Indian Journal of Pathology and Microbiology

REVIEW ARTICLE
Year
: 2021  |  Volume : 64  |  Issue : 5  |  Page : 63--68

Artificial intelligence technologies empowering identification of novel diagnostic molecular markers in gastric cancer


Ishan Pandey1, Vatsala Misra1, Aprajita T Pandey2, Pramod W Ramteke3, Ranjan Agrawal4,  
1 Department of Pathology, MotiLal Nehru Medical College, Prayagraj, India
2 Centre of Biotechnology, University of Allahabad, Prayagraj, India
3 Department of Biological Sciences, SHUATS, Prayagraj, India
4 Department of Pathology, Rohilkhand Medical College and Hospital, Bareilly, Uttar Pradesh, India

Correspondence Address:
Vatsala Misra
Department of Pathology, MotiLal Nehru Medical College, Prayagraj - 211 001, Uttar Pradesh
India

Abstract

In recent clinical practice the molecular diagnostics have been significantly empowered and upgraded by the use of Artificial Intelligence and its assisted technologies. The use of Machine leaning and Deep Learning Neural network architectures have brought in a new dimension in clinical oncological research and development. These algorithm based software system with enhanced digital image analysis have emerged into a new branch of digital pathology and contributed immensely towards precision medicine and personal diagnostics. In India, gastric cancer is one of the most common cancers in males as well as in females. Various molecular biomarkers are associated with gastric cancer development and progression of which HER2 protein, a transmembrane tyrosine kinase (TK) receptor of epidermal growth factor receptors (EGFRs) family is of prime importance. The EGF receptor expression in gastric cancer is linked with its prognostics and theragnostics. These expressions are assessed by immunohistochemistry (IHC) and molecular techniques such as Fluorescence in-situ hybridization (FISH), as per recommendations for HER2 targeted immunotherapy. These have motivated the software giants like Google Inc. to produce innovative state of art technologies mimicking human traits such as learning and problem solving skill sets. This field is still under development and is slowly evolving and capturing global importance in recent times. A literature search on PubMed was performed to access updated information for this manuscript.



How to cite this article:
Pandey I, Misra V, Pandey AT, Ramteke PW, Agrawal R. Artificial intelligence technologies empowering identification of novel diagnostic molecular markers in gastric cancer.Indian J Pathol Microbiol 2021;64:63-68


How to cite this URL:
Pandey I, Misra V, Pandey AT, Ramteke PW, Agrawal R. Artificial intelligence technologies empowering identification of novel diagnostic molecular markers in gastric cancer. Indian J Pathol Microbiol [serial online] 2021 [cited 2021 Nov 27 ];64:63-68
Available from: https://www.ijpmonline.org/text.asp?2021/64/5/63/317935


Full Text



 Introduction



Artificial intelligence (AI) is the term used to describe the use of computers and digital technology to simulate the smart behaviour and analytical thinking as compared to human beings. It refers to the branch of computer science in which machine-based approaches are used for predictions mimicking what human intelligence might do in the same situation. John McCarthy was the first to describe the term AI in 1956 as the science and engineering of making intelligent machines.[1]

It is a science discipline that explores and creates computational structures that can solve complicated tasks in ways that would typically involve human intelligence. Machine Learning (ML) over the last decades has enormously revolutionized the field of AI.

AI in medicine can be categorized into two subtypes –virtual and physical. The virtual part deals with applications such as electronic health record systems to neural network-based guidance in managing treatment. The physical part deals with robots assisting in performing surgeries, prostheses for handicapped people, and care of the elderly. Computers learn the challenge of diagnosis via two main inputs - flowcharts and database. Flowcharts form the base of AI while database is utilized for deep learning (DL) or pattern recognition that involves training a computer by the use of repetitive algorithms in identifying clinical or radiological features.[2]

AI promises to change the practice of medicine for better healthcare delivery to the masses. Deep learning (DL) is a new method of AI enabling machines to analyze various training images and extract specific clinical features using a back propagation algorithm. It uses convolutional neural networks (CNNs) that logically imitate the structure and activity of brain neurons on a computer. Implementation of AI in pathology presents a special challenge following complexity and responsibility of pathological diagnosis.[2],[3]

In the research field of machine learning, deep learning is a family of computational methods that allow an algorithm to program itself by learning from a large set of examples that demonstrate the desired behaviour, eliminating the need to explicitly specify the rules.[4] Deep learning systems such as Artificial Neural Networks (ANN), deep belief networks, Recurrent Neural Networks (RNN) and CNN have been used in fields such as computer vision, voice recognition, natural language processing, audio recognition, social network scanning, automatic translation, bioinformatics, drug design, medical image analysis, histopathological diagnosis, material inspection and board game programs, where they have produced results comparable to and in some cases superior to human experts. This new technology, in particular, deep learning algorithms can be applied for improving diagnostic accuracy and efficiency of detection of cancer.[5]

Recent studies have shown the applications of AI technology in cancer imaging that can lead to a paradigm shift in oncology diagnostics and treatments.[6] Finding from previous study shows that, since the late 1990s, the sharp rise in the technical trajectories of AI systems used in cancer imaging appears to be driven by high rates of mortality of some types of cancers (e.g., lung and breast). In cancer pathology AI is involved in tumor diagnosis, sub typing, grading, staging, prognosis prediction as well as in the identification of pathological findings, biomarkers, and genetic alterations. The therapeutic approach differs in various cancer subtypes. The tumor grading depends on the mitotic count, Ki-67 index and, the proliferation score. AI when compared to the human diagnosis showed an accuracy rate varying from 60%-89%.[7]

Tumor staging

The assessment of lymph node metastasis is necessary for tumor staging but it is time-consuming and subject to various errors. In the “Cancer Metastases in Lymph Nodes Challenge” (CAMELYON16), a competition held between AI and pathologists to assess sentinel lymph nodes in breast cancer, AI algorithms outperformed the pathologists. In the same dataset, Lymph Node Assistant (LYNA), a more optimized algorithm, gained a higher performance and sensitivity along with filtering of artifacts. Circulating tumor cells predict disease progression and survival in metastatic and early-stage cancer patients using AI.[8]

Evaluation of pathological features and biomarkers

Mitosis represents the proliferation ability while tumor budding indicates aggressiveness of tumor cells but manual counting in both is time-consuming. AI rapidly demonstrates the number of budding hotspots and the status of lymph nodes. AI algorithms evaluated the biomarkers that are involved in diagnosis, prognosis and in predicting the response to therapy. A DL-based approach was evolved to automatically detect high-proliferation regions and calculate the Ki-67 index thus selecting suitable patients for the related therapies.[7] Based on the status of human epidermal growth factor receptor 2 (HER2), Trastuzumab is helpful in patients with gastric cancer (GC).

Evaluation of genetic changes

AI models were developed to identify genetic changes depending on microsatellite instability or microsatellite stability based on H and E stained images of gastric cancer without performing actual assays. This is helpful in precision therapy.

Computational image analysis have been used to develop a solution for automated analysis and annotation of H and E tissue samples, to identify the boundary of the tumour, and to accurately measure tissue cellularity and tumour cell content.[9] Malignant tumours tend to have large and irregular nuclei or multiple nuclear structures. The cytoplasm also undergoes changes where new structures appear or normal structures disappear. Malignant cells have a small amount of cytoplasm, often with vacuoles and the ratio of cytoplasm to nucleus decreases. Algorithms have been developed to quantify these features for automatic detection and the results are acquired in the form of binary classification of two cancer classes. Benign is class 0 and the malignant one is class 1. CNN is used for the extraction of features, and classification is done using the fully-connected Artificial Neural Network (ANN).[10] This in turn instigated researchers to look out for other modern and in-silico approaches including artificial intelligence based methods for molecular biomarker analysis in Gastric Cancer.[11]

Gastric cancer (GC) is ranked third, amongst disease mortality in humans worldwide and pose a serious public health emergency for prevention and research in health establishments.[12] Advances in science have produced powerful screening instruments and a number of experimental clinical regiments for cancer treatment. Among others, biomarkers have shown a strong ability to identify early-stage tumours, enhance cancer surveillance and development, assist in prognosis, and direct the effective selection to the best care choice in cancer patients.[13] Kattan et al. in their report described in detail about the knowledge on biomarkers in patients with GC and their diagnosis, prognosis and therapeutic use in clinical practice, along with a summary of potential future consequences and prospects.

Identification of novel diagnostic markers using artificial intelligence technologies in gastric cancer

Mutations or over-expression of genes that are predictive of outcomes in GC include BRAF, HER2, KRAS, MSI, NRAS, NTRK, TP53, POLE, PIK3CA, PTEN, RSP03 and E-Cadherin.[14] Amongst them, most commonly studied markers are HER2, E-cadherin, TP53 and MSI which support evidence based diagnosis and therapy. Brief details of these molecular markers are discussed as follows. HER2 identifies itself as the first molecular biomarker used in clinical practice for patients with GC.[14] Currently, only 2 molecular subtypes have been identified which are based on the over expression of HER2 protein or the amplification of its gene ERBB2.[15] E-cadherin is a major inducer of adhesion and differentiation in epithelial gastric cells, and provides significant protection against malignant development in GC and has been shown to have a close association with invasive and metastatic activity. The TP53 is an important tumour suppressor gene and mutations in p53 gene have been detected in primary human gastric cancer and gastric cancer cells. Microsatellites are repeating 1-6 nucleotide long DNA sequence units. The addition or deletion of these repeat units contributes to genomic instability and growing vulnerability to tumour growth in GC.[14] In the past half-century, the histological definition of gastric carcinoma has been primarily based on the guidelines of Lauren, of which intestinal form and diffuse form adenocarcinoma are the two main histological subtypes, plus indeterminate type as an uncommon variant.[16] The WHO classification[17] identifies four main histological types of gastric cancers: tubular, papillary, mucinous and poorly cohesive (including signet ring cell carcinoma), plus unusual histological variations.[18] Considering the most important biomarker as HER2, its status is mainly assessed by immunohistochemistry (IHC) or Fluorescence in situ hybridization (FISH) assay.[19] Various other molecular techniques with digital assisted technologies are used for detection and analysis of cancer biomarker including polymerase chain reaction (PCR), Next Generation sequencing (NGS) etc., PCR applies the process of amplification of DNA and RNA sequences while NGS examines and detects mutations, copy number variations and gene fusions in the genome by high throughput screening software analysis system.[20] The use of artificial Intelligence technology in GC started way back in 2001.[21] In specific, the study of digital gastropathology samples focuses on the characterization of tissues, quantification of biomakers, indentification of rare events like mitosis etc.[22],[23] During the last years, an increasing number of deep learning based applications for classification procedures for pathological microscopic images in gastric cancer have been developed and effectively carried out in a broad area of applications[24],[25] Software to classify pathological images, to classify tumours, identification of metastatic cancer areas, and annotation of pathological images in the field of gastroenterology.[26],[27] To add these, further many additional softwares and networks are still under development in collaboration with software giants like Google Inc., IBM etc., which can significantly enhance the diagnosis and treatment of gastric cancer.[28],[29] Short detail of the earlier studies have been summarised in [Table 1].{Table 1}

 Challenges and Perspectives



Some limitations and challenges are still there in the implementation of AI in cancer pathology.

Validation

AI algorithms must be effectively validated using multi-institutional data before they can be implemented in actual clinical practice. Use of large-scale digital-slide libraries, training or validation datasets and feedback of the expert pathologists are required for setting the benchmark, quality control and standardization.

Interpretability

Interpretability is an important factor in the clinical implementation of AI so that the confidence of the treating clinicians as well as the patients is achieved.[30] Some pathologists are scared of the change in their workflow. When they need to report on a digital platform without the aid of a microscope, not only the interpretability but also the reporting is likely to change. In this situation selecting the best algorithm is another challenge. The cost of AI to the patient is another important key factor especially in terms of its output.

The world's first system for detecting GC using AI with deep learning was reported by Hirasawa et al.[31] The training image dataset comprised of 13,584 high-quality endoscopic images of gastric site collected from 2639 histologically proven GC patients. All images were marked manually by an expert on gastric cancer and linked to the patient's clinical data. The positive predictive value was 30.6%. The reason for misdiagnosis on evaluating of endoscopic images by AI was gastritis with redness, atrophy, and intestinal metaplasia especially in areas involved by Helicobacter pylori. These findings are at times difficult even for the endoscopists to distinguish from GC.[32] If the CNN can be trained to correctly identify the normal anatomical structures or changes along with the benign lesions, the positive predictive value of detecting cancer can be improved drastically. The outcome was the limitations of AI. First, AI is trained using high resolution images and in case the images captured have low graphics, the interpretation may change. More the computers are fed with difficult images of the same entry, more likely is the outcome. Also, the lesions that are diagnosed as positive for malignancy by AI but not by the endoscopists, if biopsied may give us some clue to the accuracy. If the images before feeding into the AI are captured using different endoscopes and video systems it is likely that the output and results obtained by AI will be more error free. Carcinoma stomach often arises from atrophic gastritis. Thus, when GC resembles gastritis, it is difficult to detect in the early stages. Use of AI is helpful in detecting cancer when images of the entire stomach are clearly fed into the computers. Wu et al. developed an AI, WISENSE (renamed ENDOANGEL), to perform a real-time stomach site check.[33]

The upper endoscopic diagnosis will vary greatly depending on the AI used. If endoscopists use AI to detect gastric cancer during upper gastrointestinal examination, they may be less likely to miss a cancer. Endoscopic examinations by the non-skilled endoscopists together with the specialist level AI are expected to shorten the time required for the non-skilled endoscopist's diagnostic technique training. The clinically relevant diagnostic ability of the CNN offers a promising applicability in routine day to day clinical practice especially in resource limited setting.

Future of AI technologies in diagnostics

AI technologies are designed to be used as a diagnostic aid by trained experts in the field of gastro-oncological image analysis. The advantages of using automated image analysis software are shorter overall analytical time and improved reproducibility and repeatability of the analysis. Design of software for AI has been far-reaching as AI applications are being researched and developed in health care from myocardial risk emergency call assessment to blood test analysis to drug discovery.[34] In parallel, the FDA has increasingly cleared AI medical applications for clinical which has increased its use in gastric cancer imaging and diagnosis.[35] Diagnostic laboratories are now also recognized as a strong candidate for AI progression, particularly in the field of gastric cancer diagnosis and tissue biomarker studies. The R&D in gastro-oncology is supporting the convergence of different research fields, such as genetics, genomics, nanotechnology, nanomedicine, computer sciences, etc., that are generating new technological pathways for diagnostics and therapeutics.[11] Health issues in gastroenterology are characterized by correct diagnosis and classification of cancers and related disorders.[5] In particular, artificial intelligence (AI), generated by converging technologies in applied sciences, can support the qualitative analysis of cancer imaging by experienced physicians, such as volumetric description of tumours as well as evaluation of the impact of disease and treatment on adjacent organs.[36] Generating high-resolution digital images, along with a high amount of data recording complicated patterns and identifying the disease is crucial, providing a significant opportunity for better implementation of Artificial Intelligence (AI).[37] Use of AI technologies in gastroenterology with special focus on genomics of gastric cancer has revolutionised this field significantly aiding the experts with better theragnostic precision. [Figure 1] shows flowchart of GC diagnosis with the use of AI networking.{Figure 1}

 Conclusion



The field of AI empowered Gastric Cancer and other oncological disease diagnostics is still young and will continue to mature as researchers, clinicians, industry, regulatory organizations and patient advocacy groups work together to innovate and deliver new technologies to health care providers which are better, faster, cheaper and, more precise in use.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

References

1McCarthy J, Minsky ML, Shannon CE. A proposal for the Dartmouth summer research project on artificial intelligence-August 31, 1955. Ai Mag 2006;27:12–4.
2Madabhushi A, Lee G. Image analysis and machine learning in digital pathology: Challenges and opportunities. Med Image Anal 2016;33:170–5.
3Liu Y, Kohlberger T, Norouzi M, Dahl GE, Smith JL, Mohtashamian A, et al. Artificial intelligence-based breast cancer nodal metastasis detection: Insights into the black box for pathologists. Arch Pathol Lab Med 2019;143:859–68.
4Bi WL, Hosny A, Schabath MB, Giger ML, Birkbak NJ, Mehrtash A, et al. Artificial intelligence in cancer imaging: Clinical challenges and applications. CA Cancer J Clin 2019;69:127–57.
5Kantarjian H, Yu PP. Artificial intelligence, big data, and cancer. JAMA Oncol 2015;1:573-4.
6Coccia M. The Nexus between technological performances of countries and incidence of cancers in society. Technol Soc 2015;42:61-70.
7Hirasawa T, Aoyama K, Tanimoto T, Ishihara S, Shichijo S, Ozawa T, et al. Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic image. Gastric Cancer 2018;21:653–60.
8Saha M, Chakraborty C, Arun I, Ahmed R, Chatterjee S. An advanced deep learning approach for Ki-67 stained hotspot detection and proliferation rate scoring for prognostic evaluation of breast cancer. Sci Rep 2017;7:3213.
9Serag A, Ion-Margineanu A, Qureshi H, McMillan R, Martin MJS, Diamond J, et al. Translational AI and deep learning in diagnostic pathology. Front Med (Lausanne) 2019;6:185.
10Dabeer S, Khan MM, Islam S. Cancer diagnosis in histopathological image: CNN based approach. Inform Med Unlocked 2019;16:100231.
11Bhattacharya S, Pradhan KB, Bashar MA, Tripathi S, Semwal J, Marzo RR, et al. Artificial intelligence enabledhealthcare: A hype, hope or harm. J Family Med Prim Care 2019;8:3461-4.
12El-Dairy WS, Goldberg RM, Lenz H-J, Shields AF, Gibney GT, Tan AR, et al. The current state of molecular testing in the treatment of patients with solid tumors. CA Cancer J Clin 2019;69:305-43.
13Nerli RB, Ghagane SC. Use of artificial intelligence in the diagnosis and treatment of prostate cancer. Indian J Health Sci Biomed Res 2020;13:1-2.
14Elimova E, Wadhwa R, Shiozaki H, Sudo K, Estrella JS, Badgwell BD, et al. Molecular biomarkers in gastric cancer. J Natl Compr Canc Netw 2015;13:e19-29.
15Marrelli D, De Stefano A, de Manzoni G, Morgagni P, Di Leo A, Roviello F. Prediction of recurrence after radical surgery for gastric cancer: A scoring system obtained from a prospective multicenter study. Ann Surg 2005;241:247-55.
16Lauren P. The two histological main types of gastric carcinoma: Diffuse and so called intestinal-type carcinoma: An attempt at a histo-clinical classification. Acta Pathol Microbiol Scand 1965;64:31-49.
17Nagtegaal ID, Odze RD, Klimstra D, Paradis V, Rugge M, Schirmacher P. The 2019 WHO classification of tumours of the digestive system. Histopathology 2020;76:182-8.
18Hu B, Hajj NE, Sittler S, Lammert N, Barnes R, Meloni-Ehrig A. Gastric cancer: Classification, histology and application of molecular pathology. J Gastrointest Oncol 2012;3:251-61.
19Matsuoka T, Yashiro M. Biomarkers of gastric cancer: Current topics and future perspective. World J Gastroenterol 2018;24:2818-32.
20Mehta S, Shelling A, Muthukaruppan A, Lasham A, Blenkiron C, Laking G, et al. Predictive and prognostic molecular markers for cancer medicine. Ther Adv Med Oncol 2010;2:125-48.
21Lerner B, Dhanjal S, Hultén MA. GELFISH--graphical environment for labelling fluorescence in-situ hybridization images. J Microsc 2001;203:258-68.
22Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: An open-source platform for biological-image analysis. Nat Methods 2012;9:676-82.
23Mea VD, Baroni GL, Pilutti D, Loreto CD. SlideJ: An ImageJ plugin for automated processing of whole slide images. PLoS One 2017;12:e0180540.
24Fang M, Zhang W, Dong D, Zhou J, Tian J. Predicting histopathological findings of gastric cancer via deep generalized multi-instance learning, Proceedings Volume 10949, Medical Imaging 2019: Image Processing.
25Tuominen VJ, Tolonen TT, Isola J. ImmunoMembrane: A publicly available web application for digital image analysis of HER2 Immunohistochemistry. Histopathology 2012;60:758-67.
26Gonçalves WGE, Santos MHdP, Lobato FMF, Santos ARd, Araujo GSd. Deep learning in gastric tissue diseases: A systematic review. BMJ Open Gastroenterol 2020;7:e000371.
27Gudla PR, Nakayama K, Pegoraro G, Misteli T. SpotLearn: Convolutional neural network for detection of fluorescence in situ hybridization (FISH) signals in high-throughput imaging approaches. Cold Spring Harb Symp Quant Biol 2017;82:57–70.
28Solovei I, Cavallo A, Schermelleh L, Jaunin F, Scasselati C, Cmarko D, et al. Spatial preservation of nuclear chromatin architecture during three-dimensional fluorescence in situ hybridization (3D-FISH). Exp Cell Res 2002;276:10–23.
29Pardo E, Morgado JMT, Malpica N. Semantic segmentation of mFISH images using convolutional networks. Cytometry A 2018;93:620-7.
30Sano T, Coit DG, Kim HH, Roviello F, Kassab P, Wittekind C, et al. Proposal of a new stage grouping of gastric cancer for TNM classification: International Gastric Cancer Association staging project. Gastric Cancer 2017;20:217–25.
31Katai H, Ishikawa T, Akazawa K, Isobe Y, Miyashiro I, Oda I, et al. Five-year survival analysis of surgically resected gastric cancer cases in Japan: A retrospective analysis of more than 100,000 patients from the nationwide registry of the Japanese Gastric Cancer Association (2001–2007). Gastric Cancer 2018;21:144-54.
32Jeon HK, Kim GH, Lee BE, Park DY, Song GA, Kim DH, et al. Long-term outcome of endoscopic submucosal dissection is comparable to that of surgery for early gastric cancer: A propensitymatched analysis. Gastric Cancer 2018;21:133-43.
33Wu L, Zhang J, Zhou W, An P, Shen L, Liu J, et al. Randomized controlled trial of WISENSE, a real-time quality improving system for monitoring blind spots during esophagogastroduodenoscopy. Gut 2019;68:2161-9.
34Zakrzewski F, de Back W, Weigert M, Wenke T, Zeugner S, Mantey R, et al. Automated detection of the HER2 gene amplification status in Fluorescence in situ hybridization images for the diagnostics of cancer tissues. Sci Rep 2019;9:8231.
35Olivo-Marin JC. Extraction of spots in biological images using multiscale products. Pattern Recognit 2002;35:1989–96.
36Akkara JD, Kuriakose A. Commentary: Rise of machine learning and artificial intelligence in ophthalmology. Indian J Ophthalmol 2019;67:1009-10.
37Amin MB, Greene FL, Edge SB, ComptonCC, Gershenwald JE, Brookland JE, et al. The Eighth Edition AJCC cancer staging manual: Continuing to build a bridge from a population-based to a more “personalized” approach to cancer staging. CA Cancer J Clin 2017;67:93-9.