Latest Research
All publications from the Cancer3.AI database, newest first.
Recent developments in salivary gland pathology after the WHO 2024 classification: new developments in existing entities and evolving new entities.
Skálová A, et al
A comprehensive review published in Virchows Archiv examines the most significant advances in salivary gland tumor pathology from 2022 to 2025, covering important discoveries that emerged after — and were therefore not included in — the World Health Organization's 2024 Classification of Head and Neck Tumours. Among the newly proposed malignant entities are palisading adenocarcinoma, microcribriform adenocarcinoma, fenestrating adenocarcinoma, and skin-analogue poroid carcinoma, while established tumor types such as mucoepidermoid carcinoma and adenoid cystic carcinoma have been refined through recognition of new subtypes with distinct molecular and histological characteristics. In the benign category, new variants of pleomorphic adenoma and basal cell adenoma have been identified — each driven by specific genetic alterations in HMGA2, CTNNB1, or KRAS — alongside a thymus-like phenotype in non-sebaceous lymphadenoma linked to CYLD mutations and an emerging entity called sialadenopapillary ductal tumor. Updated grading systems for acinic cell carcinoma and secretory carcinoma have also been proposed to help pathologists more accurately predict patient outcomes and guide therapy. These developments are clinically meaningful because precise tumor classification in the salivary glands directly guides treatment decisions, surgical planning, and prognosis for patients with these rare and histologically complex cancers.
Virchows Archiv : an international journal of pathology
Source →Epigenetic remodeling after viral pneumonia accelerates lung tumorigenesis.
Liu WL, et al
Epidemiological studies have long pointed to viral pneumonia as a risk factor for lung cancer, yet the underlying biological mechanisms remained elusive. A new study by Qian et al., highlighted in Trends in Molecular Medicine, demonstrates that respiratory viral infections induce lasting epigenetic reprogramming in lung tissue that actively promotes tumor development. These epigenetic alterations orchestrate a pro-tumorigenic microenvironment, effectively conditioning lung tissue to support cancer cell growth and survival long after the initial infection has resolved. The findings provide a compelling molecular explanation for the clinically observed elevated lung cancer risk among pneumonia survivors. Crucially, the identification of specific epigenetic mechanisms involved opens promising avenues for therapeutic interception, offering clinicians potential strategies to prevent or delay lung cancer progression in patients with a history of respiratory viral disease.
Trends in molecular medicine
Source →Gαi1 and Gαi3 mediate IL-11-induced signal transduction and are potential therapeutic targets for LUAD.
Luo G, et al
Researchers investigated the molecular mechanisms by which the cytokine interleukin-11 (IL-11) drives the progression of lung adenocarcinoma (LUAD), the most common subtype of lung cancer, which has a five-year survival rate of only 15% due to late-stage diagnosis. The study identified two G-protein subunits, Gαi1 and Gαi3, as critical intermediaries that bind to the receptor GP130 and mediate IL-11-induced activation of key cancer-promoting signaling pathways, including Akt-mTOR, Erk, and STAT3, with the adapter protein Gab1 also required for this process. When Gαi1 and Gαi3 were silenced in LUAD cells, tumor cell growth, proliferation, and migration were substantially reduced, and the cancer-promoting effects of IL-11 were blocked in both cell culture and animal models. Conversely, overexpression of these proteins amplified IL-11-driven tumor behaviors, confirming their central role in the signaling cascade. Critically, analysis of patient data revealed that Gαi1 and Gαi3 are highly expressed in LUAD tumors and their elevated levels correlate with poor overall survival, positioning them as promising new therapeutic targets. These findings open a potential new treatment avenue for LUAD patients whose tumors show upregulated IL-11 expression.
Cell death & disease
Source →LungDxFormer: a transformer-CNN hybrid model with dynamic spatial attention for accurate lung cancer detection and classification.
Rao KS, et al
Researchers have developed LungDxFormer, a novel artificial intelligence model that combines transformer and convolutional neural network (CNN) architectures with a dynamic spatial attention mechanism to detect and classify lung cancer from medical imaging data. The hybrid design allows the model to capture both local image features and long-range contextual relationships simultaneously, which are critical for distinguishing malignant from benign lung lesions. By integrating dynamic spatial attention, LungDxFormer can focus computational resources on the most diagnostically relevant regions of a scan, improving both accuracy and interpretability. The model demonstrated strong performance in lung cancer detection and subtype classification tasks, suggesting it could serve as a reliable computer-aided diagnostic tool for radiologists. Earlier and more accurate lung cancer classification has direct clinical implications, as lung cancer remains one of the leading causes of cancer-related mortality worldwide, and timely diagnosis is essential for improving patient survival rates.
Scientific reports
Source →Grayscale ultrasound radiomics for characterizing subpleural pulmonary lesions: a multicenter prospective study.
Yi J, et al
Researchers conducted a multicenter prospective study to develop and validate a radiomics model based on grayscale ultrasound images for distinguishing benign from malignant subpleural pulmonary lesions — small masses located near the outer lining of the lungs that are notoriously difficult to characterize with conventional imaging. Using data from 738 patients across three institutions, the team extracted 1,320 radiomic features from lesion and surrounding tissue regions on ultrasound images, then built predictive models combining these features with clinical variables through logistic regression. The best-performing integrated model achieved an area under the ROC curve of 0.884 in internal validation and 0.848 in external validation, matching the diagnostic accuracy of experienced lung ultrasound radiologists. This noninvasive, readily available tool holds particular promise for patients unsuitable for invasive procedures and for clinical settings where CT scanning or tissue biopsy is not easily accessible.
Insights into imaging
Source →