Background: Identifying breast cancer (BC) patients with germline breast cancer susceptibility gene (gBRCA) mutation is important. The current criteria for germline testing for BC remain controversial. This study aimed to develop a nomogram incorporating ultrasound radiomic features and clinicopathological factors to predict gBRCA mutations in patients with BC.
Materials and methods: In this retrospective study, 497 women with BC who underwent gBRCA genetic testing from March 2013 to May 2022 were included, including 348 for training (84 with and 264 without a gBRCA mutation) and 149 for validation(36 patients with and 113 without a gBRCA mutation). Factors associated with gBRCA mutations were identified to establish a clinicopathological model. Radiomics features were extracted from the intratumoral and peritumoral regions (3 mm and 5 mm) of each image. The least absolute shrinkage and selection operator regression algorithm was used to select the features and logistic regression analysis was used to construct three imaging models. Finally, a nomogram that combined clinicopathological and radiomics features was developed. The models were evaluated based on the area under the receiver operating characteristic curve (AUC), calibration, and clinical usefulness.
Results: Age at diagnosis, family history of BC, personal history of other BRCA-related cancers, and human epidermal growth factor receptor 2 status were independent predictors of the clinicopathological model. The AUC of the imaging radiomics model combining intratumoral and peritumoral 3 mm areas in the validation set was 0.783 (95% confidence interval [CI]: 0.702-0.862), which showed the best performance among three imaging models. The nomogram yielded better performance than the clinicopathological model in validation sets (AUC: 0.824 [0.755-0.894] versus 0.659 [0.563-0.755], p = 0.007).
Conclusion: The nomogram based on ultrasound images and clinicopathological factors performs well in predicting gBRCA mutations in BC patients and may help to improve clinical decisions about genetic testing.
Background: Prostate-specific membrane antigen (PSMA) PET/CT imaging is widely used for quantitative image analysis, especially in radioligand therapy (RLT) for metastatic castration-resistant prostate cancer (mCRPC). Unknown features influencing PSMA biodistribution can be explored by analyzing segmented organs at risk (OAR) and lesions. Manual segmentation is time-consuming and labor-intensive, so automated segmentation methods are desirable. Training deep-learning segmentation models is challenging due to the scarcity of high-quality annotated images. Addressing this, we developed shifted windows UNEt TRansformers (Swin UNETR) for fully automated segmentation. Within a self-supervised framework, the model's encoder was pre-trained on unlabeled data. The entire model was fine-tuned, including its decoder, using labeled data.
Methods: In this work, 752 whole-body [68Ga]Ga-PSMA-11 PET/CT images were collected from two centers. For self-supervised model pre-training, 652 unlabeled images were employed. The remaining 100 images were manually labeled for supervised training. In the supervised training phase, 5-fold cross-validation was used with 64 images for model training and 16 for validation, from one center. For testing, 20 hold-out images, evenly distributed between two centers, were used. Image segmentation and quantification metrics were evaluated on the test set compared to the ground-truth segmentation conducted by a nuclear medicine physician.
Results: The model generates high-quality OARs and lesion segmentation in lesion-positive cases, including mCRPC. The results show that self-supervised pre-training significantly improved the average dice similarity coefficient (DSC) for all classes by about 3%. Compared to nnU-Net, a well-established model in medical image segmentation, our approach outperformed with a 5% higher DSC. This improvement was attributed to our model's combined use of self-supervised pre-training and supervised fine-tuning, specifically when applied to PET/CT input. Our best model had the lowest DSC for lesions at 0.68 and the highest for liver at 0.95.
Conclusions: We developed a state-of-the-art neural network using self-supervised pre-training on whole-body [68Ga]Ga-PSMA-11 PET/CT images, followed by fine-tuning on a limited set of annotated images. The model generates high-quality OARs and lesion segmentation for PSMA image analysis. The generalizable model holds potential for various clinical applications, including enhanced RLT and patient-specific internal dosimetry.
Objective: To investigate the diagnostic value of diffusion kurtosis magnetic resonance imaging (DKI) and conventional diffusion-weighted imaging (DWI) for evaluating the response to first-line chemotherapy in unresectable pancreatic cancer.
Materials and methods: We retrospectively analyzed 21 patients with clinically and pathologically confirmed unresected pancreatic cancer who received palliative chemotherapy. Three-tesla MRI examinations containing DWI sequences with b values of 0, 100, 700, 1400, and 2100 s/mm2 were performed before and after chemotherapy. Parameters included the apparent diffusion coefficient (ADC), mean diffusion coefficient (MD), and mean diffusional kurtosis (MK). The performances of the DWI and DKI parameters in distinguishing the response to chemotherapy were evaluated by the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Overall survival (OS) was calculated from the date of first treatment to the date of death or the latest follow-up date.
Results: The ADCchange and MDchange were significantly higher in the responding group (PR group) than in the nonresponding group (non-PR group) (ADCchange: 0.21 ± 0.05 vs. 0.11 ± 0.09, P = 0.02; MDchange: 0.37 ± 0.24 vs. 0.10 ± 0.12, P = 0.002). No statistical significance was shown when comparing ADCpre, ADCpost, MKpre, MKpost, MKchange, MDpre, and MDpost between the PR and non-PR groups. The ROC curve analysis indicated that MDchange (AUC = 0.898, cutoff value = 0.7143) performed better than ADCchange (AUC = 0.806, cutoff value = 0.1369) in predicting the response to chemotherapy.
Conclusion: The ADCchange and MDchange demonstrated strong potential for evaluating the response to chemotherapy in unresectable pancreatic cancer. The MDchange showed higher specificity in the classification of PR and non-PR than the ADCchange. Other parameters, including ADCpre, ADCpost, MKpre, MKpost, MKchange, MDpre, and MDpost, are not suitable for response evaluation. The combined model SUMchange demonstrated superior performance compared to the individual DWI and DKI models. Further experiments are needed to evaluate the potential of DWI and DKI parameters in predicting the prognosis of patients with unresectable pancreatic cancer.
Background: Surgically resected grade 1-2 (G1-2) pancreatic neuroendocrine tumors (PanNETs) exhibit diverse clinical outcomes, highlighting the need for reliable prognostic biomarkers. Our study aimed to develop and validate CT-based radiomics model for predicting postsurgical outcome in patients with G1-2 PanNETs, and to compare its performance with the current clinical staging system.
Methods: This multicenter retrospective study included patients who underwent dynamic CT and subsequent curative resection for G1-2 PanNETs. A radiomics-based model (R-score) for predicting recurrence-free survival (RFS) was developed from a development set (441 patients from one institution) using least absolute shrinkage and selection operator-Cox regression analysis. A clinical model (C-model) consisting of age and tumor stage according to the 8th American Joint Committee on Cancer staging system was built, and an integrative model combining the C-model and the R-score (CR-model) was developed using multivariable Cox regression analysis. Using an external test set (159 patients from another institution), the models' performance for predicting RFS and overall survival (OS) was evaluated using Harrell's C-index. The incremental value of adding the R-score to the C-model was evaluated using net reclassification improvement (NRI) and integrated discrimination improvement (IDI).
Results: The median follow-up periods were 68.3 and 59.7 months in the development and test sets, respectively. In the development set, 58 patients (13.2%) experienced recurrence and 35 (7.9%) died. In the test set, tumors recurred in 14 patients (8.8%) and 12 (7.5%) died. In the test set, the R-score had a C-index of 0.716 for RFS and 0.674 for OS. Compared with the C-model, the CR-model showed higher C-index (RFS, 0.734 vs. 0.662, p = 0.012; OS, 0.781 vs. 0.675, p = 0.043). CR-model also showed improved classification (NRI, 0.330, p < 0.001) and discrimination (IDI, 0.071, p < 0.001) for prediction of 3-year RFS.
Conclusions: Our CR-model outperformed the current clinical staging system in prediction of the prognosis for G1-2 PanNETs and added incremental value for predicting postoperative recurrence. The CR-model enables precise identification of high-risk patients, guiding personalized treatment planning to improve outcomes in surgically resected grade 1-2 PanNETs.
Background: The state-of-the-art method for imaging men with biochemical recurrence of prostate cancer (BCR) is prostate-specific membrane antigen (PSMA)-targeted positron emission tomography/computed tomography (PET/CT) with tracers containing short-lived radionuclides, e.g., gallium-68 (68Ga; half-life: ∼67.7 min). However, such imaging not infrequently yields indeterminate findings, which remain challenging to characterize. PSMA-targeted tracers labeled with zirconium-89 (89Zr; half-life: ∼78.41 h) permit later scanning, which may help in classifying the level of suspiciousness for prostate cancer of lesions previously indeterminate on conventional PSMA-targeted PET/CT.
Methods: To assess the ability of [89Zr]Zr-PSMA-617 PET/CT to characterize such lesions, we retrospectively analyzed altogether 20 lesions that were indeterminate on prior [68Ga]Ga-PSMA-11 PET/CT, in 15 men with BCR (median prostate-specific antigen: 0.70 ng/mL). The primary endpoint was the lesions' classifications, and secondary endpoints included [89Zr]Zr-PSMA-617 uptake (maximum standardized uptake value [SUVmax]), and lesion-to-background ratio (tumor-to-liver ratio of the SUVmax [TLR]). [89Zr]Zr-PSMA-617 scans were performed 1 h, 24 h, and 48 h post-injection of 123 ± 19 MBq of radiotracer, 35 ± 35 d post-[68Ga]Ga-PSMA-11 PET/CT.
Results: Altogether, 6/20 previously-indeterminate lesions (30%) were classified as suspicious (positive) for prostate cancer, 14/20 (70%), as non-suspicious (negative). In these two categories, [89Zr]Zr-PSMA-617 uptake and lesional contrast showed distinctly different patterns. In positive lesions, SUVmax and TLR markedly rose from 1 to 48 h, with SUVmax essentially plateauing at high levels, and TLR further steeply increasing, from 24 to 48 h. In negative lesions, uptake, when present, was very low, and decreasing, while contrast was minimal, from 1 to 48 h. No adverse events or clinically-relevant vital signs changes related to [89Zr]Zr-PSMA-617 PET/CT were noted during or ~ 4 weeks after the procedure.
Conclusions: In men with BCR, [89Zr]Zr-PSMA-617 PET/CT may help characterize as suspicious or non-suspicious for prostate cancer lesions that were previously indeterminate on [68Ga]Ga-PSMA-11 PET/CT.
Trial registration: Not applicable.
Background: To investigate the association between Kirsten rat sarcoma viral oncogene homolog (KRAS) / neuroblastoma rat sarcoma viral oncogene homolog (NRAS) /v-raf murine sarcoma viral oncogene homolog B (BRAF) mutations and the tumor habitat-derived radiomic features obtained during pretreatment 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) in patients with colorectal cancer (CRC).
Methods: We retrospectively enrolled 62 patients with CRC who had undergone 18F-FDG PET/computed tomography from January 2017 to July 2022 before the initiation of therapy. The patients were randomly split into training and validation cohorts with a ratio of 6:4. The whole tumor region radiomic features, habitat-derived radiomic features, and metabolic parameters were extracted from 18F-FDG PET images. After reducing the feature dimension and selecting meaningful features, we constructed a hierarchical model of KRAS/NRAS/BRAF mutations by using the support vector machine. The convergence of the model was evaluated by using learning curve, and its performance was assessed based on the area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis. The SHapley Additive exPlanation was used to interpret the contributions of various features to predictions of the model.
Results: The model constructed by using habitat-derived radiomic features had adequate predictive power with respect to KRAS/NRAS/BRAF mutations, with an AUC of 0.759 (95% CI: 0.585-0.909) on the training cohort and that of 0.701 (95% CI: 0.468-0.916) on the validation cohort. The model exhibited good convergence, suitable calibration, and clinical application value. The results of the SHapley Additive explanation showed that the peritumoral habitat and a high_metabolism habitat had the greatest impact on predictions of the model. No meaningful whole tumor region radiomic features or metabolic parameters were retained during feature selection.
Conclusion: The habitat-derived radiomic features were found to be helpful in stratifying the status of KRAS/NRAS/BRAF in CRC patients. The approach proposed here has significant implications for adjuvant treatment decisions in patients with CRC, and needs to be further validated on a larger prospective cohort.
Objectives: Tumor spread through air spaces (STAS) is associated with poor prognosis and impacts surgical options. We aimed to develop a user-friendly model based on 2-[18F] FDG PET/CT to predict STAS in stage I lung adenocarcinoma (LAC).
Materials and methods: A total of 466 stage I LAC patients who underwent 2-[18F] FDG PET/CT examination and resection surgery were retrospectively enrolled. They were split into a training cohort (n = 232, 20.3% STAS-positive), a validation cohort (n = 122, 27.0% STAS-positive), and a test cohort (n = 112, 29.5% STAS-positive) according to chronological order. Some commonly used clinical data, visualized CT features, and SUVmax were analyzed to identify independent predictors of STAS. A prediction model was built using the independent predictors and validated using the three chronologically separated cohorts. Model performance was assessed using ROC curves and calculations of AUC.
Results: The differences in age (P = 0.009), lesion density subtype (P < 0.001), spiculation sign (P < 0.001), bronchus truncation sign (P = 0.001), and SUVmax (P < 0.001) between the positive and negative groups were statistically significant. Age ≥ 56 years [OR(95%CI):3.310(1.150-9.530), P = 0.027], lesion density subtype (P = 0.004) and SUVmax ≥ 2.5 g/ml [OR(95%CI):3.268(1.021-1.356), P = 0.005] were the independent factors predicting STAS. Logistic regression was used to build the A-D-S (Age-Density-SUVmax) prediction model, and the AUCs were 0.808, 0.786 and 0.806 in the training, validation, and test cohorts, respectively.
Conclusions: STAS was more likely to occur in older patients, in solid lesions and higher SUVmax in stage I LAC. The PET/CT-based A-D-S prediction model is easy to use and has a high level of reliability in diagnosing.
Background: To build machine learning predictive models for surgical risk assessment of extracapsular extension (ECE) in patients with prostate cancer (PCa) before radical prostatectomy; and to compare the use of decision curve analysis (DCA) and receiver operating characteristic (ROC) metrics for selecting input feature combinations in models.
Methods: This retrospective observational study included two independent data sets: 139 participants from a single institution (training), and 55 from 15 other institutions (external validation), both treated with Robotic Assisted Radical Prostatectomy (RARP). Five ML models, based on different combinations of clinical, semantic (interpreted by a radiologist) and radiomics features computed from T2W-MRI images, were built to predict extracapsular extension in the prostatectomy specimen (pECE+). DCA plots were used to rank the models' net benefit when assigning patients to prostatectomy with non-nerve-sparing surgery (NNSS) or nerve-sparing surgery (NSS), depending on the predicted ECE status. DCA model rankings were compared with those drived from ROC area under the curve (AUC).
Results: In the training data, the model using clinical, semantic, and radiomics features gave the highest net benefit values across relevant threshold probabilities, and similar decision curve was observed in the external validation data. The model ranking using the AUC was different in the discovery group and favoured the model using clinical + semantic features only.
Conclusions: The combined model based on clinical, semantic and radiomic features may be used to predict pECE + in patients with PCa and results in a positive net benefit when used to choose between prostatectomy with NNS or NNSS.
Background: Whole-body magnetic resonance imaging (WB-MRI) has emerged as a valuable tool for cancer detection. This study evaluated the prevalence rates of cancer in asymptomatic individuals undergoing WB-MRI according to the Oncologically Relevant Findings Reporting and Data System (ONCO-RADS) classifications in order to assess the reliability of the classification method.
Methods: We retrospectively enrolled 2064 asymptomatic individuals who participated in a WB-MRI cancer screening program between 2017 and 2022. WB-MRI was acquired on a 3-T system with a standard protocol, including regional multisequence and gadolinium-based contrast agent-enhanced oncologic MRI. Results of further examinations, including additional imaging and histopathology examinations, performed at our institute were used to validate the WB-MRI findings. Two radiologists blinded to the clinical outcome classified the WB-MRI findings according to the ONCO-RADS categories as follows: 1 (normal), 2 (benign finding highly likely), 3 (benign finding likely), 4 (malignant finding likely), and 5 (malignant finding highly likely). Firth logistic regression analysis was performed to determine the associations between participant characteristics and findings of ONCO-RADS category ≥ 4.
Results: Of the 2064 participants with median age of 55 years, 1120 (54.3%) were men, 43 (2.1%) had findings of ONCO-RADS category ≥ 4, and 24 (1.2%) had confirmed cancer. The cancer prevalence rates were 0.1%, 5.4%, 42.9%, and 75% for ONCO-RADS categories 2, 3, 4, and 5, respectively. In the multivariable model, older age (OR: 1.035, p = 0.029) and history of hypertension (OR: 2.051, p = 0.026), hepatitis B carrier (OR: 2.584, p = 0.013), or prior surgery (OR: 3.787, p < 0.001) were independently associated with the findings for ONCO-RADS category ≥ 4.
Conclusions: The ONCO-RADS categories for cancer risk stratification were validated and found to be positively correlated with cancer risk. The application of ONCO-RADS facilitates risk-based management after WB-MRI for cancer screening.