Search

The color variation of hematoxylin and eosin (HE)-stained tissues has presented a challenge for applications of artificial intelligence (AI) in digital pathology. Many color normalization algorithms have been developed in recent years in order to reduce the color variation between HE images. However, previous efforts in benchmarking these algorithms have produced conflicting results and none have sufficiently assessed the efficacy of the various color normalization methods for improving diagnostic performance of AI systems. In this study, we systematically investigated eight color normalization algorithms for AI-based classification of HE-stained histopathology slides, in the context of using images both from one center and from multiple centers. Our results show that color normalization does not consistently improve classification performance when both training and testing data are from a single center. However, using four multi-center datasets of two cancer types (ovarian and pleural) and objective functions, we show that color normalization can significantly improve the classification accuracy of images from external datasets (ovarian cancer: 0.25 AUC increase, p = 1.6 e-05; pleural cancer: 0.21 AUC increase, p = 1.4 e-10). Furthermore, we introduce a novel augmentation strategy by mixing color-normalized images using three easily accessible algorithms that consistently improves the diagnosis of test images from external centers, even when the individual normalization methods had varied results. We anticipate our study to be a starting point for reliable use of color normalization to improve AI-based, digital pathology-empowered diagnosis of cancers sourced from multiple centers. © 2021 The Pathological Society of Great Britain and Ireland. Published by John WileySons, Ltd.

Sarcomatoid mesothelioma is an aggressive malignancy that can be challenging to distinguish from benign spindle cell mesothelial proliferations based on biopsy, and this distinction is crucial to patient treatment and prognosis. A novel deep learning based classifier may be able to aid pathologists in making this critical diagnostic distinction. SpindleMesoNET was trained on cases of malignant sarcomatoid mesothelioma and benign spindle cell mesothelial proliferations. Performance was assessed through cross-validation on the training set, on an independent set of challenging cases referred for expert opinion (‘referral’ test set), and on an externally stained set from outside institutions (‘externally stained’ test set). SpindleMesoNET predicted the benign or malignant status of cases with AUC’s of 0.932, 0.925, and 0.989 on the cross-validation, referral and external test sets, respectively. The accuracy of SpindleMesoNET on the referral set cases (92.5%) was comparable to the average accuracy of 3 experienced pathologists on the same slide set (91.7%). We conclude that SpindleMesoNET can accurately distinguish sarcomatoid mesothelioma from benign spindle cell mesothelial proliferations. A deep learning system of this type holds potential for future use as an ancillary test in diagnostic pathology.

This is the first report of a NACC2‐NTRK2 fusion in a histological glioblastoma. Oncogenomic analysis revealed this actionable fusion oncogene in a pediatric cerebellar glioblastoma, which would not have been identified through routine diagnostics, demonstrating the value of clinical genome profiling in cancer care.

Ependymoma and chordoma are 2 tumors that occur throughout the craniospinal axis, and for which the extent of neurosurgical resection has a key prognostic role. Both tumors have distinctive pathologic features, yet can present significant diagnostic challenges to pathologists in cases without classical histology. The molecular understanding of ependymoma has had significant advances in the past decade, with the identification of 9 molecular groups with significant prognostic and clinical implications, while a comprehensive study of chordoma further emphasized the key role of brachyury overexpression in its pathogenesis. In this review, we discuss the pathogenesis, radiology and gross pathology, histology, and molecular features of these 2 tumors, as well as active research into targeted therapies, with an emphasis on practical diagnostic challenges, and the use of immunohistochemical and molecular tests in routine diagnostic practice.

Deep learning-based computer vision methods have recently made remarkable breakthroughs in the analysis and classification of cancer pathology images. However, there has been relatively little investigation of the utility of deep neural networks to synthesize medical images. In this study, we evaluated the efficacy of generative adversarial networks to synthesize high-resolution pathology images of 10 histological types of cancer, including five cancer types from The Cancer Genome Atlas and the five major histological subtypes of ovarian carcinoma. The quality of these images was assessed using a comprehensive survey of board-certified pathologists (n = 9) and pathology trainees (n = 6). Our results show that the real and synthetic images are classified by histotype with comparable accuracies and the synthetic images are visually indistinguishable from real images. Furthermore, we trained deep convolutional neural networks to diagnose the different cancer types and determined that the synthetic images perform as well as additional real images when used to supplement a small training set. These findings have important applications in proficiency testing of medical practitioners and quality assurance in clinical laboratories. Furthermore, training of computer-aided diagnostic systems can benefit from synthetic images where labeled datasets are limited (e.g. rare cancers). We have created a publicly available website where clinicians and researchers can attempt questions from the image survey (http://gan.aimlab.ca/). © 2020 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Embryonal tumor with multilayered rosettes (ETMR) is a rare and highly aggressive embryonal central nervous system tumor that primarily affects young children. It is characterized by (1) amplification of the C19MC miRNA cluster at 19q13.42 and (2) immunohistochemical tumor cell positivity for LIN28A. We describe the case of a 3-year-old girl who presented with a 2-week history of multiple neurological deficits. Based primarily on imaging findings that revealed a large pontine tumor, biopsy was not performed and the patient was clinically diagnosed with a “diffuse intrinsic pontine glioma.” She was subsequently treated with radiation and concurrent adjuvant temozolomide, but unfortunately there was minimal response and the patient died 6 months after diagnosis. Autopsy revealed an ETMR that was confirmed via C19MC fluorescence in situ hybridization and LIN28 immunohistochemistry. Although widespread central nervous system dissemination was observed, large portions of the main pontine mass exhibited evidence of extensive glial and neuronal maturation (ie, differentiation). We consider this tissue “maturation” to have been induced by chemotherapy and radiation. Herein, we discuss the importance of antemortem biopsy of intrinsic pontine tumors and the clinical significance of glial and neuronal maturation post therapy in the context of ETMR.

Deep learning refers to a set of computer models that have recently been used to make unprecedented progress in the way computers extract information from images. These algorithms have been applied to tasks in numerous medical specialties, most extensively radiology and pathology, and in some cases have attained performance comparable to human experts. Furthermore, it is possible that deep learning could be used to extract data from medical images that would not be apparent by human analysis and could be used to inform on molecular status, prognosis, or treatment sensitivity. In this review, we outline the current developments and state-of-the-art in applying deep learning for cancer diagnosis, and discuss the challenges in adapting the technology for widespread clinical deployment.

Sarcomatoid mesothelioma is an aggressive malignancy that can be challenging to distinguish from benign spindle cell mesothelial proliferations based on biopsy, and this distinction is crucial to patient treatment and prognosis. A novel deep learning based classifier may be able to aid pathologists in making this critical diagnostic distinction. SpindleMesoNET was trained on cases of malignant sarcomatoid mesothelioma and benign spindle cell mesothelial proliferations. Performance was assessed through cross-validation on the training set, on an independent set of challenging cases referred for expert opinion ('referral' test set), and on an externally stained set from outside institutions ('externally stained' test set). SpindleMesoNET predicted the benign or malignant status of cases with AUC's of 0.932, 0.925, and 0.989 on the cross-validation, referral and external test sets, respectively. The accuracy of SpindleMesoNET on the referral set cases (92.5%) was comparable to the average accuracy of 3 experienced pathologists on the same slide set (91.7%). We conclude that SpindleMesoNET can accurately distinguish sarcomatoid mesothelioma from benign spindle cell mesothelial proliferations. A deep learning system of this type holds potential for future use as an ancillary test in diagnostic pathology.

BackgroundAberrations inCapicua(CIC) have recently been implicated as a negative prognostic factor in a multitude of cancer types through the derepression of targets downstream of the mitogen-activated protein kinase (MAPK) signaling cascade, such as oncogenic E26 transformation-specific (ETS) transcription factors. The Ataxin-family protein ATXN1L has previously been reported to interact with CIC in both developmental and disease contexts to facilitate the repression of CIC target genes and promote the post-translational stability of CIC. However, little is known about the mechanisms at the base of ATXN1L-mediated CIC post-translational stability.ResultsFunctional in vitro studies utilizingATXN1LKOhuman cell lines revealed that loss of ATXN1L leads to the accumulation of polyubiquitinated CIC protein, promoting its degradation through the proteasome. Although transcriptomic signatures ofATXN1LKOcell lines indicated upregulation of the mitogen-activated protein kinase pathway, ERK activity was found to contribute to CIC function but not stability. Degradation of CIC protein following loss of ATXN1L was instead observed to be mediated by the E3 ubiquitin ligase TRIM25 which was further validated using glioma-derived cell lines and the TCGA breast carcinoma and liver hepatocellular carcinoma cohorts.ConclusionsThe post-translational regulation of CIC through ATXN1L and TRIM25 independent of ERK activity suggests that the regulation of CIC stability and function is more intricate than previously appreciated and involves several independent pathways. As CIC status has become a prognostic factor in several cancer types, further knowledge into the mechanisms which govern CIC stability and function may prove useful for future therapeutic approaches.

Deep learning-based computer vision methods have recently made remarkable breakthroughs in the analysis and classification of cancer pathology images. However, there has been relatively little investigation of the utility of deep neural networks to synthesize medical images. In this study, we evaluated the efficacy of generative adversarial networks (GANs) to synthesize high resolution pathology images of ten histological types of cancer, including five cancer types from The Cancer Genome Atlas (TCGA) and the five major histological subtypes of ovarian carcinoma. The quality of these images was assessed using a comprehensive survey of board-certified pathologists (n = 9) and pathology trainees (n = 6). Our results show that the real and synthetic images are classified by histotype with comparable accuracies, and the synthetic images are visually indistinguishable from real images. Furthermore, we trained deep convolutional neural networks (CNNs) to diagnose the different cancer types and determined that the synthetic images perform as well as additional real images when used to supplement a small training set. These findings have important applications in proficiency testing of medical practitioners and quality assurance in clinical laboratories. Furthermore, training of computer-aided diagnostic systems can benefit from synthetic images where labeled datasets are limited (e.g., rare cancers). We have created a publicly available website where clinicians and researchers can attempt questions from the image survey at http://gan.aimlab.ca/.

Background Primary central nervous system lymphoma (PCNSL) is a rare malignancy that usually arises in the context of severe immunosuppression but has incompletely understood etiology, limiting effective treatments. Case Description We present the case of an 81-year-old immunocompetent man who developed a PCNSL in the right temporal lobe, at the site of a remote episode of herpes simplex virus (HSV) encephalitis 8 years prior. There are numerous viruses with known oncogenic associations; however, to our knowledge, this is the first reported case of PCNSL with an antecedent HSV infection. Conclusions We discuss this case in the context of our current understandings of the pathogenesis of HSV encephalitis and PCNSL and postulate mechanisms through which the 2 could be associated.

Deep learning, a subset of artificial intelligence, has shown great potential in several recent applications to pathology. These have mainly involved the use of classifiers to diagnose disease, while generative modelling techniques have been less frequently used. Generative adversarial networks (GANs) are a type of deep learning model that has been used to synthesize realistic images in a range of domains, both general purpose and medical. In the GAN framework, a generator network is trained to synthesize fake images, while a dueling discriminator network aims to distinguish between the fake images and a set of real training images. As GAN training progresses, the generator network ideally learns the important features of a dataset, allowing it to create images that the discriminator cannot distinguish from the real ones. We report on our use of GANs to synthesize high resolution, realistic histopathology images of gliomas. The well- known Progressive GAN framework was trained on a set of image patches extracted from digital slides in the Cancer Genome Atlas repository, and was able to generate fake images that were visually indistinguishable from the real training images. Generative modelling in pathology has numerous potential applications, including dataset augmentation for training deep learning classifiers, image processing, and expanding educational material.LEARNING OBJECTIVESThis presentation will enable the learner to: 1.Explain basic principles of generative modelling in deep learning.2.Discuss applications of deep learning to neuropathology image synthesis.

Background: Chronic neuropathic groin pain is a common problem. It can arise following surgery or trauma, or spontaneously as part of various pelvic pain syndromes. A number of different stimulation techniques have been reported in the literature to treat this area, but due to the complex anatomy of the region, it can be difficult to target effectively with paresthesias. Objectives: In this study we report our results treating patients with chronic neuropathic groin, pelvic, and abdominal pain, using spinal cord stimulation and dorsal nerve root stimulation. Study Design: Open label, prospective study that includes all patients treated with a new trial stimulator system at a single center between July 1, 2011, and October 31, 2013. Setting: Academic university neurosurgical pain center, Canada. Methods: Thirty-two patients had trials of spinal cord stimulation and/or dorsal nerve root stimulation in the thoracic or lumbar spine. Patients were evaluated on visual analog scale pain scores, SF-36, and morphine equivalent daily dose. Data were recorded at the pre-implant visit, and 3, 6, and 12 months following permanent implant. Results: The 15 patients who went on to permanent implants had, on average, significant pain reduction and improvements in quality of life at the 12 month follow-up. The majority of patients who were taking opioids at the initial assessment were able to reduce their dose with treatment. Three patients with successful trials were long-term non-responders, of whom 2 had the permanent device removed. Limitations: This study would benefit from a larger sample size that would have adequate power for comparisons between patient subgroups and stimulation techniques. Conclusion: Dorsal nerve root stimulation is an effective long-term treatment for neuropathic groin pain. Key words: Spinal cord stimulation, nerve root stimulation, lumbar, thoracic, neuropathic pain, groin pain, pelvic pain, abdominal pain, neuromodulation, clinical effectiveness

Effective management of brain and spine tumors relies on a multidisciplinary approach encompassing surgery, radiation, and systemic therapy. In the era of personalized oncology, the latter is complemented by various molecularly targeting agents. Precise identification of cellular targets for these drugs requires comprehensive profiling of the cancer genome coupled with an efficient analytic pipeline, leading to an informed decision on drug selection, prognosis, and confirmation of the original pathological diagnosis. Acquisition of optimal tumor tissue for such analysis is paramount and often presents logistical challenges in neurosurgery. Here, we describe the experience and results of the Personalized OncoGenomics (POG) program with a focus on tumors of the central nervous system (CNS). Patients with recurrent CNS tumors were consented and enrolled into the POG program prior to accrual of tumor and matched blood followed by whole-genome and transcriptome sequencing and processing through the POG bioinformatic pipeline. Sixteen patients were enrolled into POG. In each case, POG analyses identified genomic drivers including novel oncogenic fusions, aberrant pathways, and putative therapeutic targets. POG has highlighted that personalized oncology is truly a multidisciplinary field, one in which neurosurgeons must play a vital role if these programs are to succeed and benefit our patients.

Chronic neuropathic groin pain is a common problem. It can arise following surgery or trauma, or spontaneously as part of various pelvic pain syndromes. A number of different stimulation techniques have been reported in the literature to treat this area, but due to the complex anatomy of the region, it can be difficult to target effectively with paresthesias.In this study we report our results treating patients with chronic neuropathic groin, pelvic, and abdominal pain, using spinal cord stimulation and dorsal nerve root stimulation.Open label, prospective study that includes all patients treated with a new trial stimulator system at a single center between July 1, 2011, and October 31, 2013.Academic university neurosurgical pain center, Canada.Thirty-two patients had trials of spinal cord stimulation and/or dorsal nerve root stimulation in the thoracic or lumbar spine. Patients were evaluated on visual analog scale pain scores, SF-36, and morphine equivalent daily dose. Data were recorded at the pre-implant visit, and 3, 6, and 12 months following permanent implant.The 15 patients who went on to permanent implants had, on average, significant pain reduction and improvements in quality of life at the 12 month follow-up. The majority of patients who were taking opioids at the initial assessment were able to reduce their dose with treatment. Three patients with successful trials were long-term non-responders, of whom 2 had the permanent device removed.This study would benefit from a larger sample size that would have adequate power for comparisons between patient subgroups and stimulation techniques.Dorsal nerve root stimulation is an effective long-term treatment for neuropathic groin pain.

Background Deep brain stimulation (DBS) of the ventrocaudalis nucleus of the thalamus is a last resort treatment for chronic refractory pain. DBS is generally a safe procedure, although it can result in functional disturbances depending on the site of stimulation. There has been 1 previous report of stuttering induced by microlesioning of the thalamus, as well as several reports of stuttering induced by stimulation of the thalamus and other related structures in the brain. Case Description We describe the case of a patient with trigeminal deafferentation face pain who was treated with DBS of the ventrocaudalis nucleus thalamus and developed a reversible stutter immediately on insertion of the electrode. The stutter improved significantly over 12 days after implant; however, the device was not effective in relieving the patient's pain and was removed. Conclusions Stuttering is a rare complication of deep brain exploration of the sensory thalamus. Our coordinates are near to but in a distinct anatomic region compared with cases previously described as having similar effects on speech.

A 69-year-old male presented to a peripheral emergency department with a several day history of increasing confusion and headache. On admission, his Glasgow Coma Scale (GCS) score was 10 (E3 M6 V1). He was mute but would obey commands intermittently. Cranial nerve exam was normal and there was no evidence of weakness. A computed tomogram (CT) head (Figure 1) showed a chronic subdural hematoma with midline shift, as well as a mass within the left frontal region that appeared consistent with a convexity meningioma seen on magnetic resonance imaging (MRI) two years prior (Figure 2). The patient was not on any blood thinners and had no history of falls. Past medical history also included hypertension and prostate cancer.