Deep learning applications are emerging as promising new tools that can support the diagnosis and classification of different cancer types. While such solutions hold great potential for hematological malignancies, there have been limited studies describing the use of such applications in this field. The rapid diagnosis of double/triple-hit lymphomas (DHLs/THLs) involving MYC, BCL2 and/or BCL6 rearrangements is obligatory for optimal patient care. Here, we present a novel deep learning tool for diagnosing DHLs/THLs directly from scanned images of biopsy slides. A total of 57 biopsies, including 32 in a training set (including five DH lymphoma cases) and 25 in a validation set (including 10 DH/TH cases), were included. The DHL-classifier demonstrated a sensitivity of 100%, a specificity of 87% and an AUC of 0.95, with only two false positive cases, compared to FISH. The DHL-classifier showed a 92% predictive value as a screening tool for performing conventional FISH analysis, over-performing currently used criteria. The work presented here provides the proof of concept for the potential use of an AI tool for the identification of DH/TH events. However, more extensive follow-up studies are required to assess the robustness of this tool and achieve high performances in a diverse population.

Histological evaluation plays a major role in cancer diagnosis and treatment. The appearance of H&E-stained images can vary significantly as a consequence of differences in several factors, such as reagents, staining conditions, preparation procedure and image acquisition system. Such potential sources of noise can all have negative effects on computer-assisted classification. To minimize such artefacts and their potentially negative effects several color pre-processing methods have been proposed in the literature-for instance, color augmentation, color constancy, color deconvolution and color transfer. Still, little work has been done to investigate the efficacy of these methods on a quantitative basis. In this paper, we evaluated the effects of color constancy, deconvolution and transfer on automated classification of H&E-stained images representing different types of cancers-specifically breast, prostate, colorectal cancer and malignant lymphoma. Our results indicate that in most cases color pre-processing does not improve the classification accuracy, especially when coupled with color-based image descriptors. Some pre-processing methods, however, can be beneficial when used with some texture-based methods like Gabor filters and Local Binary Patterns.

We introduce a machine learning-based analysis to predict the immunohistochemical (IHC) labeling index for the cell proliferation marker Ki67/MIB1 on cancer tissues based on morphometrical features extracted from hematoxylin and eosin (H&E)-stained formalin-fixed, paraffin-embedded (FFPE) tumor tissue samples. We provided a proof-of-concept prediction of the Ki67/MIB1 IHC positivity of cancer cells through the definition and quantitation of single nuclear features. In the first instance, we set our digital framework on Ki67/MIB1-stained OSCC (oral squamous cell carcinoma) tissue sample whole slide images, using QuPath as a working platform and its integrated algorithms, and we built a classifier in order to distinguish tumor and stroma classes and, within them, Ki67-positive and Ki67-negative cells; then, we sorted the morphometric features of tumor cells related to their Ki67 IHC status. Among the evaluated features, nuclear hematoxylin mean optical density (NHMOD) presented as the best one to distinguish Ki67/MIB1 positive from negative cells. We confirmed our findings in a single-cell level analysis of H&E staining on Ki67-immunostained/H&E-decolored tissue samples. Finally, we tested our digital framework on a case series of oral squamous cell carcinomas (OSCC), arranged in tissue microarrays; we selected two consecutive sections of each OSCC FFPE TMA (tissue microarray) block, respectively stained with H&E and immuno-stained for Ki67/MIB1. We automatically detected tumor cells in H&E slides and generated a "false color map" (FCM) based on NHMOD through the QuPath measurements map tool. FCM nearly coincided with the actual immunohistochemical result, allowing the prediction of Ki67/MIB1 positive cells in a direct visual fashion. Our proposed approach provides the pathologist with a fast method of identifying the proliferating compartment of the tumor through a quantitative assessment of the nuclear features on H&E slides, readily appreciable by visual inspection. Although this technique needs to be fine-tuned and tested on larger series of tumors, the digital analysis approach appears to be a promising tool to quickly forecast the tumor's proliferation fraction directly on routinely H&E-stained digital sections.

Reliable cytopathological diagnosis requires new methods and approaches for the rapid and accurate determination of all cell types. This is especially important when the number of cells is limited, such as in the cytological samples of fine-needle biopsy. Immunoplasmonic-multiplexed- labeling may be one of the emerging solutions to such problems. However, to be accepted and used by the practicing pathologists, new methods must be compatible and complementary with existing cytopathology approaches where counterstaining is central to the correct interpretation of immunolabeling. In addition, the optical detection and imaging setup for immunoplasmonic-multiplexed-labeling must be implemented on the same cytopathological microscope, not interfere with standard H&E imaging, and operate as a second easy-to-use imaging method. In this article, we present multiplex imaging of four types of nanoplasmonic markers on two types of H&E-stained cytological specimens (formalin-fixed paraffin embedded and non-embedded adherent cancer cells) using a specially designed adapter for SI dark-field microscopy. The obtained results confirm the effectiveness of the proposed optical method for quantitative and multiplex identification of various plasmonic NPs, and the possibility of using immunoplasmonic-multiplexed-labeling for cytopathological diagnostics.

Breast stroma plays a significant role in breast cancer risk and progression yet remains poorly understood. In breast stroma, collagen is the most abundantly expressed protein and its increased deposition and alignment contributes to progression and poor prognosis. Collagen post-translation modifications such as hydroxylated-proline (HYP) control deposition and stromal organization. The clinical relevance of collagen HYP site modifications in cancer processes remains undefined due to technical issues accessing collagen from formalin-fixed, paraffin-embedded (FFPE) tissues. We previously developed a targeted approach for investigating collagen and other extracellular matrix proteins from FFPE tissue. Here, we hypothesized that immunohistochemistry staining for fibroblastic markers would not interfere with targeted detection of collagen stroma peptides and could reveal peptide regulation influenced by specific cell types. Our initial work demonstrated that stromal peptide peak intensities when using MALD-IMS following IHC staining (αSMA, FAP, P4HA3 and PTEN) were comparable to serial sections of nonstained tissue. Analysis of histology-directed IMS using PTEN on breast tissues and TMAs revealed heterogeneous PTEN staining patterns and suggestive roles in stromal protein regulation. This study sets the foundation for investigations of target cell types and their unique contribution to collagen regulation within extracellular matrix niches.

Tumor markers that can be detected at an early stage are needed. Here, we evaluated the epiplakin expression levels in sera from patients with bladder cancer (BC). Using a micro-dot blot array, we evaluated epiplakin expression levels in 60 patients with BC, 20 patients with stone disease, and 28 healthy volunteers. The area under the curve (AUC) and best cut-off point were calculated using receiver-operating characteristic (ROC) analysis. Serum epiplakin levels were significantly higher in patients with BC than in those with stone disease (p = 0.0013) and in healthy volunteers (p < 0.0001). The AUC-ROC level for BC was 0.78 (95% confidence interval (CI) = 0.69-0.87). Using a cut-off point of 873, epiplakin expression levels exhibited 68.3% sensitivity and 79.2% specificity for BC. However, the serum epiplakin levels did not significantly differ by sex, age, pathological stage and grade, or urine cytology. We performed immunohistochemical staining using the same antibody on another cohort of 127 patients who underwent radical cystectomy. Univariate and multivariate analysis results showed no significant differences between epiplakin expression, clinicopathological findings, and patient prognoses. Our results showed that serum epiplakin might be a potential serodiagnostic biomarker in patients with BC.

Fluorescence-guided surgery is a state-of-the-art approach for intraoperative imaging during neurosurgical removal of tumor tissue. While the visualization of high-grade gliomas is reliable, lower grade glioma often lack visible fluorescence signals. Here, we present a hybrid prototype combining visible light optical coherence microscopy (OCM) and high-resolution fluorescence imaging for assessment of brain tumor samples acquired by 5-aminolevulinic acid (5-ALA) fluorescence-guided surgery. OCM provides high-resolution information of the inherent tissue scattering and absorption properties of tissue. We here explore quantitative attenuation coefficients derived from volumetric OCM intensity data and quantitative high-resolution 5-ALA fluorescence as potential biomarkers for tissue malignancy including otherwise difficult-to-assess low-grade glioma. We validate our findings against the gold standard histology and use attenuation and fluorescence intensity measures to differentiate between tumor core, infiltrative zone and adjacent brain tissue. Using large field-of-view scans acquired by a near-infrared swept-source optical coherence tomography setup, we provide initial assessments of tumor heterogeneity. Finally, we use cross-sectional OCM images to train a convolutional neural network that discriminates tumor from non-tumor tissue with an accuracy of 97%. Collectively, the present hybrid approach offers potential to translate into an in vivo imaging setup for substantially improved intraoperative guidance of brain tumor surgeries.

Few studies have examined the role of BAG2 in malignancies. We investigated the prognostic value of BAG2-expression in cancer-associated fibroblasts (CAFs) and tumor cells in predicting metastasis-free survival in patients with breast cancer. Tissue-microarray was constructed using human breast cancer tissues obtained by surgical resection between 1992 and 2015. BAG2 expression was evaluated by immunohistochemistry in CAFs or the tumor cells. BAG2 expression in the CAFs and cytoplasm of tumor cells was classified as positive and negative, and low and high, respectively. BAG2-CAF was evaluated in 310 patients and was positive in 67 (21.6%) patients. Kaplan-Meier plots showed that distant metastasis-free survival (DMFS) was lesser in patients with BAG2(+) CAF than in patients with BAG2(-) CAF (p = 0.039). Additionally, we classified the 310 patients into two groups: 109 in either BAG2-high or BAG2(+) CAF and 201 in BAG2-low and BAG2(-) CAF. DMFS was significantly reduced in patients with either BAG2-high or BAG2(+) CAF than in the patients of the other group (p = 0.005). Multivariable analysis demonstrated that DMFS was prolonged in patients with BAG2(-) CAF or BAG2-low. Evaluation of BAG2 expression on both CAFs and tumor cells could help in determining the risk of metastasis in breast cancer.

Poor generalizability is a major barrier to clinical implementation of artificial intelligence in digital pathology. The aim of this study was to test the generalizability of a pretrained deep learning model to a new diagnostic setting and to a small change in surgical indication. A deep learning model for breast cancer metastases detection in sentinel lymph nodes, trained on CAMELYON multicenter data, was used as a base model, and achieved an AUC of 0.969 (95% CI 0.926-0.998) and FROC of 0.838 (95% CI 0.757-0.913) on CAMELYON16 test data. On local sentinel node data, the base model performance dropped to AUC 0.929 (95% CI 0.800-0.998) and FROC 0.744 (95% CI 0.566-0.912). On data with a change in surgical indication (axillary dissections) the base model performance indicated an even larger drop with a FROC of 0.503 (95%CI 0.201-0.911). The model was retrained with addition of local data, resulting in about a 4% increase for both AUC and FROC for sentinel nodes, and an increase of 11% in AUC and 49% in FROC for axillary nodes. Pathologist qualitative evaluation of the retrained model´s output showed no missed positive slides. False positives, false negatives and one previously undetected micro-metastasis were observed. The study highlights the generalization challenge even when using a multicenter trained model, and that a small change in indication can considerably impact the model´s performance.

Cancer-associated fibroblasts (CAF) are highly prevalent cells in the tumor microenvironment in clear cell renal cell carcinoma (ccRCC). CAFs exhibit a pro-tumor effect in vitro and have been implicated in tumor cell proliferation, metastasis, and treatment resistance. Our objective is to analyze the geospatial distribution of CAFs with proliferating and apoptotic tumor cells in the ccRCC tumor microenvironment and determine associations with survival and systemic treatment. Pre-treatment primary tumor samples were collected from 96 patients with metastatic ccRCC. Three adjacent slices were obtained from 2 tumor-core regions of interest (ROI) per patient, and immunohistochemistry (IHC) staining was performed for αSMA, Ki-67, and caspase-3 to detect CAFs, proliferating cells, and apoptotic cells, respectively. H-scores and cellular density were generated for each marker. ROIs were aligned, and spatial point patterns were generated, which were then used to perform spatial analyses using a normalized Ripley's K function at a radius of 25 μm (nK(25)). The survival analyses used an optimal cut-point method, maximizing the log-rank statistic, to stratify the IHC-derived metrics into high and low groups. Multivariable Cox regression analyses were performed accounting for age and International Metastatic RCC Database Consortium (IMDC) risk category. Survival outcomes included overall survival (OS) from the date of diagnosis, OS from the date of immunotherapy initiation (OS-IT), and OS from the date of targeted therapy initiation (OS-TT). Therapy resistance was defined as progression-free survival (PFS) <6 months, and therapy response was defined as PFS >9 months. CAFs exhibited higher cellular clustering with Ki-67+ cells than with caspase-3+ cells (nK(25): Ki-67 1.19; caspase-3 1.05; p = 0.04). The median nearest neighbor (NN) distance from CAFs to Ki-67+ cells was shorter compared to caspase-3+ cells (15 μm vs. 37 μm, respectively; p < 0.001). Multivariable Cox regression analyses demonstrated that both high Ki-67+ density and H-score were associated with worse OS, OS-IT, and OS-TT. Regarding αSMA+CAFs, only a high H-score was associated with worse OS, OS-IT, and OS-TT. For caspase-3+, high H-score and density were associated with worse OS and OS-TT. Patients whose tumors were resistant to targeted therapy (TT) had higher Ki-67 density and H-scores than those who had TT responses. Overall, this ex vivo geospatial analysis of CAF distribution suggests that close proximity clustering of tumor cells and CAFs potentiates tumor cell proliferation, resulting in worse OS and resistance to TT in metastatic ccRCC.

SEC62 oncogene located at chromosomal region 3q26 encodes for a transmembrane protein of the endoplasmic reticulum (ER) and is expressed at high levels in numerous human malignancies. SEC62 overexpression has been associated with worse prognosis and high risk for lymphatic and distant metastases in head and neck cancer, cervical cancer, hepatocellular cancer, and lung cancer. However, its role in the development and tumor biology of melanocytic lesions has not been investigated so far. An immunohistochemical study including 209 patients with melanocytic lesions (malignant melanoma (MM), n = 93; melanoma metastases (MET), n = 28; Spitz nevi (SN), n = 29; blue nevi (BN), n = 21; congenital nevi (CN), n = 38) was conducted and SEC62 expression was correlated with clinical data including patient survival and histopathological characteristics. SN showed the highest SEC62 expression levels followed by MET, MM, CN, and BN. High SEC62 expression correlated with a shorter overall and progression-free survival in MM patients. Additionally, high Sec62 levels correlated significantly with higher tumor size (T stage), the presence of tumor ulceration, and the presence of lymph node as well as distant metastases. Strikingly, SEC62 expression showed a strong correlation with Clark level. Taken together, these data demonstrate that SEC62 is a promising prognostic marker in MM and has the potential to predict biological behavior and clinical aggressiveness of melanocytic lesions.

R0 resection is paramount in cutaneous squamous cell carcinoma (CSCC) and head and neck squamous cell carcinoma (HNSCC). However, in the setting of recurrence, immunocompromised patients, or non-keratinizing squamous cell carcinoma (SCC) with a spindle growth pattern, tumor borders are difficult, if not impossible, to determine. Fluorescence-guided surgery (FGS) aids in this differentiation. Potential targets for FGS of CSCC and HNSCC were evaluated. Most sections stained intensely for αvβ6 and epidermal growth factor receptor (EGFR) on tumor cells. Normal epithelium stained less for αvβ6 than for EGFR. In addition, soft tissue and stroma stained negative for both, allowing for clear discrimination of the soft tissue margin. Tumor cells weakly expressed urokinase plasminogen activator receptor (uPAR) while expression on stromal cells was moderate. Normal epithelium rarely expressed uPAR, resulting in clear discrimination of superficial margins. Tumors did not consistently express integrin β3, carcinoembryonic antigen, epithelial cell adhesion molecule, or vascular endothelial growth factor A. In conclusion, αvβ6 and EGFR allowed for precise discrimination of SSC at the surgically problematic soft tissue margins. Superficial margins are ideally distinguished with uPAR. In the future, FGS in the surgically challenging setting of cutaneous and mucosal SCC could benefit from a tailor-made approach, with EGFR and αvβ6 as targets.

Targeted molecular imaging may overcome current challenges in the preoperative and intraoperative delineation of pancreatic ductal adenocarcinoma (PDAC). Tumor-associated glycans Lea/c/x, sdi-Lea, sLea, sLex, sTn as well as mucin-1 (MUC1) and mucin-5AC (MU5AC) have gained significant interest as targets for PDAC imaging. To evaluate their PDAC molecular imaging potential, biomarker expression was determined using immunohistochemistry on PDAC, (surrounding) chronic pancreatitis (CP), healthy pancreatic, duodenum, positive (LN+) and negative lymph node (LN-) tissues, and quantified using a semi-automated digital image analysis workflow. Positive expression on PDAC tissues was found on 83% for Lea/c/x, 94% for sdi-Lea, 98% for sLea, 90% for sLex, 88% for sTn, 96% for MUC1 and 67% for MUC5AC, where all were not affected by the application of neoadjuvant therapy. Compared to PDAC, all biomarkers were significantly lower expressed on CP, healthy pancreatic and duodenal tissues, except for sTn and MUC1, which showed a strong expression on duodenum (sTn tumor:duodenum ratio: 0.6, p < 0.0001) and healthy pancreatic tissues (MUC1 tumor:pancreas ratio: 1.0, p > 0.9999), respectively. All biomarkers are suitable targets for correct identification of LN+, as well as the distinction of LN+ from LN- tissues. To conclude, this study paves the way for the development and evaluation of Lea/c/x-, sdi-Lea-, sLea-, sLex- and MUC5AC-specific tracers for molecular imaging of PDAC imaging and their subsequent introduction into the clinic.

Colorectal cancer (CRC) is one of the most common cancers worldwide. Although short-term cultures of tumour sections and xenotransplants have been used to determine drug efficacy, the results frequently fail to confer clinically useful information. Biomarker discovery has changed the paradigm for advanced CRC, though the presence of a biomarker does not necessarily translate into therapeutic success. To improve clinical outcomes, translational models predictive of drug response are needed. We describe a simple method for the fast establishment of CRC patient-derived explant (CRC-PDE) cultures from different carcinogenesis pathways, employing agitation-based platforms. A total of 26 CRC-PDE were established and a subset was evaluated for viability (n = 23), morphology and genetic key alterations (n = 21). CRC-PDE retained partial tumor glandular architecture and microenvironment features were partially lost over 4 weeks of culture. Key proteins (p53 and Mismatch repair) and oncogenic driver mutations of the original tumours were sustained throughout the culture. Drug challenge (n = 5) revealed differential drug response from distinct CRC-PDE cases. These findings suggest an adequate representation of the original tumour and highlight the importance of detailed model characterisation. The preservation of key aspects of the CRC microenvironment and genetics supports CRC-PDE potential applicability in pre- and co-clinical settings, as long as temporal dynamics are considered.

Alpha-synuclein (α-syn) is a protein considered to be detrimental in a number of degenerative disorders (synucleinopathies) of which α-syn aggregates are considered a pathological hallmark. The clearance of α-syn strongly depends on autophagy, which can be stimulated by inhibiting the mechanistic target of rapamycin (mTOR). Thus, the overexpression of mTOR and severe autophagy suppression may produce α-syn accumulation, including the proteinase K-resistant protein isoform. Glioblastoma multiforme (GBM) is a lethal brain tumor that features mTOR overexpression and severe autophagy inhibition. Cell pathology in GBM is reminiscent of a fast, progressive degenerative disorder. Therefore, the present work questions whether, as is analogous to neurons during degenerative disorders, an overexpression of α-syn occurs within GBM cells. A high amount of α-syn was documented in GBM cells via real-time PCR (RT-PCR), Western blotting, immunohistochemistry, immuno-fluorescence, and ultrastructural stoichiometry, compared with the amount of β- and γ-synucleins and compared with the amount of α-syn counted within astrocytes. The present study indicates that (i) α-syn is overexpressed in GBM cells, (ii) α-syn expression includes a proteinase-K resistant isoform, (iii) α-syn is dispersed from autophagy-like vacuoles to the cytosol, (iv) α-syn overexpression and cytosol dispersion are mitigated by rapamycin, and (v) the α-syn-related GBM-like phenotype is mitigated by silencing the SNCA gene.

Uveal melanoma is the most common primary intraocular malignancy in adults, with nearly half of all patients eventually developing metastases, which are invariably fatal. Manual assessment of the level of expression of the tumor suppressor BRCA1-associated protein 1 (BAP1) in tumor cell nuclei can identify patients with a high risk of developing metastases, but may suffer from poor reproducibility. In this study, we verified whether artificial intelligence could predict manual assessments of BAP1 expression in 47 enucleated eyes with uveal melanoma, collected from one European and one American referral center. Digitally scanned pathology slides were divided into 8176 patches, each with a size of 256 × 256 pixels. These were in turn divided into a training cohort of 6800 patches and a validation cohort of 1376 patches. A densely-connected classification network based on deep learning was then applied to each patch. This achieved a sensitivity of 97.1%, a specificity of 98.1%, an overall diagnostic accuracy of 97.1%, and an F1-score of 97.8% for the prediction of BAP1 expression in individual high resolution patches, and slightly less with lower resolution. The area under the receiver operating characteristic (ROC) curves of the deep learning model achieved an average of 0.99. On a full tumor level, our network classified all 47 tumors identically with an ophthalmic pathologist. We conclude that this deep learning model provides an accurate and reproducible method for the prediction of BAP1 expression in uveal melanoma.

Checkpoint blockade therapies have changed the clinical management of metastatic melanoma patients considerably, showing survival benefits. Despite the clinical success, not all patients respond to treatment or they develop resistance. Although there are several treatment predictive biomarkers, understanding therapy resistance and the mechanisms of tumor immune evasion is crucial to increase the frequency of patients benefiting from treatment. The PTEN gene is thought to promote immune evasion and is frequently mutated in cancer and melanoma. Another feature of melanoma tumors that may affect the capacity of escaping T-cell recognition is melanoma cell dedifferentiation characterized by decreased expression of the microphtalmia-associated transcription factor (MITF) gene. In this study, we have explored the role of PTEN in prognosis, therapy response, and immune escape in the context of MITF expression using immunostaining and genomic data from a large cohort of metastatic melanoma. We confirmed in our cohort that PTEN alterations promote immune evasion highlighted by decreased frequency of T-cell infiltration in such tumors, resulting in a worse patient survival. More importantly, our results suggest that dedifferentiated PTEN negative melanoma tumors have poor patient outcome, no T-cell infiltration, and transcriptional properties rendering them resistant to targeted- and immuno-therapy.

Molecular biomarkers have the potential to predict response to the treatment of rectal cancer. In this study, we aimed to evaluate the prognostic and clinicopathological implication of RAD50 (DNA repair protein RAD50 homolog) expression in rectal cancer.

Gastric cancer is an aggressive and heterogeneous malignancy that often varies in presentation and disease among racial and ethnic groups. The Alaska Native (AN) people have the highest incidence and mortality rates of gastric cancer in North America. This study examines molecular markers in solid tumor samples from eighty-five AN gastric adenocarcinoma patients using next-generation sequencing, immunohistochemistry, and in situ hybridization analysis. AN patients have a low mutation burden with fewer somatic gene mutations in their tumors compared to other populations, with the most common mutation being TP53. Epstein-Barr virus (EBV) was associated with 20% of AN gastric cancers, which is higher than the world average of 10%. The inflammation marker, cyclooxygenase-2 (COX-2), is highly expressed in patients with the lowest survival rates. Mismatch repair deficiency was present in 10% of AN patients and was associated with patients who were female, 50 years or older, gene mutations, and tumors in the distal stomach. Program death-ligand 1 (PD-L1) was expressed in 14% of AN patients who were more likely to have MMR deficiency, EBV-associated gastric cancers, and mutations in the PIK3CA gene, all of which have been linked to clinical response to PD-1 inhibitors. These studies suggest a portion of AN gastric cancer patients could be candidates for immunotherapy. Overall, this study highlights future avenues of investigation for clinical and translational studies, so that we can improve early detection and develop more effective treatments for AN patients.

To examine and compare the prognostic value of immune cell score (ICS) and Klintrup-Mäkinen (KM) grade in gastric cancer.