Developing image analysis methods for digital pathology

QuPath Edu and OpenMicroanatomy: Open-source virtual microscopy tools for medical education

Virtual microscopy is becoming increasingly common in both medical education and routine clinical practice. Virtual microscopy software is typically designed either for (1) training students in anatomy, histology, and histopathology, or (2) quantitative analysis-but not both simultaneously. QuPath is one of the most widely used software applications for histopathology image analysis in research and provides a comprehensive set of computational tools to evaluate histology slides. We have enhanced QuPath by developing a new extension, QuPath Edu, which adapts the software to function as an intuitive microanatomy learning environment. Additionally, we have created an entirely new, complementary software platform called OpenMicroanatomy, which provides an alternative way to access QuPath Edu teaching content through a web interface. These tools have been used in teaching of first year medical and dentistry students at the University of Oulu Medical Faculty, and we conducted a user survey for the Class of 2023 to assess the usability and student experience. In general, the introduced annotation and quiz features were appreciated by the students and the system usability of OpenMicroanatomy was considered excellent (SUS score 84.8). Together, these freely available tools enable teachers to develop and deploy innovative training material for anatomy, histopathology, quantitative analysis, and artificial intelligence in a wide range of contexts. This unique combination can provide the next generation of students with essential multidisciplinary skills.

Digital pathology in tau research: A comparison of QuPath and HALO

The application of digital pathology tools has expanded in recent years, but non-neoplastic human brain tissue presents unique challenges due to its complexity. This study evaluated HALO and QuPath tau quantification performance in the hippocampus and mid-frontal gyrus across various tauopathies. Percent positivity emerged as the most reliable measure, showing strong correlations with Braak stages and CERAD scores, outperforming object and optical densities. QuPath demonstrated superior correlations with Braak stages, while HALO excelled in aligning with CERAD scoring. However, HALO's optical density was less consistent. Paired t-tests revealed significant differences in object and optical densities between platforms, though percent positivity was consistent across both. QuPath's threshold-based object density showed similar agreement with manual counts compared to HALO's AI-dependent approach (all ρ > 0.70). Reanalysis of QuPath further improved its agreement with manual measurements and correlations with Braak and CERAD scores (all ρ > 0.70). HALO offers a user-friendly interface and excels in certain metrics but is hindered by frequent software malfunctions and more limited flexibility. In contrast, QuPath's customizable workflows and superior performance in Braak staging make it more suitable for advanced and larger-scale analyses. Overall, our study highlights the strengths and limitations of these platforms, helping guide their application in neuropathology.

Integrated histopathology of the human pancreas throughout stages of type 1 diabetes progression

Type 1 diabetes (T1D) is a progressive autoimmune condition that culminates in the loss of insulin-producing beta cells. Pancreatic histopathology provides essential insights into disease initiation and progression yet an integrated perspective of in situ pathogenic processes is lacking due to limited sample availability, the dispersed nature of anatomical lesions, and often restricted analytical dimensionality. Here, we combined multiplexed immunostaining, high-magnification whole-slide imaging, digital pathology, and semi-automated image analysis strategies to interrogate pancreatic tail and head regions obtained from organ donors across T1D stages including at-risk and at-onset cases. Deconvolution of architectural features, endocrine cell composition, immune cell burden, and spatial relations of ∼25,000 islets revealed a series of novel histopathological correlates especially in the prodromal disease stage preceding clinical T1D. Altogether, our comprehensive "single-islet" analyses permit the reconstruction of a revised natural T1D history with implications for further histopathological investigations, considerations of pathogenetic modalities, and therapeutic interventions.

Deep learning-based image analysis in muscle histopathology using photo-realistic synthetic data

BACKGROUND

Artificial intelligence (AI), specifically Deep learning (DL), has revolutionized biomedical image analysis, but its efficacy is limited by the need for representative, high-quality large datasets with manual annotations. While latest research on synthetic data using AI-based generative models has shown promising results to tackle this problem, several challenges such as lack of interpretability and need for vast amounts of real data remain. This study aims to introduce a new approach-SYNTA-for the generation of photo-realistic synthetic biomedical image data to address the challenges associated with state-of-the art generative models and DL-based image analysis.

METHODS

The SYNTA method employs a fully parametric approach to create photo-realistic synthetic training datasets tailored to specific biomedical tasks. Its applicability is tested in the context of muscle histopathology and skeletal muscle analysis. This new approach is evaluated for two real-world datasets to validate its applicability to solve complex image analysis tasks on real data.

RESULTS

Here we show that SYNTA enables expert-level segmentation of unseen real-world biomedical data using only synthetic training data. By addressing the lack of representative and high-quality real-world training data, SYNTA achieves robust performance in muscle histopathology image analysis, offering a scalable, controllable and interpretable alternative to generative models such as Generative Adversarial Networks (GANs) or Diffusion Models.

CONCLUSIONS

SYNTA demonstrates great potential to accelerate and improve biomedical image analysis. Its ability to generate high-quality photo-realistic synthetic data reduces reliance on extensive collection of data and manual annotations, paving the way for advancements in histopathology and medical research.

Area of Residual Tumor Measurement After Preoperative Chemotherapy as an Objective and Quantitative Method for Predicting the Prognosis of Gastric Cancer: A Single-Center Retrospective Study

BACKGROUND

Pathological regression grade after chemotherapy evaluated by surgically resected specimens is closely related with prognosis. Since usefulness of measuring the area of the residual tumor (ART) has been reported, this study aimed to evaluate the utility of ART in predicting the prognosis of patients with gastric cancer (GC) who received preoperative chemotherapy.

METHODS

This single-center retrospective study examined the relationship between ART and survival outcomes. We included 92 patients who underwent preoperative chemotherapy followed by radical surgery for GC. Digital images were used to measure the ART in the largest pathological slice of each patient's surgical tumor specimen. We simply subclassified the patients as either ART-0 (< 0.1 mm2 or carcinoma in situ) or non-ART-0 to compare the prognoses.

RESULTS

Significant differences were noted in overall survival and recurrence-free survival (RFS) between ART-0 (n = 19) and non-ART-0 (n = 73). The survival curves were similar to those of major pathological response (MPR) (n = 24) or non-MPR (n = 68), which are commonly used as surrogate endpoint presently. Multivariate analysis revealed ART and ypN independent prognostic factors for RFS. Survival curves stratified using ART and ypN to indicate risk grades (low-, moderate-, or high-) were not significantly different from those stratified using the other three existing pathological regression grade systems and ypN.

CONCLUSION

ART-based pathological assessment is a simple and useful method for predicting the prognosis in patients with GC who underwent radical surgery after chemotherapy.

Digital Pathology as an Innovative Tool for Improving Cancer Diagnosis and Treatment

For more than a century, the "gold" standard for diagnosing malignant neoplasms has been pathohistology. However, the continuous advancement of modern technologies is leading to a radical transformation of this field and the emergence of digital pathology. The main advantages of digital pathology include the convenience of the data storage and transfer, as well as the potential for automating diagnostic processes through the application of artificial intelligence technologies. Integrating digital pathology into clinical practice is expected to accelerate the analysis of histological samples, reduce the costs associated with such procedures, and enable the accumulation of large datasets for future scientific research. At the same time, the development of digital pathology faces certain challenges such as the need for technical upgrades in laboratories, ensuring data cybersecurity, and training qualified personnel.

Neuropathological stages of neuronal, astrocytic and oligodendrocytic alpha-synuclein pathology in Parkinson's disease

Alpha-synucleinopathies are neurodegenerative diseases characterized by the spread of alpha-synuclein (α-syn) aggregates throughout the central nervous system in a stereotypical manner. These diseases include Lewy body disease (LBD), which encompass Dementia with Lewy bodies (DLB), Parkinson's Disease (PD), and Parkinson's Disease Dementia (PDD), and Multiple System Atrophy (MSA). LBD and MSA chiefly contain α-syn aggregates in neurons and oligodendrocytes, respectively, although glial α-syn pathology in LBD is increasingly being recognized. Semi-quantitative and machine learning-based quantifications of neuronal, oligodendrocytic and astrocytic α-syn pathology were implemented on a cohort of LBD and MSA post-mortem tissue samples. The neuroanatomical distribution of each cell-type specific α-syn pathology was evaluated using conditional probability matrices and Subtype and Stage Inference (SuStaIn) algorithm. We revealed extensive glial α-syn pathology in LBD, emphasizing the disease- and region-specific profile of astrocytic α-syn pathology, which was absent in MSA and minimal in the substantia nigra of LBD. Furthermore, we have described distinct morphologies of astrocytic α-syn pathology, which were found to correlate with the density of astrocytic α-syn inclusions. Astrocytic α-syn pathology was mainly centered in the amygdala and exhibited a unique stereotypical progression whilst oligodendrocytes displayed a distribution akin to the established neuronal progression pattern. SuStaIn modeling was further used to test for heterogeneity in the spatiotemporal progression, revealing that a subset of cases might follow an alternative pattern. Based on these findings, we introduce a novel multimodal progression framework that integrates, for the first time, the temporal and spatial progression of astrocytic and oligodendrocytic α-syn pathology alongside neuronal pathology in PD, providing further information regarding the role of neurons and glia in disease pathogenesis.

Using digital pathology to standardize and automate histological evaluations of environmental samples

Histological evaluations of tissues are commonly used in environmental monitoring studies to assess the health and fitness status of populations or even whole ecosystems. Although traditional histology can be cost-effective, there is a shortage of proficient histopathologists and results can often be subjective between operators, leading to variance. Digital pathology is a powerful diagnostic tool that has already significantly transformed research in human health but has rarely been applied to environmental studies. Digital analyses of whole slide images introduce possibilities of highly standardized histopathological evaluations, as well as the use of artificial intelligence for novel analyses. Furthermore, incorporation of digital pathology into environmental monitoring studies using standardized bioindicator species or groups such as bivalves and fish can greatly improve the accuracy, reproducibility, and efficiency of the studies. This review aims to introduce readers to digital pathology and how it can be applied to environmental studies. This includes guidelines for sample preparation, potential sources of error, and comparisons to traditional histopathological analyses.

Use of a Pathomics Nomogram to Predict Postoperative Liver Metastasis in Patients with Stage III Colorectal Cancer

BACKGROUND

Approximately 25% of patients with stage III colorectal cancer experience liver metastasis after radical resection; however, there is currently a lack of methods to predict liver metastasis. This study aims to develop and validate a pathomics nomogram to predict liver metastasis in patients with stage III colorectal cancer.

METHODS

A total of 318 enrolled patients were divided into three cohorts: a training cohort (n = 139), a validation cohort (n = 69), and an external cohort (n = 110). A competitive risk nomogram was established by the pathomics signature and clinicopathological characteristics and assessed by calibration, discrimination, and clinical usefulness.

RESULTS

A significant correlation between the pathomics signature and liver metastasis in stage III colorectal cancer was found. Multivariate Fine-Gray analysis indicated that preoperative carcinoembryonic antigen level, postoperative chemotherapy, and pathomics signature were independent predictors of liver metastasis. A competitive risk nomogram was developed to predict liver metastasis in patients with stage III colorectal cancer. The predicting nomogram shows good discrimination and calibration, with C-indexes of 0.811 (95% confidence interval [CI] 0.651-0.971), 0.759 (95% CI 0.531-0.987), and 0.845 (95% CI 0.641-0.999), with area under the receiver operating characteristic (AUROC) curves at 5 years of 0.833 (95% CI 0.742-0.925), 0.760 (95% CI 0.652-0.893), and 0.812 (95% CI 0.692-0.931) in the training, validation, and external cohorts, respectively. Compared with the clinicopathological nomogram, the nomogram combined with the pathomics signature had better performance (AUROC 0.823 [95% CI 0.764-0.881] vs. 0.678 [95% CI 0.606-0.751]; p < 0.001).

CONCLUSIONS

The pathomics signature is a predictive indicator for liver metastasis in patients with stage III colorectal cancer, and the integrated nomogram can be used to predict liver metastasis better than the clinicopathological nomogram alone.

Analysis of gene and protein expression in the endometrium for validation of an ex vivo model of the equine uterus using PCR, digital and visual histopathology

In the past, most research in equine reproduction has been performed in vivo but the use of in vitro and ex vivo models has recently increased. This study aimed to evaluate the functional stability of an ex vivo hemoperfused model for equine uteri with molecular characterization of marker genes and their proteins. In addition, the study validated the respective protein expression and the aptness of the software QuPath for identifying and scoring immunohistochemically stained equine endometrium. After collection, uteri (n = 12) were flushed with preservation solution, transported to the laboratory on ice, and perfused with autologous blood for 6 h. Cycle stage was determined by examination of the ovaries for presence of Graafian follicles or corpora lutea and analysis of plasma progesterone concentration (estrus: n = 4; diestrus: n = 4; anestrus: n = 4). Samples were obtained directly after slaughter, after transportation, and during perfusion (240, 300, 360 min). mRNA expression levels of progesterone (PGR), estrogen (ESR1) and oxytocin (OXTR) receptor as well as of MKI67 (marker of cell growth) and CASP3 (marker of apoptosis) were analyzed by RT-qPCR, and correlation to protein abundance was validated by immunohistochemical staining. Endometrial samples were analyzed by visual and computer-assisted evaluation of stained antigens via QuPath. For PGR, effects of the perfusion and cycle stage on expression were found (P < 0.05), while ESR1 was affected only by cycle stage (P < 0.05) and OXTR was unaffected by perfusion and cycle stage. MKI67 was lower after 360 min of perfusion as compared to samples collected before perfusion (P < 0.05). For CASP3, differences in gene expression were found after transport and samples taken after 240 min (P < 0.05). Immunohistochemical staining revealed effects of perfusion on stromal and glandular cells for steroid hormone receptors, but not for Ki-67 and active Caspase 3. OXTR was visualized in all layers of the endometrium and was unaffected by perfusion. Comparison of QuPath and visual analysis resulted in similar results. For most cell types and stained antigens, the correlation coefficient was r > 0.5. In conclusion, the isolated hemoperfused model of the equine uterus was successfully validated at the molecular level, demonstrating stability of key marker gene expression. The utility of computer-assisted immunohistochemical analysis of equine endometrial samples was also confirmed.

Combination of pathological, biochemical and behavioral evaluations for peripheral neurotoxicity assessment in isoniazid-treated rats

In drug development, assessment of non-clinical peripheral neurotoxicity is important to ensure human safety. Clarifying the pathological features and mechanisms of toxicity enables the management of safety risks in humans by estimating the degree of risk and proposing monitoring strategies. Published guidelines for peripheral neurotoxicity assessment do not provide detailed information on which endpoints should be monitored preferentially and how the results should be integrated and discussed. To identify an optimal assessment method for the characterization of peripheral neurotoxicity, we conducted pathological, biochemical (biomaterials contributing to mechanistic considerations and biomarkers), and behavioral evaluations of isoniazid-treated rats. We found a discrepancy between the days on which marked pathological changes were noted and those on which biochemical and behavioral changes were noted, suggesting the importance of combining these evaluations. Although pathological evaluation is essential for pathological characterization, the results of biochemical and behavioral assessments at the same time points as the pathological evaluation are also important for discussion. In this study, since the measurement of serum neurofilament light chain could detect changes earlier than pathological examination, it could be useful as a biomarker for peripheral neurotoxicity. Moreover, examination of semi-thin specimens and choline acetyltransferase immunostaining were useful for characterizing morphological neurotoxicity, and image analysis of semi-thin specimens enabled us to objectively show the pathological features.

Glioma subtype classification from histopathological images using in-domain and out-of-domain transfer learning: An experimental study

We provide in this paper a comprehensive comparison of various transfer learning strategies and deep learning architectures for computer-aided classification of adult-type diffuse gliomas. We evaluate the generalizability of out-of-domain ImageNet representations for a target domain of histopathological images, and study the impact of in-domain adaptation using self-supervised and multi-task learning approaches for pretraining the models using the medium-to-large scale datasets of histopathological images. A semi-supervised learning approach is furthermore proposed, where the fine-tuned models are utilized to predict the labels of unannotated regions of the whole slide images (WSI). The models are subsequently retrained using the ground-truth labels and weak labels determined in the previous step, providing superior performance in comparison to standard in-domain transfer learning with balanced accuracy of 96.91% and F1-score 97.07%, and minimizing the pathologist's efforts for annotation. Finally, we provide a visualization tool working at WSI level which generates heatmaps that highlight tumor areas; thus, providing insights to pathologists concerning the most informative parts of the WSI.

TGMIL: A hybrid multi-instance learning model based on the Transformer and the Graph Attention Network for whole-slide images classification of renal cell carcinoma

BACKGROUND AND OBJECTIVES

The pathological diagnosis of renal cell carcinoma is crucial for treatment. Currently, the multi-instance learning method is commonly used for whole-slide image classification of renal cell carcinoma, which is mainly based on the assumption of independent identical distribution. But this is inconsistent with the need to consider the correlation between different instances in the diagnosis process. Furthermore, the problem of high resource consumption of pathology images is still urgent to be solved. Therefore, we propose a new multi-instance learning method to solve this problem.

METHODS

In this study, we proposed a hybrid multi-instance learning model based on the Transformer and the Graph Attention Network, called TGMIL, to achieve whole-slide image of renal cell carcinoma classification without pixel-level annotation or region of interest extraction. Our approach is divided into three steps. First, we designed a feature pyramid with the multiple low magnifications of whole-slide image named MMFP. It makes the model incorporates richer information, and reduces memory consumption as well as training time compared to the highest magnification. Second, TGMIL amalgamates the Transformer and the Graph Attention's capabilities, adeptly addressing the loss of instance contextual and spatial. Within the Graph Attention network stream, an easy and efficient approach employing max pooling and mean pooling yields the graph adjacency matrix, devoid of extra memory consumption. Finally, the outputs of two streams of TGMIL are aggregated to achieve the classification of renal cell carcinoma.

RESULTS

On the TCGA-RCC validation set, a public dataset for renal cell carcinoma, the area under a receiver operating characteristic (ROC) curve (AUC) and accuracy of TGMIL were 0.98±0.0015,0.9191±0.0062, respectively. It showcased remarkable proficiency on the private validation set of renal cell carcinoma pathology images, attaining AUC of 0.9386±0.0162 and ACC of 0.9197±0.0124. Furthermore, on the public breast cancer whole-slide image test dataset, CAMELYON 16, our model showed good classification performance with an accuracy of 0.8792.

CONCLUSIONS

TGMIL models the diagnostic process of pathologists and shows good classification performance on multiple datasets. Concurrently, the MMFP module efficiently diminishes resource requirements, offering a novel angle for exploring computational pathology images.

Hourglass, a rapid analysis framework for heterogeneous bioimaging data, identifies sex disparity in IL-6/STAT3-associated immune phenotypes in pancreatic cancer

Integration and mining of bioimaging data remains a challenge and lags behind the rapidly expanding digital pathology field. We introduce Hourglass, an open-access analytical framework that streamlines biology-driven visualization, interrogation, and statistical assessment of multiparametric datasets. Cognizant of tissue and clinical heterogeneity, Hourglass systematically organizes observations across spatial and global levels and within patient subgroups. Applied to an extensive bioimaging dataset, Hourglass promptly consolidated a breadth of known interleukin-6 (IL-6) functions via its downstream effector STAT3 and uncovered a so-far unknown sexual dimorphism in the IL-6/STAT3-linked intratumoral T-cell response in human pancreatic cancer. As an R package and cross-platform application, Hourglass facilitates knowledge extraction from multi-layered bioimaging datasets for users with or without computational proficiency and provides unique and widely accessible analytical means to harness insights hidden within heterogeneous tissues at the sample and patient level. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Comparison of FNA-based conventional cytology specimens and digital image analysis in assessment of pancreatic lesions

Objectives

Endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is one of the most important diagnostic tools for investigation of suspected pancreatic masses, although the interpretation of the results is controversial. In recent decades, digital image analysis (DIA) has been considered in pathology. The aim of this study was to assess the DIA in the evaluation of EUS-FNA based cytopathological specimens of pancreatic masses and comparing it with conventional cytology analysis by pathologist.

Material and Methods

This study was performed using cytological slides related to EUS-FNA samples of pancreatic lesions. The digital images were prepared and then analyzed by ImageJ software. Factors such as perimeter, circularity, area, minimum, maximum, mean, median of gray value, and integrated chromatin density of cell nucleus were extracted by software ImageJ and sensitivity, specificity, and cutoff point were evaluated in the diagnosis of malignant and benign lesions.

Results

In this retrospective study, 115 cytology samples were examined. Each specimen was reviewed by a pathologist and 150 images were prepared from the benign and malignant lesions and then analyzed by ImageJ software and a cut point was established by SPSS 26. The cutoff points for perimeter, integrated density, and the sum of three factors of perimeter, integrated density, and circularity to differentiate between malignant and benign lesions were reported to be 204.56, 131953, and 24643077, respectively. At this cutting point, the accuracy of estimation is based on the factors of perimeter, integrated density, and the sum of the three factors of perimeter, integrated density, and circularity were 92%, 92%, and 94%, respectively.

Conclusion

The results of this study showed that digital analysis of images has a high accuracy in diagnosing malignant and benign lesions in the cytology of EUS-FNA in patients with suspected pancreatic malignancy and by obtaining cutoff points by software output factors; digital imaging can be used to differentiate between benign and malignant pancreatic tumors.

Quantitative multiplexed imaging technologies for single-cell analysis to assess predictive markers for immunotherapy in thoracic immuno-oncology: promises and challenges

The past decade has witnessed a revolution in cancer treatment by the shift from conventional drugs (chemotherapies) towards targeted molecular therapies and immune-based therapies, in particular the immune-checkpoint inhibitors (ICIs). These immunotherapies selectively release the host immune system against the tumour and have shown unprecedented durable remission for patients with cancers that were thought incurable such as advanced non-small cell lung cancer (aNSCLC). The prediction of therapy response is based since the first anti-PD-1/PD-L1 molecules FDA and EMA approvals on the level of PD-L1 tumour cells expression evaluated by immunohistochemistry, and recently more or less on tumour mutation burden in the USA. However, not all aNSCLC patients benefit from immunotherapy equally, since only around 30% of them received ICIs and among them 30% have an initial response to these treatments. Conversely, a few aNSCLC patients could have an efficacy ICIs response despite low PD-L1 tumour cells expression. In this context, there is an urgent need to look for additional robust predictive markers for ICIs efficacy in thoracic oncology. Understanding of the mechanisms that enable cancer cells to adapt to and eventually overcome therapy and identifying such mechanisms can help circumvent resistance and improve treatment. However, more than a unique universal marker, the evaluation of several molecules in the tumour at the same time, particularly by using multiplex immunostaining is a promising open room to optimise the selection of patients who benefit from ICIs. Therefore, urgent further efforts are needed to optimise to individualise immunotherapy based on both patient-specific and tumour-specific characteristics. This review aims to rethink the role of multiplex immunostaining in immuno-thoracic oncology, with the current advantages and limitations in the near-daily practice use.

Multiplex imaging in immuno-oncology

Multiplex imaging has emerged as an invaluable tool for immune-oncologists and translational researchers, enabling them to examine intricate interactions among immune cells, stroma, matrix, and malignant cells within the tumor microenvironment (TME). It holds significant promise in the quest to discover improved biomarkers for treatment stratification and identify novel therapeutic targets. Nonetheless, several challenges exist in the realms of study design, experiment optimization, and data analysis. In this review, our aim is to present an overview of the utilization of multiplex imaging in immuno-oncology studies and inform novice researchers about the fundamental principles at each stage of the imaging and analysis process.

Deep learning for the prediction of the chemotherapy response of metastatic colorectal cancer: comparing and combining H&E staining histopathology and infrared spectral histopathology

Colorectal cancer is a global public health problem with one of the highest death rates. It is the second most deadly type of cancer and the third most frequently diagnosed in the world. The present study focused on metastatic colorectal cancer (mCRC) patients who had been treated with chemotherapy-based regimen for which it remains uncertainty about the efficacy for all eligible patients. This is a major problem, as it is not yet possible to test different therapies in view of the consequences on the health of the patients and the risk of progression. Here, we propose a method to predict the efficacy of an anticancer treatment in an individualized way, using a deep learning model constructed on the retrospective analysis of the primary tumor of several patients. Histological sections from tumors were imaged by standard hematoxylin and eosin (HE) staining and infrared spectroscopy (IR). Images obtained were then processed by a convolutional neural network (CNN) to extract features and correlate them with the subsequent progression-free survival (PFS) of each patient. Separately, HE and IR imaging resulted in a PFS prediction with an error of 6.6 and 6.3 months respectively (28% and 26% of the average PFS). Combining both modalities allowed to decrease the error to 5.0 months (21%). The inflammatory state of the stroma seemed to be one of the main features detected by the CNN. Our pilot study suggests that multimodal imaging analyzed with deep learning methods allow to give an indication of the effectiveness of a treatment when choosing.

Unleashing the potential of AI for pathology: challenges and recommendations

Computational pathology is currently witnessing a surge in the development of AI techniques, offering promise for achieving breakthroughs and significantly impacting the practices of pathology and oncology. These AI methods bring with them the potential to revolutionize diagnostic pipelines as well as treatment planning and overall patient care. Numerous peer-reviewed studies reporting remarkable performance across diverse tasks serve as a testimony to the potential of AI in the field. However, widespread adoption of these methods in clinical and pre-clinical settings still remains a challenge. In this review article, we present a detailed analysis of the major obstacles encountered during the development of effective models and their deployment in practice. We aim to provide readers with an overview of the latest developments, assist them with insights into identifying some specific challenges that may require resolution, and suggest recommendations and potential future research directions. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

The area of residual tumor predicts esophageal squamous cell carcinoma prognosis following neoadjuvant chemotherapy

PURPOSE

To clarify the utility of the area of residual tumor for patients with esophageal squamous cell cancer treated with neoadjuvant chemotherapy.

METHODS

We enrolled 186 patients with esophageal squamous cell cancer who underwent surgical resection following neoadjuvant chemotherapy at our hospital. Using digital images, we measured the area of residual tumor at the maximum plane of the specimen and divided the patient into three groups as follows: 0 (area = 0 mm2), low (area = 0-40 mm2), and high (area ≥ 40 mm2). The clinicopathological factors and prognosis were compared among these groups.

RESULTS

The median area of the residual tumor was 15.0 mm2　(range 0-1,448.8 mm2). Compared with the 0 and low group, the high group was significantly associated with poorer recurrence-free survival (all P < .001) and overall survival (P < .001 [vs. 0] and P = .017 [vs low]). The area of residual tumor, ypN, tumor regression grade, and lymphovascular invasion were independent predictors of recurrence-free survival. By dividing the patients using a combination of the area of residual tumor and lymphovascular invasion, the high and/or lymphovascular invasion ( +) group displayed significantly poor recurrence-free survival than the 0 group and low/lymphovascular invasion ( -) group. However, there was no significant difference in the recurrence-free survival between the 0 group and low/lymphovascular invasion ( -) group.

CONCLUSION

The area of residual tumor is a promising histopathological prognostic factor for patients with esophageal squamous cell cancer treated with neoadjuvant chemotherapy. Moreover, it is a possible candidate histopathological factor for postoperative chemotherapy selection.

Predicting peritoneal recurrence in gastric cancer with serosal invasion using a pathomics nomogram

Peritoneal recurrence is the most frequent and lethal recurrence pattern in gastric cancer (GC) with serosal invasion after radical surgery. However, current evaluation methods are not adequate for predicting peritoneal recurrence in GC with serosal invasion. Emerging evidence shows that pathomics analyses could be advantageous for risk stratification and outcome prediction. Herein, we propose a pathomics signature composed of multiple pathomics features extracted from digital hematoxylin and eosin-stained images. We found that the pathomics signature was significantly associated with peritoneal recurrence. A competing-risk pathomics nomogram including carbohydrate antigen 19-9 level, depth of invasion, lymph node metastasis, and pathomics signature was developed for predicting peritoneal recurrence. The pathomics nomogram had favorable discrimination and calibration. Thus, the pathomics signature is a predictive indicator of peritoneal recurrence, and the pathomics nomogram may provide a helpful reference for predicting an individual's risk in peritoneal recurrence of GC with serosal invasion.

Design and Validation of a Custom-Made Hyperspectral Microscope Imaging System for Biomedical Applications

Hyperspectral microscope imaging (HMI) is an emerging modality that integrates spatial information collected by standard laboratory microscopy and the spectral-based contrast obtained by hyperspectral imaging and may be instrumental in establishing novel quantitative diagnostic methodologies, particularly in histopathology. Further expansion of HMI capabilities hinges upon the modularity and versatility of systems and their proper standardization. In this report, we describe the design, calibration, characterization, and validation of the custom-made laboratory HMI system based on a Zeiss Axiotron fully motorized microscope and a custom-developed Czerny-Turner-type monochromator. For these important steps, we rely on a previously designed calibration protocol. Validation of the system demonstrates a performance comparable to classic spectrometry laboratory systems. We further demonstrate validation against a laboratory hyperspectral imaging system for macroscopic samples, enabling future comparison of spectral imaging results across length scales. An example of the utility of our custom-made HMI system on a standard hematoxylin and eosin-stained histology slide is also shown.

Annotating for Artificial Intelligence Applications in Digital Pathology: A Practical Guide for Pathologists and Researchers

Training machine learning models for artificial intelligence (AI) applications in pathology often requires extensive annotation by human experts, but there is little guidance on the subject. In this work, we aimed to describe our experience and provide a simple, useful, and practical guide addressing annotation strategies for AI development in computational pathology. Annotation methodology will vary significantly depending on the specific study's objectives, but common difficulties will be present across different settings. We summarize key aspects and issue guiding principles regarding team interaction, ground-truth quality assessment, different annotation types, and available software and hardware options and address common difficulties while annotating. This guide was specifically designed for pathology annotation, intending to help pathologists, other researchers, and AI developers with this process.

Reciprocal effects of mTOR inhibitors on pro-survival proteins dictate therapeutic responses in tuberous sclerosis complex

mTORC1 is aberrantly activated in cancer and in the genetic tumor syndrome tuberous sclerosis complex (TSC), which is caused by loss-of-function mutations in the TSC complex, a negative regulator of mTORC1. Clinically approved mTORC1 inhibitors, such as rapamycin, elicit a cytostatic effect that fails to eliminate tumors and is rapidly reversible. We sought to determine the effects of mTORC1 on the core regulators of intrinsic apoptosis. In TSC2-deficient cells and tumors, we find that mTORC1 inhibitors shift cellular dependence from MCL-1 to BCL-2 and BCL-XL for survival, thereby altering susceptibility to BH3 mimetics that target specific pro-survival BCL-2 proteins. The BCL-2/BCL-XL inhibitor ABT-263 synergizes with rapamycin to induce apoptosis in TSC-deficient cells and in a mouse tumor model of TSC, resulting in a more complete and durable response. These data expose a therapeutic vulnerability in regulation of the apoptotic machinery downstream of mTORC1 that promotes a cytotoxic response to rapamycin.

Evaluation of MetaSystems Automated Fluorescent Microscopy System for the Machine-Assisted Detection of Acid-Fast Bacilli in Clinical Samples

Manual reading of fluorescent acid-fast bacilli (AFB) microscopy slides is time-intensive and technically demanding. The aim of this study was to evaluate the accuracy of MetaSystems' automated fluorescent AFB slide scanner and analyzer. Auramine O-stained slides corresponding to 133 culture-positive and 363 culture-negative respiratory (n = 284), tissue (n = 120), body fluid (n = 81), and other (n = 11) sources were evaluated with the MetaSystems Mycobacteria Scanner running the NEON Metafer AFB Module. The sensitivity and specificity of the MetaSystems platform was measured as a standalone diagnostic and as an assistant to technologists to review positive images. Culture results were used as the reference method. The MetaSystems platform failed to scan 57 (11.5%) slides. The MetaSystems platform used as a standalone had a sensitivity of 97.0% (129/133; 95% CI 92.5 to 99.2) and specificity of 12.7% (46/363; 95% CI 9.4 to 16.5). When positive scans were used to assist technologists, the MetaSystems platform had a sensitivity of 70.7% (94/133; 95% CI 62.2 to 78.3) and specificity of 89.0% (323/363; 95% CI 85.3 to 92.0). The manual microscopy method had a sensitivity of 79.7% (106/133; 95% CI 71.9 to 86.2) and specificity of 98.6% (358/363; 95% CI 96.8 to 99.6). The sensitivity of the MetaSystems platform was not impacted by smear grade or mycobacterial species. The majority (70.3%) of false positive smears had ≥2+ smear results with the MetaSystems platform. Further performance improvements are needed before the MetaSystems' automated fluorescent AFB slide reader can be used to assist microscopist in the clinical laboratory.

Validating Automatic Concept-Based Explanations for AI-Based Digital Histopathology

Digital histopathology poses several challenges such as label noise, class imbalance, limited availability of labelled data, and several latent biases to deep learning, negatively influencing transparency, reproducibility, and classification performance. In particular, biases are well known to cause poor generalization. Proposed tools from explainable artificial intelligence (XAI), bias detection, and bias discovery suffer from technical challenges, complexity, unintuitive usage, inherent biases, or a semantic gap. A promising XAI method, not studied in the context of digital histopathology is automated concept-based explanation (ACE). It automatically extracts visual concepts from image data. Our objective is to evaluate ACE's technical validity following design science principals and to compare it to Guided Gradient-weighted Class Activation Mapping (Grad-CAM), a conventional pixel-wise explanation method. To that extent, we created and studied five convolutional neural networks (CNNs) in four different skin cancer settings. Our results demonstrate that ACE is a valid tool for gaining insights into the decision process of histopathological CNNs that can go beyond explanations from the control method. ACE validly visualized a class sampling ratio bias, measurement bias, sampling bias, and class-correlated bias. Furthermore, the complementary use with Guided Grad-CAM offers several benefits. Finally, we propose practical solutions for several technical challenges. In contradiction to results from the literature, we noticed lower intuitiveness in some dermatopathology scenarios as compared to concept-based explanations on real-world images.