighting scheme.

Results: When compared to widely used deep learning approaches, our method achieved the best AUC of 0.743.

Impact: The method explored the interrelated and complementary features of different MRI sequences, which outperformed widely used deep learning methods in terms of performance. This method was expected to have a positive impact on the accurate prediction of HER2 expression status.

Keywords: Diagnosis/Prediction, Breast

Motivation: Breast cancer (BrCa) is the most prevalent malignancy among women. MRI is a useful tool for BrCa early detection and characterization. However, high false-positive rates can lead to unnecessary biopsies and patient distress. To enhance diagnostic accuracy, deep learning presents a promising avenue, but training deep neural networks (DNN) requires a large, annotated dataset.

Goal(s): Introduce a novel method for BrCa classification, utilizing a minimally labeled dataset.

Approach:  We employ a few-shot learning (FSL) approach to differentiate between benign and malignant breast tumors.

Results: Our FSL-based model significantly surpasses the diagnostic performance of trained radiologists in breast cancer classification (p < 0.0001).

Impact: Our FSL model streamlines machine learning by reducing data labeling needs outperforms radiologists in detecting breast cancer, and could reduce unnecessary biopsies, sparing patients from potential harm.

Keywords: Diagnosis/Prediction, Breast

Motivation: ALN status is crucial for clinical staging, prognosis assessment, and treatment decision for breast cancer patients.

Goal(s): We aimed to assess feasibility of ML based on mpMRI for predicting the risk of NSLN metastasis in breast cancer patients.

Approach: mpMRI including T1WI, T2WI, DWI, and DCE-MRI was acquired, and four ML models were constructed.

Results: ML model incorporating mpMRI features and clinical factors can predict NSLN metastasis with fair accuracy for breast cancer, with an AUC of 0.781 in test dataset. Five factors for NSLN metastasis were found, including histological grade, cortical thickness, fatty hilum, short axis of lymph node, and age.

Impact: The proposed ML model may benefit for breast cancer patients with 1-2 positive SLN but consistently negative NSLN to avoid overtreatment and improve individualized axillary management.

Keywords: Diagnosis/Prediction, Machine Learning/Artificial Intelligence

Motivation: Fluid and fat accumulation can be observed in MRI images of patients with PLEL; however, the microscopic characteristics of the different components of PLEL are currently unknown.

Goal(s): This study aimed to explore the MRI radiomics features of different components of subcutaneous soft tissues in patients with PLEL, such as simple fat, mixed fat and water, fat interstitial edema, and effusion

Approach: We propose a machine learning model to analyze the radiomics characteristics of different tissue components of lower extremity lymphedema in MRI.

Results: he four-class model, using 15 selected radiomics features, shows outstanding performance with an overall accuracy of 0.866.

Impact: The different components of subcutaneous soft tissues of PLEL patients, such as simple fat, mixed fat and water, adipose interstitial edema and effusion, have unique radiomic features.

Keywords: Diagnosis/Prediction, Liver, deep learning，small hepatocellular carcinomas，Dysplastic Nodule

Motivation: In light of the overlapping image features between small hepatocellular carcinoma (sHCC) and benign precancerous nodules, the detection of sHCC from cirrhosis liver is deemed difficult and challenging.

Goal(s): To develop a fully automatic deep learning approach for the detection of sHCC in cirrhotic livers, utilizing Gd-EOB-DTPA-enhanced MRI.

Approach: A 3D nnU-Net deep learning network was trained to perform automatic segmentation and detection of sHCC lesions.

Results: 120 patients were included. The AUCs for discriminating between sHCC  lesions and non-sHCC  lesions were 0.967 and 0.864 in the training and test cohorts,, with both P<0.001.

Impact: Deep learning holds promise for the noninvasive detection of sHCC, offering the potential to alleviate the workload of radiologists and mitigate the necessity for biopsies along with their associated complications.