Deep-learning-based radiomics of intratumoral and peritumoral MRI images to predict the pathological features of adjuvant radiotherapy in early-stage cervical squamous cell carcinoma.

BACKGROUND: Surgery combined with radiotherapy substantially escalates the likelihood of encountering complications in early-stage cervical squamous cell carcinoma(ESCSCC). We aimed to investigate the feasibility of Deep-learning-based radiomics of intratumoral and peritumoral MRI images to predict the pathological features of adjuvant radiotherapy in ESCSCC and minimize the occurrence of adverse events associated with the treatment. METHODS: A dataset comprising MR images was obtained from 289 patients who underwent radical hysterectomy and pelvic lymph node dissection between January 2019 and April 2022. The dataset was randomly divided into two cohorts in a 4:1 ratio.The postoperative radiotherapy options were evaluated according to the Peter/Sedlis standard. We extracted clinical features, as well as intratumoral and peritumoral radiomic features, using the least absolute shrinkage and selection operator (LASSO) regression. We constructed the Clinical Signature (Clinic_Sig), Radiomics Signature (Rad_Sig) and the Deep Transformer Learning Signature (DTL_Sig). Additionally, we fused the Rad_Sig with the DTL_Sig to create the Deep Learning Radiomic Signature (DLR_Sig). We evaluated the prediction performance of the models using the Area Under the Curve (AUC), calibration curve, and Decision Curve Analysis (DCA). RESULTS: The DLR_Sig showed a high level of accuracy and predictive capability, as demonstrated by the area under the curve (AUC) of 0.98(95% CI: 0.97-0.99) for the training cohort and 0.79(95% CI: 0.67-0.90) for the test cohort. In addition, the Hosmer-Lemeshow test, which provided p-values of 0.87 for the training cohort and 0.15 for the test cohort, respectively, indicated a good fit. DeLong test showed that the predictive effectiveness of DLR_Sig was significantly better than that of the Clinic_Sig(P < 0.05 both the training and test cohorts). The calibration plot of DLR_Sig indicated excellent consistency between the actual and predicted probabilities, while the DCA curve demonstrating greater clinical utility for predicting the pathological features for adjuvant radiotherapy. CONCLUSION: DLR_Sig based on intratumoral and peritumoral MRI images has the potential to preoperatively predict the pathological features of adjuvant radiotherapy in early-stage cervical squamous cell carcinoma (ESCSCC).

Subphrenic Lymph Node Metastasis Predicts Poorer Prognosis for Nasopharyngeal Carcinoma Patients With Metachronous Metastasis.

AIM: We retrospectively analyzed the distribution of distant lymph node metastasis and its impact on prognosis in patients with metastatic NPC after treatment. METHODS: From 2010 to 2016, 219 NPC patients out of 1,601 (182 from the Affiliated Cancer Hospital and Institute of Guangzhou Medical University, and 37 from the Affiliated Dongguan Hospital, Southern Medical University) developed distant metastasis after primary radiation therapy. Metastatic lesions were divided into groups according to location: bones above the diaphragm (supraphrenic bone, SUP-B); bones below the diaphragm (subphrenic bone, SUB-B); distant lymph nodes above the diaphragm (supraphrenic distant lymph nodes, SUP-DLN); distant lymph nodes below the diaphragm (subphrenic distant lymph nodes, SUB-DLN), liver, lung, and other lesions beyond bone/lung/distant lymph node above the diaphragm (supraphrenic other lesions, SUP-OL); other lesions beyond bone/liver/distant lymph node below the diaphragm (subphrenic other lesions, SUB-OL); the subtotal above the diaphragm (supraphrenic total lesions, SUP-TL); and the subtotal below the diaphragm (subphrenic total lesions, SUB-TL). Kaplan-Meier methods were used to estimate the probability of patients' overall survival (OS). Univariate and multivariate analyses were applied using the Cox proportional hazard model to explore prediction factors of OS. RESULTS: The most frequent metastatic locations were bone (45.2%), lung (40.6%), liver (32.0%), and distant lymph nodes (20.1%). The total number of distant lymph node metastasis was 44, of which 22 (10.0%) were above the diaphragm, 18 (8.2%) were below the diaphragm, and 4 (1.8%) were both above and below the diaphragm. Age (HR: 1.02, 95% CI: 1.00, 1.03, p = 0.012), N stage (HR: 1.26, 95% CI: 1.04, 1.54, p = 0.019), number of metastatic locations (HR: 1.39, 95% CI: 1.12, 1.73, p = 0.003), bone (HR: 1.65, 95% CI: 1.20, 2.25, p = 0.002), SUB-B (HR: 1.51, 95% CI: 1.07, 2.12, p = 0.019), SUB-DLN (HR: 1.72, 95% CI: 1.03, 2.86, p = 0.038), and SUB-O L(HR: 4.46, 95% CI: 1.39, 14.3, p = 0.012) were associated with OS. Multivariate analyses revealed that a higher N stage (HR: 1.23, 95% CI: 1.00, 1.50, p = 0.048), SUB-DLN (HR: 1.72, 95% CI: 1.02, 2.90, p = 0.043), and SUB-OL (HR: 3.72, 95% CI: 1.14, 12.16, p = 0.029) were associated with worse OS. CONCLUSION: Subphrenic lymph node metastasis predicts poorer prognosis for NPC patients with metachronous metastasis; however, this needs validation by large prospective studies.

Cocluster analysis of thalamo-cortical fibre tracts extracted from diffusion tensor MRI.

As the central relay station of the human brain, the thalamus modulates sensory signals to and from the cerebral cortex. The reciprocal connectivity between the cerebral cortex and the thalamus is believed to play an essential role in consciousness and various neurological disorders. Thus, in-vivo analysis of thalamo-cortical connectivity is important for our understanding of normal and pathological brain processes. In this paper: We propose a new partitioning paradigm, called coclustering, in order to segment the thalamus into thalamic nuclei based on their cortical projections. In contrast to the traditional clustering paradigm, a coclustering procedure not only simultaneously partitions cortical voxels and thalamic voxels into groups, but also identifies the corresponding strong connectivities between the two classes of groups. We develop the first coclustering algorithm, Genetic Coclustering Algorithm (GCA), to solve the coclustering problem. We apply GCA to segment the thalamus into thalamic nuclei and visualise main thalamo-cortical fibre tracts.