Contribution of whole slide imaging-based deep learning in the assessment of intraoperative and postoperative sections in neuropathology.

The pathological diagnosis of intracranial germinoma (IG), oligodendroglioma, and low-grade astrocytoma on intraoperative frozen section (IFS) and hematoxylin-eosin (HE)-staining section directly determines patients' treatment options, but it is a difficult task for pathologists. We aimed to investigate whether whole-slide imaging (WSI)-based deep learning can contribute new precision to the diagnosis of IG, oligodendroglioma, and low-grade astrocytoma. Two types of WSIs (500 IFSs and 832 HE-staining sections) were collected from 379 patients at multiple medical centers. Patients at Center 1 were split into the training, testing, and internal validation sets (3:1:1), while the other centers were the external validation sets. First, we subdivided WSIs into small tiles and selected tissue tiles using a tissue tile selection model. Then a tile-level classification model was established, and the majority voting method was used to determine the final diagnoses. Color jitter was applied to the tiles so that the deep learning (DL) models could adapt to the variations in the staining. Last, we investigated the effectiveness of model assistance. The internal validation accuracies of the IFS and HE models were 93.9% and 95.3%, respectively. The external validation accuracies of the IFS and HE models were 82.0% and 76.9%, respectively. Furthermore, the IFS and HE models can predict Ki-67 positive cell areas with R2 of 0.81 and 0.86, respectively. With model assistance, the IFS and HE diagnosis accuracy of pathologists improved from 54.6%-69.7% and 53.5%-83.7% to 87.9%-93.9% and 86.0%-90.7%, respectively. Both the IFS model and the HE model can differentiate the three tumors, predict the expression of Ki-67, and improve the diagnostic accuracy of pathologists. The use of our model can assist clinicians in providing patients with optimal and timely treatment options.

A CNN-LASSO ensemble classification model for incomplete antibody reactants screening in coombs test.

BACKGROUND: Precise classification of incomplete antibody reactants (IAR) in the Coombs test is the primary means to prevent incompatible blood transfusions. Currently, an automatic and contactless method is required for accurate IAR classification to avoid human error. OBJECTIVE: We present an ensemble learning algorithm that integrates five convolutional neural networks and the least absolute shrinkage and selection operator (LASSO) regression algorithm into an IAR intensity classification model. METHODS: A dataset including 1628 IAR and corresponding labels of IAR intensity categories ((-), (1+), (2+), (3+), and (4+)) was used. We trained the ensemble model using 1302 IAR and validated its performance using 326 IAR. The optimal ensemble model was used to assist immunologists in classifying IAR. The chord diagrams based on the human-machine interaction were established. RESULTS: The ensemble model achieved 98.8%, 98.4%, 99.7%, 99.5%, and 99.4% accuracies in the (-), (1+), (2+), (3+), and (4+) categories, respectively. The results were compared with those of manual classification by immunologists (average accuracy: 99.2% vs. 75.6%). Using the model, all three immunologists achieved increased accuracy (average accuracy: +8.4%). CONCLUSIONS: The proposed algorithm can thus effectively improve the accuracy and efficiency of IAR intensity classification and facilitate the automation of haemolytic disease screening equipment.

Convolutional neural network-based automatic classification for incomplete antibody reaction intensity in solid phase anti-human globulin test image.

The precise classification of incomplete antibody reaction intensity (IARI) in hydrogel chromatography medium high density medium solid-phase Coombs test is essential for haemolytic disease screening. However, an automatic and contactless method is required for accurate classification of IARI. Here, we present a deep ensemble learning model that integrates five different convolutional neural networks into a single model for IARI classification. A dataset, including 1628 IARI images and corresponding labels of IARI categories ((-), (1 +), (2 +), (3 +), and (4 +)), was used. We trained our model using 1302 IARIs and validated its performance using 326 IARIs. The proposed model achieved 100%, 99.4%, 99.4%, 100%, and 100% accuracies in the ( -), (1 +), (2 +), (3 +), and (4 +) categories, respectively. The results were compared with those of manual classification by immunologists (average accuracy: 99.8% vs. 88.3%, p < 0.01). Following model assistance, all three immunologists achieved increased accuracy (average accuracy: + 6.1%), with the average accuracy of junior immunologists maximum increasing by 11.3%. The time required for model classification was 0.094 s·image-1, whereas that required manually was 5.528 s·image-1. The proposed model can thus substantially improve the accuracy and efficiency of IARI classification and facilitate the automation of haemolytic disease screening equipment.

Surface values, volumetric measurements and radiomics of structural MRI for the diagnosis and subtyping of attention-deficit/hyperactivity disorder.

Attention-deficit/hyperactivity disorder (ADHD) is diagnosed subjectively based on an individual's behaviour and performance. The clinical community has no objective biomarker to inform the diagnosis and subtyping of ADHD. This study aimed to explore the potential diagnostic biomarkers of ADHD among surface values, volumetric metrics and radiomic features that were extracted from structural MRI images. Public data of New York University and Peking University were downloaded from the ADHD-200 Consortium. MRI T1-weighted images were pre-processed using CAT12. We calculated surface values based on the Desikan-Killiany atlas. The volumetric metrics (mean grey matter volume and mean white matter volume) and radiomic features within each automated anatomical labelling (AAL) brain area were calculated using DPABI and IBEX, respectively. The differences among three groups of participants were tested using ANOVA or Kruskal-Wallis test depending on the normality of the data. We selected discriminative features and classified typically developing controls (TDCs) and ADHD patients as well as two ADHD subtypes using least absolute shrinkage and selection operator and support vector machine algorithms. Our results showed that the radiomics-based model outperformed the others in discriminating ADHD from TDC and classifying ADHD subtypes (area under the curve [AUC]: 0.78 and 0.94 in training test; 0.79 and 0.85 in testing set). Combining grey matter volumes, surface values and clinical factors with radiomic features can improve the performance for classifying ADHD patients and TDCs with training and testing AUCs of 0.82 and 0.83, respectively. This study demonstrates that MRI T1-weighted features, especially radiomic features, are potential diagnostic biomarkers of ADHD.

Metabolomic profiling identifies novel biomarkers and mechanisms in human bladder cancer treated with submucosal injection of gemcitabine.

Bladder cancer (BCa) is a common urinary tract malignancy with frequent recurrences after initial resection. Submucosal injection of gemcitabine prior to transurethral resection of bladder tumor (TURBT) may prevent recurrence of urothelial cancer. However, the underlying mechanism remains unknown. In the present study, ultra­performance liquid chromatography Q­Exactive mass spectrometry was used to profile tissue metabolites from 12 BCa patients. The 48 samples included pre­ and post­gemcitabine treatment BCa tissues, as well as adjacent normal tissues. Principal component analysis (PCA) revealed that the metabolic profiles of pre­gemcitabine BCa tissues differed significantly from those of pre­gemcitabine normal tissues. A total of 34 significantly altered metabolites were further analyzed. Pathway analysis using MetaboAnalyst identified three metabolic pathways closely associated with BCa, including glutathione, purine and thiamine metabolism, while glutathione metabolism was also identified by the enrichment analysis using MetaboAnalyst. In search of the possible targets of gemcitabine, metabolite profiles were compared between the pre­gemcitabine normal and post­gemcitabine BCa tissues. Among the 34 metabolites associated with BCa, the levels of bilirubin and retinal recovered in BCa tissues treated with gemcitabine. When comparing normal bladder tissues with and without gemcitabine treatment, among the 34 metabolites associated with BCa, it was observed that histamine change may be associated with the prevention of relapse, whereas thiamine change may be involved in possible side effects. Therefore, by employing a hypothesis­free tissue­based metabolomics study, the present study investigated the metabolic signatures of BCa and found that bilirubin and retinal may be involved in the mechanism underlying the biomolecular action of submucosal injection of gemcitabine in urothelial BCa.