A systematic pan-cancer study on deep learning-based prediction of multi-omic biomarkers from routine pathology images.

BACKGROUND: The objective of this comprehensive pan-cancer study is to evaluate the potential of deep learning (DL) for molecular profiling of multi-omic biomarkers directly from hematoxylin and eosin (H&E)-stained whole slide images. METHODS: A total of 12,093 DL models predicting 4031 multi-omic biomarkers across 32 cancer types were trained and validated. The study included a broad range of genetic, transcriptomic, and proteomic biomarkers, as well as established prognostic markers, molecular subtypes, and clinical outcomes. RESULTS: Here we show that 50% of the models achieve an area under the curve (AUC) of 0.644 or higher. The observed AUC for 25% of the models is at least 0.719 and exceeds 0.834 for the top 5%. Molecular profiling with image-based histomorphological features is generally considered feasible for most of the investigated biomarkers and across different cancer types. The performance appears to be independent of tumor purity, sample size, and class ratio (prevalence), suggesting a degree of inherent predictability in histomorphology. CONCLUSIONS: The results demonstrate that DL holds promise to predict a wide range of biomarkers across the omics spectrum using only H&E-stained histological slides of solid tumors. This paves the way for accelerating diagnosis and developing more precise treatments for cancer patients.

Molecular profiling tests are used to check cancers for changes in certain genes, proteins, or other molecules. Results of such tests can be used to identify the most effective treatment for cancer patients. Faster and more accessible alternatives to standard tests are needed to improve cancer care. This study investigates whether deep learning (DL), a series of advanced computer techniques, can perform molecular profiling directly from routinely-collected images of tumor specimens used for diagnostic purposes. Over 12,000 DL models were utilized to evaluate thousands of biomarkers using statistical approaches. The results indicate that DL can effectively detect molecular changes in a tumor from these images, for many biomarkers and tumor types. The study shows that DL-based molecular profiling from images is possible. Introducing this type of approach into routine clinical workflows could potentially accelerate treatment decisions and improve outcomes.

The Role of Neutrophil-Lymphocyte-Ratio (NLR) and Platelet-Lymphocyte-Ratio (PLR) as a Biomarker for Distinguishing Between Complicated and Uncomplicated Appendicitis.

Introduction Acute appendicitis (AA) is one of the most common acute general surgical presentations affecting 7% of the population at some point in their lifetime. The ability to assess the risk of complicated appendicitis (CA) from uncomplicated appendicitis (UA) in acute appendicitis (AA) could reduce the associated morbidity and mortality. The value of platelet lymphocyte ratio (PLR) as an inflammatory marker increases when its fluctuations are interpreted along with other complementary hematologic indices, such as neutrophil-to-lymphocyte ratio (NLR), which provides additional information about the disease activity. Hence, we postulated that NLR and/or PLR could serve as a potential surrogate marker in assessing the severity of AA. Aim This study aims to investigate the use of PLR and/or NLR as a surrogate biomarker in differentiating uncomplicated from complicated appendicitis. Material and methods This retrospective study was conducted at Russells Hall Hospital from January 1, 2017, to December 31, 2020. Data of all patients over age 16 years that had histologically confirmed appendicitis were retrieved. NLR and PLR were calculated from the admission hemogram, and the ratios were compared between uncomplicated (UA) or complicated appendicitis (CA). Cut-off values were calculated using the summarized ROC curve; in addition, the sensitivity and specificity with 95% confidence intervals were determined using SPSS 25.0 (IBM Corp., Armonk, NY). Results A total of 799 patients were analyzed, of which 469 (58.7%) were female. The median age was 31.2 years. The difference between NLR and PLR within the two appendicitis groups was significant (P=0.05; Kruskal-Wallis). Cohen's kappa (degree of inter-rater agreement) between NLR and PLR showed a moderate agreement of 0.589 (P<0.001). We equally demonstrated an exponential relationship between PLR and NLR (R2 =0.510, P<0.05). For UA, the area under the curve (AUC) and the cut-off for NLR and PLR were 0.715, 4.75 with a confidence interval (CI) of 0.678-0.653 and 0.632, 155 with a CI of 0.591-0.672, respectively. For CA, using NLR and PLR, the AUC and cut-off were 0.727, 6.96 with a CI of 0.687-0.768 and 0.653, 180.5 with a CI of 0.602-0.703, respectively; all were significant with a P of <0.001. Conclusion NLR and PLR are a reliable, less cumbersome surrogate biomarker for assessing the severity of acute appendicitis.

Weakly Supervised Learning for Categorization of Medical Inquiries for Customer Service Effectiveness.

With the growing unstructured data in healthcare and pharmaceutical, there has been a drastic adoption of natural language processing for generating actionable insights from text data sources. One of the key areas of our exploration is the Medical Information function within our organization. We receive a significant amount of medical information inquires in the form of unstructured text. An enterprise-level solution must deal with medical information interactions via multiple communication channels which are always nuanced with a variety of keywords and emotions that are unique to the pharmaceutical industry. There is a strong need for an effective solution to leverage the contextual knowledge of the medical information business along with digital tenants of natural language processing (NLP) and machine learning to build an automated and scalable process that generates real-time insights on conversation categories. The traditional supervised learning methods rely on a huge set of manually labeled training data and this dataset is difficult to attain due to high labeling costs. Thus, the solution is incomplete without its ability to self-learn and improve. This necessitates techniques to automatically build relevant training data using a weakly supervised approach from textual inquiries across consumers, healthcare professionals, sales, and service providers. The solution has two fundamental layers of NLP and machine learning. The first layer leverages heuristics and knowledgebase to identify the potential categories and build an annotated training data. The second layer, based on machine learning and deep learning, utilizes the training data generated using the heuristic approach for identifying categories and sub-categories associated with verbatim. Here, we present a novel approach harnessing the power of weakly supervised learning combined with multi-class classification for improved categorization of medical information inquiries.

Meningomyelocele: the tip of the iceberg.

Meningomyelocele is one of the commonest neural tube closure defects. A 2-month-old girl presented with meningomyelocele in the lumbosacral region that had been present since birth. She had experienced paraparesis for 1 month. Terminally, she had excessive crying, and died within 1 week of presentation. An autopsy revealed focally infected meningomyelocele with features of Arnold-Chiari II malformation. Associated malformations included polymicrogyria, neuronal heterotopias and neuronal cytomegaly in the brainstem. Cortical malformations may be detected radiologically, thus prompting surgical intervention and reduced mortality. By contrast, the presence of cytomegaly may hinder postsurgical improvement, affecting intellectual outcome in survivors in particular. New genetic revelations also offer a scope for genetic counselling in these conditions.