Slide Scanner Research Articles

Artificial intelligence can be trained to predict c-KIT-11 mutational status of canine mast cell tumors from hematoxylin and eosin-stained histological slides.

Numerous prognostic factors are currently assessed histologically and immunohistochemically in canine mast cell tumors (MCTs) to evaluate clinical behavior. In addition, polymerase chain reaction (PCR) is often performed to detect internal tandem duplication (ITD) mutations in exon 11 of the c-KIT gene (c-KIT-11-ITD) to predict the therapeutic response to tyrosine kinase inhibitors. This project aimed at training deep learning models (DLMs) to identify MCTs with c-KIT-11-ITD solely based on morphology. Hematoxylin and eosin (HE) stained slides of 368 cutaneous, subcutaneous, and mucocutaneous MCTs (195 with ITD and 173 without) were stained consecutively in 2 different laboratories and scanned with 3 different slide scanners. This resulted in 6 data sets (stain-scanner variations representing diagnostic institutions) of whole-slide images. DLMs were trained with single and mixed data sets and their performances were assessed under stain-scanner variations (domain shifts). The DLM correctly classified HE slides according to their c-KIT-11-ITD status in up to 87% of cases with a 0.90 sensitivity and a 0.83 specificity. A relevant performance drop could be observed when the stain-scanner combination of training and test data set differed. Multi-institutional data sets improved the average accuracy but did not reach the maximum accuracy of algorithms trained and tested on the same stain-scanner variant (ie, intra-institutional). In summary, DLM-based morphological examination can predict c-KIT-11-ITD with high accuracy in canine MCTs in HE slides. However, staining protocol and scanner type influence accuracy. Larger data sets of scans from different laboratories and scanners may lead to more robust DLMs to identify c-KIT mutations in HE slides.

Rapid 3D imaging at cellular resolution for digital cytopathology with a multi-camera array scanner (MCAS)

Optical microscopy has long been the standard method for diagnosis in cytopathology. Whole slide scanners can image and digitize large sample areas automatically, but are slow, expensive and therefore not widely available. Clinical diagnosis of cytology specimens is especially challenging since these samples are both spread over large areas and thick, which requires 3D capture. Here, we introduce a new parallelized microscope for scanning thick specimens across extremely wide fields-of-view (54 × 72 mm2) at 1.2 and 0.6 μm resolutions, accompanied by machine learning software to rapidly assess these 16 gigapixel scans. This Multi-Camera Array Scanner (MCAS) comprises 48 micro-cameras closely arranged to simultaneously image different areas. By capturing 624 megapixels per snapshot, the MCAS is significantly faster than most conventional whole-slide scanners. We used this system to digitize entire cytology samples (scanning three entire slides in 3D in just several minutes) and demonstrate two machine learning techniques to assist pathologists: first, an adenocarcinoma detection model in lung specimens (0.73 recall); second, a slide-level classification model of lung smears (0.969 AUC).

Efficient, gigapixel-scale, aberration-free whole slide scanner using angular ptychographic imaging with closed-form solution.

Whole slide imaging provides a wide field-of-view (FOV) across cross-sections of biopsy or surgery samples, significantly facilitating pathological analysis and clinical diagnosis. Such high-quality images that enable detailed visualization of cellular and tissue structures are essential for effective patient care and treatment planning. To obtain such high-quality images for pathology applications, there is a need for scanners with high spatial bandwidth products, free from aberrations, and without the requirement for z-scanning. Here we report a whole slide imaging system based on angular ptychographic imaging with a closed-form solution (WSI-APIC), which offers efficient, tens-of-gigapixels, large-FOV, aberration-free imaging. WSI-APIC utilizes oblique incoherent illumination for initial high-level segmentation, thereby bypassing unnecessary scanning of the background regions and enhancing image acquisition efficiency. A GPU-accelerated APIC algorithm analytically reconstructs phase images with effective digital aberration corrections and improved optical resolutions. Moreover, an auto-stitching technique based on scale-invariant feature transform ensures the seamless concatenation of whole slide phase images. In our experiment, WSI-APIC achieved an optical resolution of 772 nm using a 10×/0.25 NA objective lens and captures 80-gigapixel aberration-free phase images for a standard 76.2 mm × 25.4 mm microscopic slide.

Concordance of whole slide imaging and conventional light microscopy for assessment of pathologic response following neoadjuvant therapy for lung cancer

Pathologic response is an endpoint in many ongoing clinical trials for neoadjuvant regimens, including immune checkpoint blockade and chemotherapy. Whole slide scanning of glass slides generates high resolution digital images and allows for remote review and potential measurement with image analysis tools, but concordance of pathologic response assessment on digital scans compared to glass slides has yet to be evaluated. Such a validation goes beyond previous concordance studies which focused on establishing surgical pathology diagnoses, as it requires quantitative assessment of tumor, necrosis, and regression. Further, as pathologic response assessment is being used as an endpoint, such concordance studies have regulatory implications. The purpose of this study was two fold: firstly, to determine the concordance between pathologic response assessed on glass slides and on digital scans; and secondly, to determine if pathologists benefited from using measurement tools when determining pathologic response. To that end, H&E-stained glass slides from 64 non-small cell lung carcinoma specimens were visually assessed for percent residual viable tumor (%RVT). The sensitivity and specificity for digital vs. glass reads of complete pathologic response (pCR, 0% RVT) and major pathologic response (MPR, ≤10% RVT) were all >95%. When %RVT was considered as a continuous variable, intraclass correlation coefficient of digital vs. glass reads was 0.94. The visual assessments of pathologic response were supported by pathologist annotations of residual tumor and tumor bed areas. In a separate subset of H&E-stained glass slides, several measurement approaches to quantifying %RVT were performed. Pathologist estimates strongly reflected measured %RVT. This study demonstrates the high level of concordance between glass slides evaluated using light microscopy and digital whole slide images for pathologic response assessments. Pathologists did not require measurement tools to generate robust %RVT values from slide annotations. These findings have broad implications for improving clinical workflows and multisite clinical trials.

Performance of the Vetscan Imagyst in point-of-care detection of Giardia duodenalis in canine fecal samples.

Giardia duodenalis is a common parasite of the gastrointestinal tract of dogs, with an especially high prevalence in dogs <1-y-old. Methods for detecting G. duodenalis are point-of-care (POC) tests such as lateral-flow tests or fecal flotation. The Vetscan Imagyst (Zoetis) is a new POC device for the detection of G. duodenalis in fecal samples using zinc sulfate flotation, automated slide scanning, and image recognition with artificial intelligence. Vetscan results are the number of Giardia cysts per coverslip. We compared the performance of the Vetscan and another POC test (SNAP Giardia test; Idexx) with a direct immunofluorescence assay (IFA) performed in a specialized parasitology laboratory as the reference test. We included 164 dogs <19-mo-old. We used pooled fecal samples from 3 defecations gained within 2-3 d and tested the repeatability of the Vetscan by triplicate measurement. Compared to IFA, Vetscan had a diagnostic sensitivity of 88.4% and specificity of 98.1%; SNAP had a diagnostic sensitivity of 74.4% and specificity of 98.1%. A variation coefficient of 67.0% was determined for the Vetscan results. The performance of the Vetscan is acceptable for the qualitative evaluation of fecal samples (Giardia positive or negative), and the device can be used by untrained personnel. Given its high variation coefficient, we do not recommend the Vetscan for monitoring the number of cysts.

Effect of Vernonia Cinerea Water Extract on the Protection of Liver Tissue in Chronic Nicotine-Treated Rats

Nicotine, a highly toxic alkaloid found in tobacco, is known for its addictive properties and systemic side effects, including carcinogenic potential and multi-organ damage. Recent evidence indicates that nicotine can significantly impair liver function and regeneration by promoting fibrogenesis and oxidative stress. This study aimed to investigate the protective effects of Vernonia cinerea (VC) against nicotine-induced liver damage in Wistar rats, focusing on inflammation and pro-fibrosis markers. Thirty male Wistar rats were divided into three groups: a control group, a nicotine group (1 mg/kg/day), and a nicotine plus VC group (1 mg/kg/day nicotine and 100 mg/kg/day VC). Liver tissues were examined using Hematoxylin and Eosin staining, and Masson's trichrome staining to evaluate histological changes. Kupffer cell counts were determined using a Panoramic digital slide scanner. Statistical analyses were conducted to compare the groups. Nicotine exposure led to significant liver damage, characterized by increased inflammation, collagen deposition, and disruption of liver architecture. VC supplementation ameliorated these effects, reducing inflammation and fibrosis markers. Kupffer cell counts were lower in the nicotine plus VC group compared to the nicotine group alone. Conclusion: VC demonstrates a protective role in mitigating nicotine-induced liver damage, highlighting its potential as a therapeutic agent for preventing liver fibrosis and maintaining liver health in nicotine-exposed individuals.

A histology and immunochemistry study comparing open versus close and open-method endoscopic saphenous vein harvesting.

The introduction of endoscopic saphenous vein harvesting (ESVH) has been reported to reduce wound pain and infection, compared with open saphenous vein harvesting (OSVH) techniques. There are still controversies regarding this technique. The aim of our study is to investigate the endothelial preservation of saphenous vein (SV) grafts harvested by different techniques. Further observations were made for harvesting and closure time, incision length and effect of pressure distension of the veins to the vein endothelium. Prospective observational study of sixty human saphenous vein grafts was performed to evaluate endothelial preservation by haematoxylin-eosin and Cluster of Differentiation 31 (CD 31) staining. Saphenous vein was harvested endoscopically either by closed CO2 (carbon dioxide) ESVH, open CO2 ESVH or OSVH harvesting technique. Demographic data and intra-operative data were collected. Two saphenous vein samples were collected from each patient to compare differences before and after distension of the veins. Both haematoxylin-eosin and immunohistochemistry slides were imaged by a high-resolution slide scanning system. Open CO2 ESVH group showed the highest number of endothelial detachments. Mean scoring of the immunohistochemistry method using the CD31 antibody was much lower in the open CO2 ESVH group (33.25% ± 28.71, P < 0.0003). This represents a more poorly preserved endothelial cells in the Open CO2 ESVH than the closed CO2 ESVH. Closure time and incision lengths were significantly shorter in both ESVH groups compared to the OSVH group. Significant low scores of immunohistochemistry for samples were seen in distended veins (39.0% ± 30.08, p = 0.004). The OSVH in random sample B, which represents the conduit that will be used, had a far better endothelium preservation and less endothelial detachment when compared to ESVH. We observed more endothelial detachment in the open CO2 ESVH group, due to lack of subcutaneous tissue separation, poor visualization and traction stress across the wall of the saphenous vein. The closed CO2 ESVH group had more endothelial cells preserved, but the OSVH group fared the best with the least number of endothelial cell detachment and a higher score of CD31 antibody. The online version contains supplementary material available at 10.1007/s12055-024-01752-3.

A Microscope Setup and Methodology for Capturing Hyperspectral and RGB Histopathological Imaging Databases.

The digitization of pathology departments in hospitals around the world is now a reality. The current commercial solutions applied to digitize histopathological samples consist of a robotic microscope with an RGB-type camera attached to it. This technology is very limited in terms of information captured, as it only works with three spectral bands of the visible electromagnetic spectrum. Therefore, we present an automated system that combines RGB and hyperspectral technology. Throughout this work, the hardware of the system and its components are described along with the developed software and a working methodology to ensure the correct capture of histopathological samples. The software is integrated by the controller of the microscope, which features an autofocus functionality, whole slide scanning with a stitching algorithm, and hyperspectral scanning functionality. As a reference, the time to capture and process a complete sample with 20 regions of high biological interest using the proposed method is estimated at a maximum of 79 min, reducing the time required by a manual operator by at least three times. Both hardware and software can be easily adapted to other systems that might benefit from the advantages of hyperspectral technology.

Diagnosis of Hirschsprung disease by analyzing acetylcholinesterase staining using artificial intelligence.

Classical Hirschsprung disease (HD) is defined by the absence of ganglion cells in the rectosigmoid colon. The diagnosis is made from rectal biopsy, which reveals the aganglionosis and the presence of cholinergic hyperinnervation. However, depending on the method of rectal biopsy, the quality of the specimens and the related diagnostic accuracy varies substantially. To facilitate and objectify the diagnosis of HD, we investigated whether software-based identification of cholinergic hyperinnervation in digitalized histopathology slides is suitable for distinguishing healthy individuals from affected individuals. N = 190 samples of 112 patients who underwent open surgical rectal biopsy at our pediatric surgery center between 2009 and 2019 were included in this study. Acetylcholinesterase (AChE) stained slides of these samples were collected and digitalized via slide scanning and analyzed using two digital imaging software programs (HALO, QuPath). The AChE-positive staining area in the mucosal layers of the intestinal wall was determined. In the next step machine learning was employed to identify patterns of cholinergic hyperinnervation. The area of AChE-positive staining was greater in HD patients compared to healthy individuals (p < 0.0001). Artificial intelligence-based assessment of parasympathetic hyperinnervation identified HD with a high precision (area under the curve [AUC] 0.96). The accuracy of the prediction model increased when nonrectal samples were excluded (AUC 0.993). Software-assisted machine-learning analysis of AChE staining is suitable to improve the diagnostic accuracy of HD.

Deep Learning Artificial Intelligence Predicts Homologous Recombination Deficiency and Platinum Response From Histologic Slides.

Cancers with homologous recombination deficiency (HRD) can benefit from platinum salts and poly(ADP-ribose) polymerase inhibitors. Standard diagnostic tests for detecting HRD require molecular profiling, which is not universally available. We trained DeepHRD, a deep learning platform for predicting HRD from hematoxylin and eosin (H&E)-stained histopathological slides, using primary breast (n = 1,008) and ovarian (n = 459) cancers from The Cancer Genome Atlas (TCGA). DeepHRD was compared with four standard HRD molecular tests using breast (n = 349) and ovarian (n = 141) cancers from multiple independent data sets, including platinum-treated clinical cohorts with RECIST progression-free survival (PFS), complete response (CR), and overall survival (OS) endpoints. DeepHRD predicted HRD from held-out H&E-stained breast cancer slides in TCGA with an AUC of 0.81 (95% CI, 0.77 to 0.85). This performance was confirmed in two independent primary breast cancer cohorts (AUC, 0.76 [95% CI, 0.71 to 0.82]). In an external platinum-treated metastatic breast cancer cohort, samples predicted as HRD had higher complete CR (AUC, 0.76 [95% CI, 0.54 to 0.93]) with 3.7-fold increase in median PFS (14.4 v 3.9 months; P = .0019) and hazard ratio (HR) of 0.45 (P = .0047). There were no significant differences in nonplatinum treatment outcome by predicted HRD status in three breast cancer cohorts, including CR (AUC, 0.39) and PFS (HR, 0.98, P = .95) in taxane-treated metastatic breast cancer. Through transfer learning to high-grade serous ovarian cancer, DeepHRD-predicted HRD samples had better OS after first-line (HR, 0.46; P = .030) and neoadjuvant (HR, 0.49; P = .015) platinum therapy in two cohorts. DeepHRD can predict HRD in breast and ovarian cancers directly from routine H&E slides across multiple external cohorts, slide scanners, and tissue fixation variables. When compared with molecular testing, DeepHRD classified 1.8- to 3.1-fold more patients with HRD, which exhibited better OS in high-grade serous ovarian cancer and platinum-specific PFS in metastatic breast cancer.

Global contrast-masked autoencoders are powerful pathological representation learners

Using digital pathology slide scanning technology, artificial intelligence algorithms, particularly deep learning, have achieved significant results in the field of computational pathology. Compared to other medical images, pathology images are more difficult to annotate, and thus, there is an extreme lack of available datasets for conducting supervised learning to train robust deep learning models. In this paper, we introduce a self-supervised learning (SSL) model, the Global Contrast-masked Autoencoder (GCMAE), designed to train encoders to capture both local and global features of pathological images and significantly enhance the performance of transfer learning across datasets. Our study demonstrates the capability of the GCMAE to learn transferable representations through extensive experiments on three distinct disease-specific hematoxylin and eosin (H&E)-stained pathology datasets: Camelyon16, NCT-CRC, and BreakHis. Moreover, we propose an effective automated pathology diagnosis process based on the GCMAE for clinical applications. The source code of this paper is publicly available at https://github.com/StarUniversus/gcmae.

Quantitative cellular pathology of the amygdala in temporal lobe epilepsy and correlation with magnetic resonance imaging volumetry, tissue microstructure, and sudden unexpected death in epilepsy risk factors.

Amygdala enlargement can occur in temporal lobe epilepsy, and increased amygdala volume is also reported in sudden unexpected death in epilepsy (SUDEP). Apnea can be induced by amygdala stimulation, and postconvulsive central apnea (PCCA) and generalized seizures are both known SUDEP risk factors. Neurite orientation dispersion and density imaging (NODDI) has recently provided additional information on altered amygdala microstructure in SUDEP. In a series of 24 surgical temporal lobe epilepsy cases, our aim was to quantify amygdala cellular pathology parameters that could predict enlargement, NODDI changes, and ictal respiratory dysfunction. Using whole slide scanning automated quantitative image analysis methods, parallel evaluation of myelin, axons, dendrites, oligodendroglia, microglia, astroglia, neurons, serotonergic networks, mTOR-pathway activation (pS6) and phosphorylated tau (pTau; AT8, AT100, PHF) in amygdala, periamygdala cortex, and white matter regions of interest were compared with preoperative magnetic resonance imaging data on amygdala size, and in 13 cases with NODDI and evidence of ictal-associated apnea. We observed significantly higher glial labeling (Iba1, glial fibrillary acidic protein, Olig2) in amygdala regions compared to cortex and a strong positive correlation between Olig2 and Iba1 in the amygdala. Larger amygdala volumes correlated with lower microtubule-associated protein (MAP2), whereas higher NODDI orientation dispersion index correlated with lower Olig2 cell densities. In the three cases with recorded PCCA, higher MAP2 and pS6-235 expression was noted than in those without. pTau did not correlate with SUDEP risk factors, including seizure frequency. Histological quantitation of amygdala microstructure can shed light on enlargement and diffusion imaging alterations in epilepsy to explore possible mechanisms of amygdala dysfunction, including mTOR pathway activation, that in turn may increase the risk for SUDEP.

Automated Whole Slide Imaging for Label-Free Histology Using Photon Absorption Remote Sensing Microscopy.

Pathologists rely on histochemical stains to impart contrast in thin translucent tissue samples, revealing tissue features necessary for identifying pathological conditions. However, the chemical labeling process is destructive and often irreversible or challenging to undo, imposing practical limits on the number of stains that can be applied to the same tissue section. Here we present an automated label-free whole slide scanner using a PARS microscope designed for imaging thin, transmissible samples. Peak SNR and in-focus acquisitions are achieved across entire tissue sections using the scattering signal from the PARS detection beam to measure the optimal focal plane. Whole slide images (WSI) are seamlessly stitched together using a custom contrast leveling algorithm. Identical tissue sections are subsequently H&E stained and brightfield imaged. The one-to-one WSIs from both modalities are visually and quantitatively compared. PARS WSIs are presented at standard 40x magnification in malignant human breast and skin samples. We show correspondence of subcellular diagnostic details in both PARS and H&E WSIs and demonstrate virtual H&E staining of an entire PARS WSI. The one-to-one WSI from both modalities show quantitative similarity in nuclear features and structural information. PARS WSIs are compatible with existing digital pathology tools, and samples remain suitable for histochemical, immunohistochemical, and other staining techniques. This work is a critical advance for integrating label-free optical methods into standard histopathology workflows.

Biological subgroups in melanoma: A correlation of tumor genomic and transcriptomic analysis with spatial immunophenotyping.

e14667 Background: A prominent challenge advancing immunotherapy is the heterogeneity of resistance mechanisms leading to drug development failure despite efficacy in subgroup of patients. To address these shortcomings, precision medicine platforms aimed at prospectively stratifying patients into treatment groups based on tumor mutational burden (TMB) and tumor inflammation scores by gene expression profiling (GEP) are now being explored. Whether these gene-based variables reflect the true immunologic state of the tumor remains an open question. Here, we sought to evaluate TMB and GEP as markers of intratumoral inflammatory state by single-cell spatial analysis. Methods: Under an IRB approved protocol through the Huntsman Cancer Institute, 25 melanoma tumors with available clinical, WES, and RNAseq data were stained with a multiplex immunofluorescence-based protocol using Opal dye and Leica Bond Rx (CD8, CD68, PD1, PD-L1, Ki67, SOX10, DAPI) and scanned with Akoya Polaris whole slide scanner. Images were analyzed using HALO. Results: Melanoma tumors were stratified into 4 subgroups based on TMB (&lt;10 vs ≥10 m/Mb) by WES, and inflammatory (Inflm) gene signature expression (GSEA &gt;0 vs ≤0) by RNAseq, denoted as S1 (TMBhi/Inflmhi; n=4), S2 (TMBhi/Inflmlo; n=4), S3 (TMBlo/Inflmhi; n=8), and S4 (TMBlo/Inflmlo; n=9). The rate of disease progression was 0% (S1), 25% (S2), 25% (S3), and 56% (S4). Infiltration analysis of CD8 T cells showed elevated intratumoral CD8 density in S1 (249/mm2) and S3 (157/mm2) vs S2 (65/mm2) and S4 (65/mm2; p=0.0002), consistent with GEP findings. Spatial analysis showed a similar pattern of CD8 distribution across subgroups with peak CD8 density in the tumor periphery. Proximity analyses showed average distance of CD8 T cell to the nearest tumor cell was not different across subgroups. Furthermore, we found Inflmhi tumors had a higher frequency of PD1+ CD8 T cells relative to Inflmlo tumors (12.9 vs 2.8 cells/HPF; p=0.0002), suggesting GEP also reflects CD8 activation and functional status. Increased frequency of PD1+ CD8 T cells was correlated with a much improved overall survival (HR=3.3; p=0.017). A similar analysis performed on CD68+ macrophages showed significantly increased peri-tumoral macrophage recruitment in S4 (445/mm2) vs S1 (209/mm2; p&lt;0.0001), S2 (91/mm2; p&lt;0.0001), and S3 (267/mm2; p=0.0003) subgroups. Interestingly, Inflmhi tumors were associated with an increased proportion of PDL1+ macrophages regardless of TMB status (43% vs 3%; p&lt;0.0001), suggestive of an upregulated adaptive immune response in tumors with active inflammation. Conclusions: Our findings illustrate that biological groupings based on TMB and GEP biomarkers correlate with CD8 spatial location and activation state. Each subgroup has unique tumor immune features that can be related to immune resistance and warrant further investigation.

Canine Cerebrospinal Fluid Analysis Using Two New Automated Techniques: The Sysmex XN-V Body Fluid Mode and an Artificial-Intelligence-Based Algorithm.

Cerebrospinal fluid analysis is an important diagnostic test when assessing a neurological canine patient. For this analysis, the total nucleated cell count and differential cell counts are routinely taken, but both involve time-consuming manual methods. To investigate faster automated methods, in this study, the Sysmex XN-V body fluid mode and the deep-learning-based algorithm generated by the Olympus VS200 slide scanner were compared with the manual methods in 161 canine cerebrospinal fluid samples for the total nucleated cell count and in 65 samples with pleocytosis for the differential counts. Following incorrect gating by the Sysmex body fluid mode, all samples were reanalyzed with manually set gates. The Sysmex body fluid mode then showed a mean bias of 15.19 cells/μL for the total nucleated cell count and mean biases of 4.95% and -4.95% for the two-part differential cell count, while the deep-learning-based algorithm showed mean biases of -7.25%, -0.03% and 7.27% for the lymphocytes, neutrophils and monocytoid cells, respectively. Based on our findings, we propose that the automated Sysmex body fluid mode be used to measure the total nucleated cell count in canine cerebrospinal fluid samples after making adjustments to the predefined settings from the manufacturer. However, the two-part differential count of the Sysmex body fluid mode and the deep-learning-based algorithm require some optimization.

Prediction of Mismatch Repair Status in Endometrial Cancer from Histological Slide Images Using Various Deep Learning-Based Algorithms.

The application of deep learning algorithms to predict the molecular profiles of various cancers from digital images of hematoxylin and eosin (H&E)-stained slides has been reported in recent years, mainly for gastric and colon cancers. In this study, we investigated the potential use of H&E-stained endometrial cancer slide images to predict the associated mismatch repair (MMR) status. H&E-stained slide images were collected from 127 cases of the primary lesion of endometrial cancer. After digitization using a Nanozoomer virtual slide scanner (Hamamatsu Photonics), we segmented the scanned images into 5397 tiles of 512 × 512 pixels. The MMR proteins (PMS2, MSH6) were immunohistochemically stained, classified into MMR proficient/deficient, and annotated for each case and tile. We trained several neural networks, including convolutional and attention-based networks, using tiles annotated with the MMR status. Among the tested networks, ResNet50 exhibited the highest area under the receiver operating characteristic curve (AUROC) of 0.91 for predicting the MMR status. The constructed prediction algorithm may be applicable to other molecular profiles and useful for pre-screening before implementing other, more costly genetic profiling tests.

A beginner’s guide to immunohistochemistry

Using the affinity between antibodies and antigens, immunohistochemistry (IHC) is a helpful method for analysing the cellular components within tissue sections placed on glass slides. The chemistry of antibodies was not discovered until 1941, and since then, their use has increased due to the discovery of enzymes and fluorochromes conjugated to antibodies, enabling observations through optical or fluorescence microscopy (FM). Hence, immunofluorescence (IF), or IHC with fluorescent antibodies, has been associated with the advancement of FM and fluorochromes. Specifically, the discovery of three fluorochromes that emit blue, red and green led to a rise in IF application in the context of visualizing numerous components in a single cell. The proximity ligation assay (PLA) method, so termed because the antibody is conjugated with short DNA strands called PLA probes, is the most recent development in the field of antibody conjugation. Protein–protein interactions and sub-localization of proteins can both be studied using this technique. The improvement of conjugation techniques has occurred in parallel with the development of sophisticated microscopies, including confocal and slide scanners. Confocal microscopy allows for the reconstruction of three-dimensional structures while a slide scanner can digitally convert up to 100 slides' worth of high-resolution pictures. The first step in the IHC process is the collection and fixation of tissues from animal models and humans. Paraformaldehyde is the most commonly used fixative. Fixation prevents antigen blocking and creates cross-links between proteins to preserve tissue structures. Fixation is followed by cutting tissues using microtome or cryostat, then slides are treated for antigen exposition (antigen retrieval) and then included in Triton X solution, which increases membrane permeability. Subsequently, incubations with enzyme- or fluorochrome-conjugated antibodies enable observations via optical microscopy or FM, respectively. In clinical practice and research, IHC and IF are still helpful methods for identifying the antigen-proteins linked to specific diseases.

Abstract PO2-14-12: Accurate quantification of slide-level HER2 scores in breast cancer using a machine-learning model, AIM-HER2 Breast Cancer

Abstract Background: HER2 expression level is a key factor in determining the optimal treatment course for breast cancer patients. Roughly 15% of breast cancers are HER2(+), and determination of HER2 status is routinely assessed by immunohistochemistry (IHC). Accurate assessment of the HER2 IHC score (0, 1+, 2+, 3+) by pathologists is therefore critical, especially in light of novel therapeutic approaches demonstrating efficacy in the HER2-low setting (IHC scores 1+, and 2+/FISH-)1,2. To assist pathologists with the consistent provision of reproducible and accurate scores across the entire HER2 scoring range, we developed a machine-learning model (“AIM-HER2”) to generate accurate, slide-level HER2 scores aligned with ASCO-CAP guidelines in clinical breast cancer HER2 IHC specimens. Methods: AIM-HER2 was developed using whole-slide images (WSI; N=4261) from clinical and commercial sources. WSI were split into training (N=2694, 63%) and optimization (N=1567, 37%) sets. An additive multiple instance learning (aMIL) model3 was trained to predict HER2 scores directly from WSI and create interpretable heatmaps that depict HER2 predictions in tissue images. Image artifacts and in situ carcinomas were identified using previously trained artifact and tissue segmentation models and were excluded, leaving only regions of invasive carcinoma to be analyzed. AIM-HER2 performance was assessed on additional slides obtained from five academic or commercial sources (N=804 total, 770 evaluable) on which HER2 IHC was performed. Board-certified pathologists (N=52) with relevant experience provided manual HER2 scores based on ASCO-CAP guidelines. Nested pairwise non-inferiority analysis4 was used to compare model performance to that of pathologists (N=3 pathologists per slide). In the nested pairwise framework, agreement among pathologists was compared to agreement between AIM-HER2 and pathologists via linear kappa, so that summary metrics account for inter-pathologist variability. Results: High concordance was observed between AIM-HER2-predicted and pathologist-labeled slide-level HER2 scores, both overall and for each scoring level. Similar results were observed when assessing AIM-HER2 performance on multiple slide scanners and after IHC with multiple HER2 IHC antibody clones. Results are summarized in Table 1. Conclusions: We developed AIM-HER2, a novel aMIL-based approach for predicting slide-level HER2 IHC scores. AIM-HER2 has similar levels of agreement with pathologists as pathologists have with each other for determining HER2 score. This result is upheld when slides imaged using multiple scanning platforms and stained using multiple HER2 antibody clones. The performance of AIM-HER2 on multiple scanners and after multiple assays supports broad applicability of this algorithm in clinical laboratories, including for the identification of HER2-low cases. Work is ongoing to perform similar analyses in an independent, real-world dataset.

Abstract PO5-06-06: Tumor/stromal expression of CD3/CD8/PD-1/PD-L1 and overall survival (OS) in patients (pts) with metastatic (met) HER2+ breast cancer (BC)

Abstract Background: The “Thousand Patient HER-2 database” project at Saint Vincent’s University Hospital (SVUH) Dublin contains a cohort of met HER2+ BC pts treated with trastuzumab, with OS ranging from short term to durable complete response (never relapsed). While the majority of pts progress on treatment, higher stromal tumor immune infiltrate has been associated with longer OS in met HER2+ BC. Using the SVUH database, we have identified a cohort of met HER2+ BC pts that received trastuzumab and have a broad OS range (0.3 months (mos) to 200.9 mos). This study examines the tumor and stromal levels of pan T cell marker CD3, cytotoxic T cell marker CD8, and immune checkpoints PD-L1 and PD-1 by immunohistochemistry (IHC) in primary and met samples from this cohort. Methods: Clinico-pathological data was available for formalin-fixed, paraffin-embedded biopsy specimens (n=45 primary, n=25 matched met biopsies). IHC staining was conducted for CD3 (Agilent IR50361-2), CD8 (Agilent IR62361-2), PD-L1, (Agilent M365329-2), and PD-1 (Roche 07099029001). A DAKO Link 48 Autostainer was utilised for staining, incorporating positive (tonsil tissue) and negative controls (isotype controls). Slide processing used an Aperio AT2 Digital Slide Scanner (Leica Biosystems), Aperio ImageScope 12.4 software (Leica Biosystems) and QuPath analysis software (University of Edinburgh). Following annotation of tumor areas, an algorithm was trained to identify immune cells and designate them as tumor or stromal. Data was expressed as the number of positively stained cells/mm2 tissue. A median cut-off was applied to denote “high” and “low” expression for CD3 and CD8. For PD-1 and PD-L1, samples with ≥ 1 stained cell/mm2 were designated as positive (pos) and samples with zero stained cells as negative (neg). The Kaplan Meier method was utilised for survival studies. A paired Student’s t test was utilised for primary vs metastatic site comparisons. Results: 51.1% of primary samples displayed high CD3 (23/45) and CD8 (23/45) expression in both the stromal and tumor compartments. Within the tumor compartment, 18.6% (8/43) of samples were pos for PD-L1 expression, and 57.8% (26/45) were pos for PD-1 expression. In the stromal compartment, 30.2% (13/43) of samples were pos for PD-L1 expression, and 68.9% (31/45) were pos for PD-1 expression. High tumor levels of CD3 (median OS CD3 high 113.4 mos vs CD3 low 7.9 mos, HR 0.2938 (95% CI 0.130-0.618), p=0.0012) and CD8 (median OS CD8 high 73.9 mos vs CD8 low 10.8 mos, HR 0.4764 (95% CI 0.238-0.954), p=0.0365), but not stromal levels of these markers, were associated with improved OS. There were no significant differences in CD3, CD8, PD-1 and PD-L1 levels between matched primary and met samples (p &amp;gt;0.05). Tumor, but not stromal, PD-L1 expression was associated with longer OS (HR 0.3120 (95% CI 0.146-0.667), p=0.0027). Interestingly, stromal PD-1 expression, but not tumor PD-1 expression, was also associated with longer OS (median OS PD-1 pos 64.7 mos vs. PD-1 neg 5.6 mos, HR 0.158 (95% CI 0.060-0.411), p=0.0002). Conclusion: Our results report a link between OS and tumor (CD3+, CD8+ and PD-L1+)/stromal (PD-1+) immune cell infiltrate in met HER2+ BC patients treated with trastuzumab. Further expansion of this preliminary dataset is warranted. Citation Format: Denis Collins, Janet McCormack, Laura Ivers, Jose Javier Berenguer-Pina, Jo Ballot, Cecily Quinn, Darko Skrobo, Alex Eustace, Naomi Walsh, Aurelie Fabre, John Crown. Tumor/stromal expression of CD3/CD8/PD-1/PD-L1 and overall survival (OS) in patients (pts) with metastatic (met) HER2+ breast cancer (BC) [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-06-06.

Abstract PO3-13-11: Quantitative Multiplex Immunofluorescence Assay for TROP2 and HER2 Expression in Breast Cancer: Towards Guiding Patient Selection for Antibody Drug Conjugate Therapies

Abstract Emerging antibody drug conjugate (ADC) therapies targeting human trophoblast cell-surface antigen (TROP2) and human epidermal growth factor receptor 2 (HER2) are transforming the treatment landscape for breast cancer. Sacituzumab govitecan (SG) and trastuzumab deruxtecan (T-DXd) have gained approval for an overlapping set of "HER2-low" metastatic breast cancers, including hormone receptor (HR)-positive HER2 non-amplified and "triple-negative" subtypes. Nevertheless, the optimal selection of patients and treatment sequencing for these ADC therapies remains a clinical challenge. Clinical trial objective response rates to SG are approximately 30%, compared to 30-90% for T-DXd depending on HER2 expression levels. While both drugs are thought to be targeted therapies, the value of measuring the target and the best methods to do so are still not established. We believe that quantitative measurement of TROP2 and HER2 antigen expression levels could establish thresholds for responders, enabling more effective patient selection for ADC therapies. Here, we present a TROP2, high-sensitivity HER2, and cytokeratin (CK) quantitative immunofluorescence (QIF) multiplex assay. Using a ten-cell line standard array and proteomic mass spectrometry, we can convert tumor compartment QIF intensity to protein concentration in fmol/mm2 for tissue specimens. Anti-HER2, anti-TROP2 antibodies, and fluorescence detection systems were titrated and combined to maximize signal-to-noise ratio on our cell line standard array and breast cancer tissue microarrays (TMA). The multiplex assay was designed for automated slide stainers (Leica BOND Rx) and fluorescence slide scanning (Rarecyte CyteFinder II HT). We perform our analyses in QuPath using an image processing plugin developed for automated QIF/IHC analysis (Qymia). Reproducible TROP2 and HER2 QIF scoring (R² &amp;gt; 0.95) was achieved across multiple staining batches using serial sections of breast cancer TMAs and cell standard arrays. This assay has a TROP2 linear range between 0.63 - 9.17 fmol/mm2 (about 1 million TROP2 receptors/cell) and HER2 linear range between 0.09 - 0.565 fmol/mm2 (about 60,000 HER2 receptors/cell). We then applied this multiplex assay to two serial retrospective primary breast cancer cohorts from Yale University to quantitatively measure TROP2 and HER2 expression (338 clinical cases). We find a weak negative correlation between TROP2 and HER2 expression in our breast cancer cohorts (Pearson r = -0.14, p = 0.0097, n = 338). TROP2 expression levels were above the limit of detection (LOD) in 90.2% of cases, with 4.1% exceeding the limit of linearity (LOL), and a mean TROP2 expression of 4.05 fmol/mm2. For HER2, 67.2% of cases were above the LOD, with 7.1% exceeding the LOL, and a mean HER2 expression of 0.186 fmol/mm2. Both TROP2 and HER2 were below the LOD in 3.0% of cases, which we define as “negative”. We found 29.9% expressed TROP2 and were HER2-negative, and 6.8% expressed HER2 and were TROP2-negative. Our future studies will aim to quantitatively define expression thresholds for T-DXd and SG response with the goal to produce a clinical grade assay for ADC patient selection and determine the value of the assay to help select which ADC to give first. HER2 and TROP2 protein expression summary in Yale breast cancer cohort Table 1: Summary of HER2 and TROP2 protein expression levels in serial retrospective primary breast cancer cohort of 338 cases using our high-sensitivity HER2 and TROP2 multiplex immunofluorescent assay Citation Format: Charles Robbins, Mengni He, Revekka Khaimova, Katherine Bates, Nay Nwe Nyein Chan, Daniel Liebler, Regan Fulton, David Rimm. Quantitative Multiplex Immunofluorescence Assay for TROP2 and HER2 Expression in Breast Cancer: Towards Guiding Patient Selection for Antibody Drug Conjugate Therapies [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO3-13-11.