Search

Breast cancer has reached the highest incidence rate worldwide among all malignancies since 2020. Breast imaging plays a significant role in early diagnosis and intervention to improve the outcome of breast cancer patients. In the past decade, deep learning has shown remarkable progress in breast cancer imaging analysis, holding great promise in interpreting the rich information and complex context of breast imaging modalities. Considering the rapid improvement in deep learning technology and the increasing severity of breast cancer, it is critical to summarize past progress and identify future challenges to be addressed. This paper provides an extensive review of deep learning-based breast cancer imaging research, covering studies on mammograms, ultrasound, magnetic resonance imaging, and digital pathology images over the past decade. The major deep learning methods and applications on imaging-based screening, diagnosis, treatment response prediction, and prognosis are elaborated and discussed. Drawn from the findings of this survey, we present a comprehensive discussion of the challenges and potential avenues for future research in deep learning-based breast cancer imaging.

Gastric cancer (GC) is a prevalent malignancy worldwide, ranking as the fifth most common cancer with over 1 million new cases and 700 thousand deaths in 2020. Locally advanced gastric cancer (LAGC) accounts for approximately two-thirds of GC diagnoses, and neoadjuvant chemotherapy (NACT) has emerged as the standard treatment for LAGC. However, the effectiveness of NACT varies significantly among patients, with a considerable subset displaying treatment resistance. Ineffective NACT not only leads to adverse effects but also misses the optimal therapeutic window, resulting in lower survival rate. However, existing multimodal learning methods assume the availability of all modalities for each patient, which does not align with the reality of clinical practice. The limited availability of modalities for each patient would cause information loss, adversely affecting predictive accuracy. In this study, we propose an incomplete multimodal data integration framework for GC (iMD4GC) to address the challenges posed by incomplete multimodal data, enabling precise response prediction and survival analysis. Specifically, iMD4GC incorporates unimodal attention layers for each modality to capture intra-modal information. Subsequently, the cross-modal interaction layers explore potential inter-modal interactions and capture complementary information across modalities, thereby enabling information compensation for missing modalities. To evaluate iMD4GC, we collected three multimodal datasets for GC study: GastricRes (698 cases) for response prediction, GastricSur (801 cases) for survival analysis, and TCGA-STAD (400 cases) for survival analysis. The scale of our datasets is significantly larger than previous studies. The iMD4GC achieved impressive performance with an 80.2% AUC on GastricRes, 71.4% C-index on GastricSur, and 66.1% C-index on TCGA-STAD, significantly surpassing other compared methods., Comment: 27 pages, 9 figures, 3 tables (under review)

Breast cancer has reached the highest incidence rate worldwide among all malignancies since 2020. Breast imaging plays a significant role in early diagnosis and intervention to improve the outcome of breast cancer patients. In the past decade, deep learning has shown remarkable progress in breast cancer imaging analysis, holding great promise in interpreting the rich information and complex context of breast imaging modalities. Considering the rapid improvement in deep learning technology and the increasing severity of breast cancer, it is critical to summarize past progress and identify future challenges to be addressed. This paper provides an extensive review of deep learning-based breast cancer imaging research, covering studies on mammogram, ultrasound, magnetic resonance imaging, and digital pathology images over the past decade. The major deep learning methods and applications on imaging-based screening, diagnosis, treatment response prediction, and prognosis are elaborated and discussed. Drawn from the findings of this survey, we present a comprehensive discussion of the challenges and potential avenues for future research in deep learning-based breast cancer imaging., Comment: IEEE RBME 2024

Computational cytology is a critical, rapid-developing, yet challenging topic in medical image computing concerned with analyzing digitized cytology images by computer-aided technologies for cancer screening. Recently, an increasing number of deep learning (DL) approaches have made significant achievements in medical image analysis, leading to boosting publications of cytological studies. In this article, we survey more than 120 publications of DL-based cytology image analysis to investigate the advanced methods and comprehensive applications. We first introduce various deep learning schemes, including fully supervised, weakly supervised, unsupervised, and transfer learning. Then, we systematically summarize public datasets, evaluation metrics, versatile cytology image analysis applications including cell classification, slide-level cancer screening, nuclei or cell detection and segmentation. Finally, we discuss current challenges and potential research directions of computational cytology. © 2022 Elsevier B.V.

Rapid and high-resolution histological imaging with minimal tissue preparation has long been a challenging and yet captivating medical pursuit. Here, the authors propose a promising and transformative histological imaging method, termed computational high-throughput autofluorescence microscopy by pattern illumination (CHAMP). With the assistance of computational microscopy, CHAMP enables high-throughput and label-free imaging of thick and unprocessed tissues with large surface irregularity at an acquisition speed of 10 mm2/10 s with 1.1-µm lateral resolution. Moreover, the CHAMP image can be transformed into a virtually stained histological image (Deep-CHAMP) through unsupervised learning within 15 s, where significant cellular features are quantitatively extracted with high accuracy. The versatility of CHAMP is experimentally demonstrated using mouse brain/kidney and human lung tissues prepared with various clinical protocols, which enables a rapid and accurate intraoperative/postoperative pathological examination without tissue processing or staining, demonstrating its great potential as an assistive imaging platform for surgeons and pathologists to provide optimal adjuvant treatment.

Rapid and high-resolution histological imaging with minimal tissue preparation has long been a challenging and yet captivating medical pursuit. Here, the authors propose a promising and transformative histological imaging method, termed computational high-throughput autofluorescence microscopy by pattern illumination (CHAMP). With the assistance of computational microscopy, CHAMP enables high-throughput and label-free imaging of thick and unprocessed tissues with large surface irregularity at an acquisition speed of 10 mm2/10 s with 1.1-µm lateral resolution. Moreover, the CHAMP image can be transformed into a virtually stained histological image (Deep-CHAMP) through unsupervised learning within 15 s, where significant cellular features are quantitatively extracted with high accuracy. The versatility of CHAMP is experimentally demonstrated using mouse brain/kidney and human lung tissues prepared with various clinical protocols, which enables a rapid and accurate intraoperative/postoperative pathological examination without tissue processing or staining, demonstrating its great potential as an assistive imaging platform for surgeons and pathologists to provide optimal adjuvant treatment.

Computational cytology is a critical, rapid-developing, yet challenging topic in the field of medical image computing which analyzes the digitized cytology image by computer-aided technologies for cancer screening. Recently, an increasing number of deep learning (DL) algorithms have made significant progress in medical image analysis, leading to the boosting publications of cytological studies. To investigate the advanced methods and comprehensive applications, we survey more than 120 publications of DL-based cytology image analysis in this article. We first introduce various deep learning methods, including fully supervised, weakly supervised, unsupervised, and transfer learning. Then, we systematically summarize the public datasets, evaluation metrics, versatile cytology image analysis applications including classification, detection, segmentation, and other related tasks. Finally, we discuss current challenges and potential research directions of computational cytology.

Rapid and high-resolution histological imaging with minimal tissue preparation has long been a challenging and yet captivating medical pursuit. Here, the authors propose a promising and transformative histological imaging method, termed computational high-throughput autofluorescence microscopy by pattern illumination (CHAMP). With the assistance of computational microscopy, CHAMP enables high-throughput and label-free imaging of thick and unprocessed tissues with large surface irregularity at an acquisition speed of 10 mm2/10 s with 1.1-µm lateral resolution. Moreover, the CHAMP image can be transformed into a virtually stained histological image (Deep-CHAMP) through unsupervised learning within 15 s, where significant cellular features are quantitatively extracted with high accuracy. The versatility of CHAMP is experimentally demonstrated using mouse brain/kidney and human lung tissues prepared with various clinical protocols, which enables a rapid and accurate intraoperative/postoperative pathological examination without tissue processing or staining, demonstrating its great potential as an assistive imaging platform for surgeons and pathologists to provide optimal adjuvant treatment.

Visual-spatial attention has been shown to influence literacy development, yet studies investigating its influence on readingin non-alphabetic scripts such as Chinese are scarce, despite recent studies demonstrating orthographic and visuo-spatialskills to be key deficits in people with dyslexia in Chinese. Here, we investigate visual-spatial processing skills in Chi-nese adolescents by measuring their 1) exogenous and endogenous attentional orienting, and 2) holistic processinga phe-nomenon typically demonstrated in face perceptionin Chinese character recognition. Compared with typically developingstudents, Chinese high-school students with dyslexia showed deficits in both endogenous and exogenous visual-spatialattention. Dyslexics also perceived characters more holistically than the controls, suggesting that they selectively attendedto individual components within Chinese characters less readily. These results demonstrated irregularities in visual-spatialprocessing skills in students in dyslexia. This study provides implications for reading intervention programs in order tofacilitate selective attention to character components to enhance literacy.

The present study examined the relationship between Holistic processing (HP), a well-established perceptual-expertisephenomenon for visual-object recognition, and Chinese literacy ability in Hong Kong preschoolers. The literacy abilityof participants was assessed by The Hong Kong Reading Ability Screening Test for Preschool Children (RAST-K); whileHP of Chinese characters was measured by adopting the complete composite paradigm from Hsiao & Cottrell (2009). Inline with the previous findings, preschoolers also showed HP in Chinese character perception, with a negative correlationbetween HP and writing ability when other measurements were controlled. This study provides a theoretical contributionon the study of Chinese writing difficulties among preschoolers. Educational implications will also be discussed.

Recent research on visual object recognition has suggested that the right hemisphere can engage either holistic or part-based processing depending on whether the recognition relies on configural (exact distances among features) or featural information, respectively. Consistent with this finding, expert Chinese reading has been marked by a left-side bias (an indication of right-hemisphere lateralization) with decreased holistic processing (as assessed using the composite paradigm) due to its reliance on featural information. Here we examine two common perceptual expertise phenomena in object recognition – holistic processing and left-side bias – of Chinese characters in adolescents with developmental dyslexia and matched controls. We found that those with dyslexia showed stronger holistic processing, a weaker left-side bias, and worse performance in Chinese character dictation than controls. This was in contrast to Limited writers (proficient readers with limited writing experience) reported in Tso, Au, and Hsiao (Psychological Science, 25, 1757–1767, 2014), who showed stronger holistic processing and worse dictation performance, but the same level of left-side bias as controls. This result demonstrated two different perceptual mechanisms underlying holistic processing: Limited writers’ holistic processing may be due to difficulties in de-emphasizing configural information unimportant to Chinese characters, whereas readers with dyslexia may have deficits selectively attending to character components to form appropriate part-based representations in the right hemisphere. © 2020, The Psychonomic Society, Inc.

Expert face recognition has been marked by holistic processingand left-side bias/right hemisphere involvement. Hencerecognition for Chinese characters, sharing many visualperceptual properties with face perception, was thought toinduce stronger holistic processing and left-side bias effect.However, Hsiao & Cottrell (2009) showed that expertise inChinese character recognition involved reduced holisticprocessing, while Tso, Au & Hsiao (2014) suggested this effectmay be modulated by writing experiences; in contrast, left-sidebias was found to be a consistent expertise marker regardlessof writing experiences. Here we examine holistic processingand left-side bias effect of Chinese character recognitionbetween adolescents with and without dyslexia. Students withdyslexia were found to recognize Chinese characters with astronger holistic processing effect than the typical controls.However, compared with the controls, dyslexics showed amore reduced left-side bias in processing mirror-symmetricChinese characters. The theoretical and educationalimplications of these results were discussed.

Expert face recognition has been marked by holistic processing and left-side bias/right hemisphere involvement. Hence recognition for Chinese characters, sharing many visual perceptual properties with face perception, was thought to induce stronger holistic processing and left-side bias effect. However, Hsiao & Cottrell (2009) showed that expertise in Chinese character recognition involved reduced holistic processing, while Tso, Au & Hsiao (2014) suggested this effect may be modulated by writing experiences; in contrast, left-side bias was found to be a consistent expertise marker regardless of writing experiences. Here we examine holistic processing and left-side bias effect of Chinese character recognition between adolescents with and without dyslexia. Students with dyslexia were found to recognize Chinese characters with a stronger holistic processing effect than the typical controls. However, compared with the controls, dyslexics showed a more reduced left-side bias in processing mirror-symmetric Chinese characters. The theoretical and educational implications of these results were discussed. © Cognitive Science Society: Creativity + Cognition + Computation, CogSci 2019.All rights reserved.

As an engineering system becomes more complex and widespread, the inherent behaviour tends to become less obvious and difficult to quantify. Over the past decade, researchers have attempted to create ‘super-models’ in order to mimic the behaviour of a complex system. They hoped that these ‘super-models’ would eventually uncover important information and knowledge of the system. Unfortunately, many of these models were insufficient in robustness and offered very little predictive power over the entire lifespan of the machine. As a result, these state-of-the-art models are rarely implemented in industry.In order to bridge this knowledge gap, partial least squares (PLS) analysis is introduced as an innovative modelling technique in the mechanical and manufacturing engineering discipline. This research aims to develop a methodology to establish casual relationships between latent (unobservable) constructs embedded in complex systems. This research argues the importance of first utilising ‘soft data’ (expert knowledge) to establish the nature of the issue. By conducting experiments and obtaining ‘hard data’ (manifest observations), conceptual models are translated into scientific models that can be analysed using partial least squares path modelling (PLSPM). This research is supported by a world-leading medical manufacturer and supplier. The company’s intravenous bag production line has a performance issue from the overall equipment effectiveness (OEE) perspective. Using the methodology established in this research, a causal network is constructed for the production line. There are two general outcomes obtained using PLSPM. First, it identifies critical variables that offer a high degree of contribution in predicting the variation of the construct. Second, it establishes a causal network that maps out the relationships between low-level constructs (such as the operating environment) and high-level constructs (such as machine quality).There are two major conclusions drawn

As an engineering system becomes more complex and widespread, the inherent behaviour tends to become less obvious and difficult to quantify. Over the past decade, researchers have attempted to create ‘super-models’ in order to mimic the behaviour of a complex system. They hoped that these ‘super-models’ would eventually uncover important information and knowledge of the system. Unfortunately, many of these models were insufficient in robustness and offered very little predictive power over the entire lifespan of the machine. As a result, these state-of-the-art models are rarely implemented in industry.In order to bridge this knowledge gap, partial least squares (PLS) analysis is introduced as an innovative modelling technique in the mechanical and manufacturing engineering discipline. This research aims to develop a methodology to establish casual relationships between latent (unobservable) constructs embedded in complex systems. This research argues the importance of first utilising ‘soft data’ (expert knowledge) to establish the nature of the issue. By conducting experiments and obtaining ‘hard data’ (manifest observations), conceptual models are translated into scientific models that can be analysed using partial least squares path modelling (PLSPM). This research is supported by a world-leading medical manufacturer and supplier. The company’s intravenous bag production line has a performance issue from the overall equipment effectiveness (OEE) perspective. Using the methodology established in this research, a causal network is constructed for the production line. There are two general outcomes obtained using PLSPM. First, it identifies critical variables that offer a high degree of contribution in predicting the variation of the construct. Second, it establishes a causal network that maps out the relationships between low-level constructs (such as the operating environment) and high-level constructs (such as machine quality).There are two major conclusions drawn

The evolutionarily conserved peripheral benzodiazepine receptor (PBR), or 18-kDa translocator protein (TSPO), is thought to be essential for cholesterol transport and steroidogenesis, and thus life. TSPO has been proposed as a biomarker of neuroinflammation and a new drug target in neurological diseases ranging from Alzheimer’s disease to anxiety. Here we show that global C57BL/6-Tspotm1GuWu(GuwiyangWurra)-knockout mice are viable with normal growth, lifespan, cholesterol transport, blood pregnenolone concentration, protoporphyrin IX metabolism, fertility and behaviour. However, while the activation of microglia after neuronal injury appears to be unimpaired, microglia from GuwiyangWurraTSPO knockouts produce significantly less ATP, suggesting reduced metabolic activity. Using the isoquinoline PK11195, the ligand originally used for the pharmacological and structural characterization of the PBR/TSPO, and the imidazopyridines CLINDE and PBR111, we demonstrate the utility of GuwiyangWurraTSPO knockouts to provide robust data on drug specificity and selectivity, both in vitro and in vivo, as well as the mechanism of action of putative TSPO-targeting drugs.

Background The mycobacteriophage large serine recombinase Bxb1 catalyzes site-specific recombination between its corresponding attP and attB recognition sites. Previously, we and others have shown that Bxb1 has catalytic activity in various eukaryotic species including Nicotiana tabacum, Schizosaccharomyces pombe, insects and mammalian cells. Results In this work, the Bxb1 recombinase gene was transformed and constitutively expressed in Arabidopsis thaliana plants harboring a chromosomally integrated attP and attB-flanked target sequence. The Bxb1 recombinase successfully excised the target sequence in a conservative manner and the resulting recombination event was heritably transmitted to subsequent generations in the absence of the recombinase transgene. In addition, we also show that Bxb1 recombinase expressing plants can be manually crossed with att-flanked target transgenic plants to generate excised progeny. Conclusion The Bxb1 large serine recombinase performs site-specific recombination in Arabidopsis thaliana germinal tissue, producing stable lines free of unwanted DNA. The precise site-specific deletion produced by Bxb1 in planta demonstrates that this enzyme can be a useful tool for the genetic engineering of plants without selectable marker transgenes or other undesirable exogenous sequences.

Background The mycobacteriophage large serine recombinase Bxb1 catalyzes site-specific recombination between its corresponding attP and attB recognition sites. Previously, we and others have shown that Bxb1 has catalytic activity in various eukaryotic species including Nicotiana tabacum, Schizosaccharomyces pombe, insects and mammalian cells. Results In this work, the Bxb1 recombinase gene was transformed and constitutively expressed in Arabidopsis thaliana plants harboring a chromosomally integrated attP and attB-flanked target sequence. The Bxb1 recombinase successfully excised the target sequence in a conservative manner and the resulting recombination event was heritably transmitted to subsequent generations in the absence of the recombinase transgene. In addition, we also show that Bxb1 recombinase expressing plants can be manually crossed with att-flanked target transgenic plants to generate excised progeny. Conclusion The Bxb1 large serine recombinase performs site-specific recombination in Arabidopsis thaliana germinal tissue, producing stable lines free of unwanted DNA. The precise site-specific deletion produced by Bxb1 in planta demonstrates that this enzyme can be a useful tool for the genetic engineering of plants without selectable marker transgenes or other undesirable exogenous sequences.

Background The large serine recombinase phiC31 from broad host range Streptomyces temperate phage, catalyzes the site-specific recombination of two recognition sites that differ in sequence, typically known as attachment sites attB and attP. Previously, we characterized the phiC31 catalytic activity and modes of action in the fission yeast Schizosaccharomyces pombe. Results In this work, the phiC31 recombinase gene was placed under the control of the Arabidopsis OXS3 promoter and introduced into Arabidopsis harboring a chromosomally integrated attB and attP-flanked target sequence. The phiC31 recombinase excised the attB and attP-flanked DNA, and the excision event was detected in subsequent generations in the absence of the phiC31 gene, indicating germinal transmission was possible. We further verified that the genomic excision was conservative and that introduction of a functional recombinase can be achieved through secondary transformation as well as manual crossing. Conclusion The phiC31 system performs site-specific recombination in germinal tissue, a prerequisite for generating stable lines with unwanted DNA removed. The precise site-specific deletion by phiC31 in planta demonstrates that the recombinase can be used to remove selectable markers or other introduced transgenes that are no longer desired and therefore can be a useful tool for genome engineering in plants.

Background The large serine recombinase phiC31 from broad host range Streptomyces temperate phage, catalyzes the site-specific recombination of two recognition sites that differ in sequence, typically known as attachment sites attB and attP. Previously, we characterized the phiC31 catalytic activity and modes of action in the fission yeast Schizosaccharomyces pombe. Results In this work, the phiC31 recombinase gene was placed under the control of the Arabidopsis OXS3 promoter and introduced into Arabidopsis harboring a chromosomally integrated attB and attP-flanked target sequence. The phiC31 recombinase excised the attB and attP-flanked DNA, and the excision event was detected in subsequent generations in the absence of the phiC31 gene, indicating germinal transmission was possible. We further verified that the genomic excision was conservative and that introduction of a functional recombinase can be achieved through secondary transformation as well as manual crossing. Conclusion The phiC31 system performs site-specific recombination in germinal tissue, a prerequisite for generating stable lines with unwanted DNA removed. The precise site-specific deletion by phiC31 in planta demonstrates that the recombinase can be used to remove selectable markers or other introduced transgenes that are no longer desired and therefore can be a useful tool for genome engineering in plants.