Digital pathology has recently led to significant advancements in the field of microscopic image analysis, particularly regarding the increasing use of Deep Learning methods. These models represent the state-of-the-art in histopathological slide analysis, but Deep Learning features remain difficult to interpret, despite recent developments in post hoc explainability frameworks. In contrast, features extracted from biological objects—such as nuclei, cells or tissues—are supposed to be more grounded in pathologists' a priori knowledge. Accordingly, Machine Learning based on handcrafted features represents another paradigm of explainability and may stand as a complementary method to Deep Learning to assist pathologists.In order to perceive how biological features have been used in hematoxylin eosin microscopic images to address medical questions, we conducted a systematic review of articles published from January 2005 to May 2025, adhering to PRISMA guidelines.A total of 97 articles were analyzed from the PubMed, IEEE, and ACM databases. Three primary categories of features—texture/color, morphology and topology—were both identified and thoroughly described. These features were most frequently derived from segmented cells and tissues in 80 and 28 studies, respectively. They were used to address seven types of medical questions: “normal vs diseased”, disease subtyping, tumor grading, phenotyping, object detection, prognosis and treatment-response prediction.We discussed methodological and reporting limitations of these studies, highlighting the difficulty to assess the potential impact of such methods. Among the most common concerns, we found features difficult to interpret, data leakage, and inadequate sample sizes. Nevertheless, we also focused on promising domain-inspired feature engineering that provides better explainability and specificity. This kind of features associated with more methodological rigor may increase the relevance and reliability of AI models, and also raise new research avenues in pathology.

Bispecific antibodies (bsAbs) such as glofitamab represent a promising therapeutic approach for relapsed/refractory B-cell non-Hodgkin's lymphoma (R/R B-NHL), but resistance mechanisms remain poorly understood. This study aimed to identify predictive markers of bsAbs resistance based on the response of 3D patient-derived lymphoma spheroids (PDLS) established from 39 R/R B-NHL samples. PDLS were treated with glofitamab for 3 days and B-cell depletion was quantified to assess the ex-vivo treatment response. Comprehensive immune profiling was performed on patient samples using multiparametric flow cytometry, single-cell RNA sequencing, CODEX spatial proteomics and functional assays. High responders to glofitamab possessed CD8+ T-cells with consistently higher cytotoxic and activation signatures across effector differentiation states, while low responders showed enrichment of exhausted CD8+ T-cell with enhanced expression of exhaustion markers (TIGIT, LAG3, PD1). Furthermore, low responders exhibited elevated functional CD4+ T-follicular helper (Tfh) cells in close proximity to malignant B-cell thus promoting their survival through IL21 and CXCL13 signaling pathways. Analysis of pretreatment RNA-seq data from 48 R/R B-NHL patients confirmed that high Tfh abundance is associated with poor glofitamab response. In PDLS, anti-TIGIT co-treatment enhanced glofitamab efficacy in low responders, and Tfh depletion experiments confirmed that reducing Tfh activity increased B-cell depletion. Together, these findings identify CD8+ T-cell exhaustion and functionally activated Tfh cells as key factors associated with glofitamab resistance in R/R B-NHL. This work supports their potential use as predictive biomarkers for selecting patients with higher probability of response and provides a foundation for future combination therapeutic strategies.

Les dernières classifications (ICC et OMS 2022) ont renforcé la distinction entre le lymphome de Hodgkin classique et le lymphome de Hodgkin à prédominance lymphocytaire nodulaire qui présentent des caractéristiques morphologiques, immunophénotypiques, biologiques et cliniques distinctes. À ce titre, le lymphome de Hodgkin à prédominance lymphocytaire nodulaire a été renommé lymphome B à prédominance lymphocytaire nodulaire et déplacé dans le groupe des lymphomes B dans la classification ICC 2022. Les critères diagnostiques du lymphome de Hodgkin classique demeurent inchangés et reposent sur la présence de cellules de Hodgkin et de Reed-Sternberg noyées dans un microenvironnement immunitaire riche, témoignant paradoxalement d’un échec du système immunitaire à éliminer les cellules tumorales. Les cellules de Hodgkin et de Reed-Sternberg dérivent de lymphocytes B ayant perdu l’essentiel du répertoire B, à l’exception du facteur de transcription PAX5, marqueurs diagnostiques clés. Le lymphome de Hodgkin/lymphome B à prédominance lymphocytaire nodulaire se caractérise par une prolifération de petits lymphocytes B réactionnels mêlés à de grandes cellules tumorales LP (lymphocyte predominant) ou « cellules popcorn » avec différents types architecturaux dont certains pourraient avoir une valeur pronostique. Des difficultés diagnostiques persistent, notamment entre le lymphome de Hodgkin classique riche en lymphocytes et le lymphome de Hodgkin/lymphome B à prédominance lymphocytaire nodulaire. En cas de localisation médiastinale le diagnostic différentiel avec les lymphomes B médiastinaux peut également s’avérer compliqué. Enfin, le diagnostic différentiel avec un lymphome T TFH peut également être délicat en particulier sur des biopsies à l’aiguille. Dans ce contexte diagnostique complexe, les analyses moléculaires constituent un apport diagnostique majeur.

Tertiary lymphoid structures (TLS) have emerged as critical immune niches within the tumor microenvironment across various cancers. However, their structural organization and functional roles in non-muscle-invasive bladder cancer (NMIBC) remain poorly characterized. In this study, we comprehensively characterized TLS in NMIBC using en bloc resected primary diagnostic specimens, which enabled high-resolution spatial mapping and quantitative histological assessment across a large cohort of primary tumors (97 patients). Compared with TLS in inflammation-driven cystitis, tumor-associated TLS showed heterogeneous spatial organization, altered immune composition, and disrupted follicular dendritic cell (FDC) networks with reduced high endothelial venules density. TLS density increased with NMIBC grade and stage and correlated with higher recurrence risk and shorter progression-free survival in multivariable analysis. Paired analysis of primary and recurrent tumors showed a reduction in FDC density and a shift from M1 to M2 differentiation in tumor recurrences, suggesting the development of an immunosuppressive microenvironment. Single-cell RNA sequencing of TLS from NMIBC and cystitis samples revealed impaired germinal center activity in tumor-associated TLS, with asynchronous B-cell maturation and reduced B-cell interactions with both T follicular helper cells and FDC. Tumor-associated TLS also showed increased myeloid infiltration and altered dendritic-cell function, including reduced MHC class II antigen presentation and downregulation of co-stimulatory (CD86-CD28) and migration-related (CD99) pathways compared to TLS from cystitis samples. Overall, we found that NMIBC-associated TLS are structurally and functionally impaired, which may limit effective local immune responses and suggests they may be relevant as biomarkers and targets for immunomodulatory therapy in bladder cancer.

In this commentary, we open the debate on what can be expected from artificial intelligence (AI) in the diagnosis of hematologic cancers. We discuss the key factors that make AI solutions robust, trustworthy, and, above all, generalizable, with particular emphasis on the importance of dataset quality in shaping the performance and effectiveness of AI models.

Background: Inflammatory bowel disease (IBD), including ulcerative colitis (UC) and Crohn’s disease (CD), increase colorectal cancer (CRC) risk. Methods: Mouse IBD and CRC models with a combination of pharmacological, knockout and knock-in approaches was employed to analyze the involvement of TOP2s and NOS2 in CRC tumorigenesis. Key pathologies, such as inflammatory and neoplastic scores, were examined by immunohistochemical assays. Results: In colon tissues from acute, chronic colitis and CRC mouse models and from CD patients, the biomarkers γH2AX and 53BP1pS25/S29 of DNA breaks (mainly representing DSBs) accumulated, alongside increases in topoisomerase II (TOP2) and nitric oxide synthase 2 (NOS2). Genetic ablation of NOS2 (Nos2-/-) or TOP2β (Top2βf/f) as well as pharmacological inhibition with ICRF-193 (a TOP2 inhibitor) or PTIO (a NO scavenger) reduced DSB formation and disease severity. Consistently, Nos2-/-, or ICRF-treated, mice exhibited decreased tumor burden. DSBs and tumor accumulation were pronounced in the distal colon, mirroring human CRC distribution. While ICRF-193 suppressed tumor growth, Top2βf/f deficiency (with a compensatory TOP2α upregulation) enhanced tumor development, indicating potential roles for TOP2 isozymes in tumor formation and progression. Conclusion: Collectively, these findings identify the cooperative action of TOP2 and NOS2 in driving DSBs, highlighting a potential therapeutic target in inflammation-associated CRC.

MET amplification (METamp) is a noteworthy genomic alteration that can occur in patients with non-small cell lung cancer (NSCLC). It has been demonstrated to occur as a primary oncogenic driver that may exist prior to any treatment and is referred to as de novo METamp. Despite the recognized significance of this genetic alteration, routine large-scale screening for the early detection of de novo METamp is currently lacking in clinical practice and the clinical impact of de novo METamp on NSCLC remains poorly investigated. In this study, we developed a next-generation sequencing-based screening method for detecting and stratifying METamp optimized in silico, validated in a patient cohort (n = 72) and applied to 1932 patients with NSCLC. Clinical outcomes (overall survival [OS] and progression-free survival) were assessed in de novo METamp cases (n = 46). The optimized next-generation sequencing-based method achieved high confidence (F-score > 0.99) during in silico optimization. In vivo validation demonstrated high sensitivity (0.93) and specificity (0.97) compared with fluorescence in situ hybridization. De novo METamp was found in 2.4% of cases stratified into the following distinct amplification groups based on the amplification copy number ratio (CNR): low (1.5 < CNR < 2.2), medium (2.2 < CNR < 4), and high amplification ( CNR > 4). Significant differences in patient outcome (P < .001) were observed between the low- (median OS: 35.9 months), medium- (median OS: 14.3 months) and high-amplification (median OS: 3.3 months) groups. Progression-free survival under chemotherapy was notably reduced in the medium-/high-amplification groups compared with the low-amplification group (P = .001). Screening for METamp detection followed by stratification based on METamp levels may be considered in all patients with NSCLC at diagnosis. This approach could potentially enhance treatment management effectiveness by facilitating inclusion in clinical trials. (c) 2025 United States & Canadian Academy of Pathology. Published by Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Innate lymphoid cells (ILC) distribution and compartmentalization in human lymphoid tissues are incompletely described. Through combined multiplex immunofluorescence, multispectral imaging, and advanced computer vision methods, we provide a map of ILCs at the whole-slide single-cell resolution level, and study their proximity to T helper (Th) cells. The results show that ILC2 predominates in thymic medulla; by contrast, immature Th cells prevail in the cortex. Unexpectedly, we find that Th2-like and Th17-like phenotypes appear before complete T cell receptor gene rearrangements in these immature thymocytes. In the periphery, ILC2 are more abundant in lymph nodes and tonsils, penetrating lymphoid follicles. NK cells are uncommon in lymphoid tissues but abundant in the spleen, whereas ILC1 and ILC3 predominate in the ileum and appendix. Under pathogenic conditions, a deep perturbation of both ILC and Th populations is seen in follicular lymphoma compared with non-neoplastic conditions. Lastly, all ILCs are preferentially in close proximity to their Th counterparts. In summary, our histopathology tool help present a spatial mapping of human ILCs and Th cells, in normal and neoplastic lymphoid tissues.

The EWSR1::CREM rearranged intra-abdominal malignant epithelioid neoplasm is an emerging tumor, with only a few publications describing it to date. Here, we report two new cases of this highly aggressive tumor, primarily involving the peritoneal surface. The tumors presented as a widespread diffuse peritoneal lesion associated with a 4-cm pelvic mass in a 28-year-old woman (Case 1) and as a 10-cm intra-abdominal mass infiltrating the stomach with multiple hepatic metastases in a 53-year-old woman (Case 2). The tumors shared predominant epithelioid morphology with minimal nuclear polymorphism. One of them additionally harbored spindle and rhabdoid cell populations. Both tumors displayed immunoreactivity for pan-cytokeratins, EMA, and CD99, and variable positivity for MUC4, progesterone and estrogen receptors, pan-NTRK, and synaptophysin. This misleading histology and immunophenotype give rise to a wide spectrum of differential diagnoses and highlight the crucial role of RNA sequencing in diagnostic accuracy and thus in appropriate therapeutic approaches.

Osteosarcoma is the most common malignant bone tumor among children and young adults. Distinguishing between osteoblastoma and osteosarcoma can be particularly challenging, especially in small tissue samples and for the osteoblastoma-like osteosarcoma subtype. Recent studies with conflicting results have suggested a potential malignant transformation process from osteoblastoma to osteosarcoma. This study aims to investigate the hypothesis of osteoblastoma evolving into osteosarcoma and to discuss its clinical implications. We conducted a retrospective multicentric case-series study, collecting clinical, radiological, histological, and follow-up data from osteosarcoma cases suspected to have originated from malignant transformation of osteoblastoma within the French ResOs network. Molecular analyses (fluorescence in situ hybridization (FISH) and next-generation sequencing (NGS)) were performed. We included two cases (one female and one male), with a median age at osteosarcoma diagnosis of 42 and 73 years old, respectively. One patient had tumor located in the axial skeleton, and both cases exhibited features of osteoblastoma-like osteosarcoma. Notably, one of the patients had a documented history of osteoblastoma diagnosed sixteen years earlier. FISH analysis revealed FOS rearrangements in both osteosarcoma cases, with tumors presenting uncommon fusion transcripts (FOS::VGLL4 and FOS::COL5A2) identified through NGS. Both patients were alive at last follow-up. Our findings suggest that osteosarcoma can rarely present with FOS gene fusions and be associated with an indolent progression, challenging the specificity of such signatures for osteoblastoma diagnosis. This discovery also raises the hypothesis of malignant transformation from osteoblastoma to osteosarcoma and underscores the necessity for diligent monitoring of FOS-rearranged bone-forming tumors for optimal therapeutic management.

Classic Hodgkin’s lymphoma (cHL) is a curable cancer with a disease-free survival rate of over 10 years. Over 80% of diagnosed patients respond favorably to first-line chemotherapy, but few biomarkers exist that can predict the 15–20% of patients who experience refractory or early relapsed disease. To date, the identification of patients who will not respond to first-line therapy based on disease staging and traditional clinical risk factor analysis is still not possible. Three-dimensional (3D) telomere analysis using the TeloView® software platform has been shown to be a reliable tool to quantify genomic instability and to inform on disease progression and patients’ response to therapy in several cancers. It also demonstrated telomere dysfunction in cHL elucidating biological mechanisms related to disease progression. Here, we report 3D telomere analysis on a multicenter cohort of 156 cHL patients. We used the cohort data as a training data set and identified significant 3D telomere parameters suitable to predict individual patient outcomes at the point of diagnosis. Multivariate analysis using logistic regression procedures allowed for developing a predictive scoring model using four 3D telomere parameters as predictors, including the proportion of t-stumps (very short telomeres), which has been a prominent predictor for cHL patient outcome in a previously published study using TeloView® analysis. The percentage of t-stumps was by far the most prominent predictor to identify refractory/relapsing (RR) cHL prior to initiation of adriamycin, bleomycin, vinblastine, and dacarbazine (ABVD) therapy. The model characteristics include an AUC of 0.83 in ROC analysis and a sensitivity and specificity of 0.82 and 0.78 respectively.

Background: MET amplification (METamp) is a noteworthy genomic alteration that can occur in patients with non-small cell lung cancer (NSCLC). It has been demonstrated to occur as a primary oncogenic driver that may exist prior to any treatment and is referred to as de novo METamp. Despite the recognized significance of this genetic alteration, routine large-scale screening for the early detection of de novo METamp is currently lacking in clinical practice and the clinical impact of de novo METamp in NSCLC remains poorly investigated. Methods: In this study, we developed a NGS-based screening method for detecting and stratifying METamp optimized in silico, validated in a patient cohort (n = 72) and applied to 1,932 NSCLC patients. Clinical outcomes (OS and PFS) were assessed in de novo METamp cases (n = 46). Results: The optimized NGS-based method achieved high confidence (F-score > 0.99) during in silico optimization. In vivo validation demonstrated high sensitivity (0.93) and specificity (0.97) compared to fluorescent in situ hybridization. de novo METamp was found in 2.4% of cases stratified into distinct amplification groups based on the amplification copy number ratio (CNR): Low- (1.5 < CNR ≤ 2.2), Medium- (2.2 < CNR ≤ 4), and High-amplification (CNR > 4). Significant differences in patient outcome (p < 0.001) were observed between the Low- (median OS: 35.9 months), Medium- (median OS: 14.3 months) and High-amplification (median OS: 3.3 months) groups. PFS under chemotherapy was notably reduced in the Medium/High-amplification groups compared to the Low-amplification group (p = 0.001). Conclusions: Screening for METamp detection followed by stratification based on METamp levels may be considered in all NSCLC patients at diagnosis. This approach could potentially enhance treatment management effectiveness by facilitating inclusion in clinical trials.

Depuis plus d’une dizaine d’années, la France a mis en place plusieurs initiatives et programmes visant à promouvoir et à exploiter le potentiel des techniques de next-generation sequencing, ou séquençage nouvelle génération (NGS) dans différents domaines. Le pays dispose d’un certain nombre de plateformes de séquençage nouvelle génération hautement équipées et de centres de recherche spécialisés dans le NGS. Ces infrastructures fournissent des services de séquençage à des chercheurs académiques, à des institutions médicales et à des partenaires industriels. Le NGS est largement utilisé dans la recherche en biologie médicale, en particulier en génétique moléculaire pour développer des approches de médecine personnalisée et dans la recherche en génomique. Au travers plusieurs auditions d’experts dans le domaine de la génétique médicale et paleogénomique, nous avons obtenu un aperçu de la contribution de ces techniques en médecine et en génétique des populations. Nous envisageons plusieurs perspectives d’utilisation du NGS en parallèle des progrès biotechnologiques et bioinformatiques qui lui sont attribués.

The t(14;19)(q32;q13) often juxtaposes BCL3 with immunoglobulin heavy chain (IGH) resulting in overexpression of the gene. In contrast to other oncogenic translocations, BCL3 rearrangement (BCL3-R) has been associated with a broad spectrum of lymphoid neoplasms. Here we report an integrative whole-genome sequence, transcriptomic, and DNA methylation analysis of 13 lymphoid neoplasms with BCL3-R. The resolution of the breakpoints at single base-pair revealed that they occur in two clusters at 5' (n=9) and 3' (n=4) regions of BCL3 associated with two different biological and clinical entities. Both breakpoints were mediated by aberrant class switch recombination of the IGH locus. However, the 5' breakpoints (upstream) juxtaposed BCL3 next to an IGH enhancer leading to overexpression of the gene whereas the 3' breakpoints (downstream) positioned BCL3 outside the influence of the IGH and were not associated with its expression. Upstream BCL3-R tumors had unmutated IGHV, trisomy 12, and mutated genes frequently seen in chronic lymphocytic leukemia (CLL) but had an atypical CLL morphology, immunophenotype, DNA methylome, and expression profile that differ from conventional CLL. In contrast, downstream BCL3-R neoplasms were atypical splenic or nodal marginal zone lymphomas (MZL) with mutated IGHV, complex karyotypes and mutated genes typical of MZL. Two of the latter four tumors transformed to a large B-cell lymphoma. We designed a novel fluorescence in situ hybridization assay that recognizes the two different breakpoints and validated these findings in 17 independent tumors. Overall, upstream or downstream breakpoints of BCL3-R are mainly associated with two subtypes of lymphoid neoplasms with different (epi)genomic, expression, and clinicopathological features resembling atypical CLL and MZL, respectively.

The assessment of chemotherapy response in osteosarcoma (OS), based on the average percentage of viable cells, is limited, as it overlooks the spatial heterogeneity of tumor cell response (foci of resistant cells), immune microenvironment, and bone microarchitecture. Despite the resulting positive classification for response to chemotherapy, some patients experience early metastatic recurrence, demonstrating that our conventional tools for evaluating treatment response are insufficient. We studied the interactions between tumor cells, immune cells (lymphocytes, histiocytes, and osteoclasts), and bone extracellular matrix (ECM) in 18 surgical resection samples of OS using multiplex and conventional immunohistochemistry (IHC: CD8, CD163, CD68, and SATB2), combined with multiscale characterization approaches in territories of good and poor response (GRT/PRT) to treatment. GRT and PRT were defined as subregions with <10% and >10% of viable tumor cells, respectively. Local correlations between bone ECM porosity and density of immune cells were assessed in these territories. Immune cell density was then correlated to overall patient survival. Two patterns were identified for histiocytes and osteoclasts. In poor responder patients, CD68 osteoclast density exceeded that of CD163 histiocytes but was not related to bone ECM load. Conversely, in good responder patients, CD163 histiocytes were more numerous than CD68 osteoclasts. For both of them, a significant negative local correlation with bone ECM porosity was found (P < ,01). Moreover, in PRT, multinucleated osteoclasts were rounded and intermingled with tumor cells, whereas in GRT, they were elongated and found in close contact with bone trabeculae. CD8 levels were always low in metastatic patients, and those initially considered good responders rapidly died from their disease. The specific recruitment of histiocytes and osteoclasts within the bone ECM, and the level of CD8 represent new features of OS response to treatment. The associated / (2024) prognostic signatures should be integrated into the therapeutic stratification algorithm of patients after surgery. (c) 2024 United States & Canadian Academy of Pathology. Published by Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Artificial intelligence (AI)-assisted diagnosis is an ongoing revolution in pathology. However, a frequent drawback of AI models is their propension to make decisions based rather on bias in training dataset than on concrete biological features, thus weakening pathologists' trust in these tools. Technically, it is well known that microscopic images are altered by tissue processing and staining procedures, being one of the main sources of bias in machine learning for digital pathology. So as to deal with it, many teams have written about color normalization and augmentation methods. However, only a few of them have monitored their effects on bias reduction and model generalizability. In our study, two methods for stain augmentation (AugmentHE) and fast normalization (HEnorm) have been created and their effect on bias reduction has been monitored. Actually, they have also been compared to previously described strategies. To that end, a multicenter dataset created for breast cancer histological grading has been used. Thanks to it, classification models have been trained in a single center before assessing its performance in other centers images. This setting led to extensively monitor bias reduction while providing accurate insight of both augmentation and normalization methods. AugmentHE provided an 81% increase in color dispersion compared to geometric augmentations only. In addition, every classification model that involved AugmentHE presented a significant increase in the area under receiving operator characteristic curve (AUC) over the widely used RGB shift. More precisely, AugmentHE-based models showed at least 0.14 AUC increase over RGB shift-based models. Regarding normalization, HEnorm appeared to be up to 78x faster than conventional methods. It also provided satisfying results in terms of bias reduction. Altogether, our pipeline composed of AugmentHE and HEnorm improved AUC on biased data by up to 21.7% compared to usual augmentations. Conventional normalization methods coupled with AugmentHE yielded similar results while being much slower. In conclusion, we have validated an open-source tool that can be used in any deep learning-based digital pathology project on H&E whole slide images (WSI) that efficiently reduces stain-induced bias and later on might help increase pathologists' confidence when using AI-based products.

For more than ten years, France has implemented several initiatives and programs aimed at promoting and exploiting the potential of next-generation sequencing (NGS) techniques in various fields. The country has a number of highly equipped next-generation sequencing platforms and research centers specializing in NGS. These infrastructures provide sequencing services to academic researchers, medical institutions and industrial partners. NGS is widely used in medical biology, particularly in molecular medicine and genetics, to develop personalized medicine approaches and in genomics research. Through several hearings with experts in the field of medical genetics and paleo-genomics, we obtained an overview of the contribution of NGS techniques to medicine and population genetics. We envisage several perspectives for the use of NGS in parallel with the biotechnological and bioinformatics progresses dedicated to it. (c) 2024 Published by Elsevier Masson SAS on behalf of l'Academie nationale de medecine.