Emerging Advanced Technologies Developed by IPR for Bio Medical Applications ‑.A Review
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.279707
Source of Support: None, Conflict of Interest: None
Keywords: Atmospheric pressure plasma jet, deep learning, plasma active medium, Surface-enhanced Raman spectroscopy
Key Message: Institute for plasma research (IPR) has developed various plasma-based technologies that could be helpful in the diagnosis and treatment of various diseases. The cold atmospheric plasma (CAP) generated by helium and argon gases generate reactive oxygen species (ROS) upon interaction with living cells. Thus, CAP jet has been useful in argeting tumor cells in various cell lines like lung cancer and glioblastoma and also in various animal models of tumor. CAP treatment does not affect the normal cells, possibly due to the intact metabolic processes, unlike that in tumor cells. Even CAP treated media has been shown to be effective against the cancerous cells. IPR along with AIIMS, New Delhi is also developing deep learning algorithms for the neurological applications like EEG and MRI the diagnosis of epilepsy, Alzheimer's disease and Parkinson's Disease and with TMC, Mumbai for cancer screening. IPR has also developed AI based models for the detection of chest ailments using images of chest X- rays. This article is providing an insight into the echnologies developed by IPR for biomedical applications.
The fourth state of matter in the universe is plasma with solid, liquid, and gas being the familiar former common states. About 99% of the universe is plasma. Plasma consists of a collection of charged ions, electrons, and neutral particles. Plasmas are seen in household fluorescent light bulbs and neon signboards. The gas present in them gets ionized when subjected to a high voltage and becomes a conductive plasma. The light emitted from the bulbs and signboards are due to the emitted photons when the excited electrons drop back into their previous energy levels. Plasmas are also used in fusion reactors that can transform fusion energy into electrical energy as an alternative renewable source of energy.
The above examples can be categorized into two types of plasma, referred to as cold (non-thermal) and hot (equilibrium) plasmas. The classification is based on the light of thermal equilibrium in the plasma. Cold plasmas have Te>> Tgas where T is temperature, e is electrons and g is gas. It is obtained in low-pressure discharges and has temperatures of less than 40°C. As a result, it can be used for the treatment of living tissues.,,,, Recently, such cold plasmas have also been explored for possible biological applications and this has spurred an emerging field called plasma medicine., Whereas, hot plasmas have Te~ Tgas. It is produced in plasma torches used in pyrolysis and metallurgy applications or in high-pressure discharges.
Non-thermal plasmas can be formed using dielectric barrier discharge (DBD) and atmospheric pressure plasma jet (APPJ) methods. APPJ is also referred to as cold atmospheric plasma (CAP) jet. In DBD, an electrical discharge is generated between two electrodes separated by an insulating dielectric barrier and the plasma is directly in contact with the treated objects. On the other hand, in APPJ plasma is generated inside a quartz tube. The plasma is not in contact with the biological samples and treats them remotely. This device is used globally in the field of medicine and the plasma so generated is touchable by bare hands without causing any harm. The temperature of the plasma varies from 30° to 45°. The gases used for the production of plasma in these devices are helium, argon, a mixture of oxygen, and argon. One of the most important requirements of all these methods is to have cold plasma at room temperature and the inhibition from glow to arc transition. The dielectric layer (quartz) collects the charge and causes a drop in the voltage across the plasma every time when it is ignited. This is because the discharge current does not increase to a level that induces arcing. APPJ produces non-uniform plasma in the form of streamers or filaments. These filaments are cylindrical in shape having sub millimetre radii and a lifetime of few nanoseconds. Under certain conditions, the APPJ is also seen to form uniform diffuse plasma, i.e. filament free. The use of APPJ in the field of medicine has been studied in detail by many researchers like Laroussi et al., Fridmen et al. and Lloyd et al.,, There are several applications of APPJ and these include Sterilization, disinfection, and decontamination; wound healing; dentistry; cancer treatments; dermatology; plasma treatment of implants for biocompatibility. In the mid-1990s, experiments were conducted to show that atmospheric pressure plasmas jets can be used to inactivate bacteria such as Geobacillus stearothermophilus, bacillus cereus, Escherichia More Details coli.,, Starting around the mid-2000s several investigators reported in vitro and in vivo experiments showing that APPJ can destroy the various cancerous cell of pancreas, skin, lung, bladder, colon, breast, brain, and many more.,,,,,,,,,,,,,,,,,
Role of Species present in Plasma
Investigators reported that the effects of atmospheric pressure plasma, when interacted with living cells are mainly facilitated by reactive oxygen as well as with nitrogen species (RONS) These species include hydroxyl (OH), atomic oxygen (O), singlet delta oxygen (O2 (1D)), superoxide (O2-), hydrogen peroxide (H2O2), and nitric oxide (NO). For example, the hydroxyl radical leads to the peroxidation of unsaturated fatty acids. Because of its strong oxidative properties, it affects lipids, proteins, and DNA (breakage of single and possibly double strands). Nitric oxide, which behaves as an intracellular messenger and regulator in biological functions, affects the regulation of immune deficiencies, cell proliferation, induction of phagocytosis, regulation of collagen synthesis, and angiogenesis. Moreover, it is observed that in cancer cells, the plasma is known to increase the intracellular reactive oxygen species (ROS). This leads to cell cycle arrest at the S-phase, DNA double-strand breaks, and also induction of apoptosis. It has also been reported that the RONS react with cell membranes and can even penetrate the cells. They eventually induce subsequent reactions within the cells that trigger cell-signaling cascades, which finally leads to apoptosis in cancer cells.,,,,,, Duan et al. have shown that RONS interacts not only with the cells on the surface but with those underneath as they can penetrate biological tissues up to depths of more than 1 mm.
The application of these APPJ for neuroblastoma and glioblastoma has been studied recently. Atmospheric pressure plasma was used in the treatment of neuroblastoma, which is the most common solid extracranial malignancy in children. Both in vitro and in vivo efficacy for tumor ablation was studied by Walk et al. In in-vitro studies, APPJ decreased the metabolic activity, induced apoptosis, and reduced viability of cancer cells whereas, in in-vivo, a single treatment ablated tumors and eventual tumor growth was decelerated. Malignant gliomas is an aggressive type of tumor. Even after treating them with chemo and radiotherapy, it is seen that the survival for patients is very short (8–15 months)., Most of the study of plasma interaction is done with Human Glioblastoma cells. Nagendra et al. studied the effect of APPJ on T98G cell lines and used cell growth kinetic assay, MIT assay and clonogenic survival assay to understand the interactions. They found that plasma increases the cell death and loss of the clonogenicity. Xiaoqian et al. demonstrated the effect of a plasma jet on U87 and E6/E7 cells considering the treatment time, voltage and flow rate as the parameters. They observed that all these parameters play an important role in increasing the concentration of reactive species. They established that by, the rate of ionization increased by increasing the voltage and the number of species increased by increasing the time that can interact with cells whereas the flow rate creates a volume that facilitates ionization. Among the cells, U87 cells are more active towards plasma as compared to E6/E7 cells. Vandamme et al. demonstrated a significant antitumor effect of plasma treatment in U87 glioma xenografts. Moreover, a marked decrease of tumor volume was also observed. However, further studies are still necessary to explain the mechanisms involved in this shrinkage of tumor volume. Vermeylen et al. studied the plasma effect on two types of cancer cell lines, i.e. melanoma and glioblastoma. They found that most of the dead cells stain both annexin V and propidium iodide, indicating that the plasma-treated non-viable cells are both apoptotic or necrotic in nature. Angela et al. worked on the cold atmospheric plasma (CAP) and demonstrated the effect of short and long-lived reactive species on the human glioblastoma cells. They concluded that CAP reduces 3D glioblastoma spheroid growth, cell migration, and cell proliferation. Yan et al. in his review paper also presented the latest development in plasma medicine in the area of the central nervous system.
Recently a few researchers reported that the cancer cells are not only affected directly in contact with the use of plasma jet but also indirectly by putting the cancer cells in a medium which is activated by plasma and such liquids/mediums are called plasma-activated medium (PAL/PAM). The interaction of plasma with the liquid can be classified into two categories, once of which is a direct mode (discharge within the liquid), another one is the indirect mode (discharge close to the liquid surface). This interaction phenomenon change the physico-chemical properties of these liquids. When water is used as a liquid during the generation of PAL, this water is known as plasma-activated water (PAW). Past reported work of various author's shows that PAL can be used for various applications such as surface disinfectant or sanitizer,,, cancer cell treatment, animal vaccine, dentistry, etc. Xiang et al. and Subramanian et al. demonstrated the disinfectant activity of PAW on E. coli, P. deceptionensis, and P. aeruginosa.,, Their reported work claims around 8 log (CFU/ml) reduction in E. Coli colonies and 5 log (CFU/ml) reduction in P. Deceptionensis colonies. Hongzhuan et al. tried to use PAW as a vaccine for newcastle disease virus (NDV) in chickens. Li et al. and Pan et al. showed a potential of PAW in dental studies. Li et al. used PAW as a replacement of mouthwash, as they observed a significant reduction in colonies of S. Mutans, A. Viscosus, and P. Gingivalis. Pathogen E. Faecalis is known to be a major contributor to the diseases caused in root canal treatment. Hence, Pan et al. targeted these pathogens by using PAW, and reported 7 log (CFU/ml) reduction in their conferred work. PAW can also induce apoptosis in cancer cells as shown by Li et al. They showed 18% apoptotic effect in PAW treated cervical cancer cells compared to 7% apoptotic effect in DBD plasma.,
One of major application of PAM is in cancer cell treatment. Effect of PAM on a different type of cancer cells already investigated by Florian et al., Kumar et al., Judee et al., and Takeda et al.,, PAM selectivity on HCT116 cancerous cells and GM637 human fibroblast was studied by Florian et al. Their study shows PAM only damaged HCT116 cells using γH2AX expression. Kumar et al. used micropulse plasma to generate PAM and used this PAM to trigger apoptosis in H460 lung cancer cells. Use of PAM was also studied to inhibit the growth of multicellular tumor spheroid (MCTS) by Judee et al. Effect of PAM treatment time (up to 48 hours) on MCTS (induce reactive species in MCTS) was observed on MCTS growth and DNA damage. Major finding of this showed species of H2O2 which responsible for genotoxic effect and other reactive species were responsible for growth inhibition. Takeda et al. shows PAM effectiveness against gastric cancer in the mouse. In vivo experiments were conducted in which 60% reduction in the peritoneal metastases in gastric cancer was observed in mouse model.
Tanaka et al. treated glioblastoma cells with plasma-activated medium (PAM) and concluded that PAM induced morphological changes along with apoptosis in glioblastoma cells and shrinkage of cell sizes. Moreover, PAM also down regulated the AKT, ERK survival signals in cancers, which mediate survival and proliferation by inhibiting apoptosis in glioblastoma cells. They also found that PAM down regulated expression of the membrane-bound receptor, CD-44 which activates those pathways.
Currently, researchers are trying to diagnose diseases using label-free sensing methods, which are capable to sense low biomolecule concentrations present in the biological mediums. Surface-enhanced Raman spectroscopy (SERS) have shown potential, for label-free, sensitive, and selective detection. Using SERS, minor changes in the cellular process can be traced when the disease is just initiated. There are other techniques also available that offer similar sensitivity, but they involve extensive sample processing and complex instrumentation. Particularly in the contest of neurological problems, using this method, changes in the concentrations of neurotransmitters in various zones of the brain would be helpful to diagnosis neurological diseases. SERS has proven to be an excellent technique for monitoring neurotransmitters and other biological molecules associated with neurological diseases by the use of metal nano particles (NPs). Moore et al. in his review article elaborates the recent developments in SERS for biosensing of neurological, diabetes, cardiovascular, cancer, and viral diseases. Tiwari et al. studied the feasibility of SERS for the detection of amino acid neurotransmitters such as glutamate and γ-amino butyric acid (GABA) using Ag NPs. It was observed that the limits of detection (LODs) were 10−7 M for glutamate and 10−4 M for GABA in aqueous solutions. These amino acid neurotransmitters are important for neuroendocrine control and are also linked to epilepsy. Siek et al. demonstrated the detection of choline, acetylcholine, dopamine, and epinephrine neurotransmitters with Ag NPs. An enhancement factor of 107 was obtained by using SERS method and the detection limits achieved for these neurotransmitters were 2 μM for choline, 4 μM for acetylcholine, 10 μM for dopamine, and 0.7 μM for epinephrine.
Deep learning (DL) algorithms are becoming an important ally today to improve medical/clinical practices.,,, Today, it addresses a large spectrum of problems from cancer screening and disease monitoring to personalized treatment suggestion. It is reported that DL approaches have been evaluated for Model Reference Adaptive Control (MRAC) (termed deep MRAC) in brain Positron Emission Tomography/Magnetic Resonance imaging PET/MR imaging, and neurosurgical imaging. Deep learning finds its application for advanced deformable image registration, enabling quantitative analysis across different physical imaging modalities. Deep learning is also used in neuroimaging and neuroradiology, brain segmentation stroke imaging, and imaging in oncology., It is reported that Deep Learning is used in MR image segmentation, disease detection and disease prediction based on images and text data (reports). Supervised and unsupervised machine learning methods are used for MR image segmentation and tissue classification.,,, Research on Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) based image reconstruction methods are rapidly increasing. In MRI, it is reported to be applied for MR image reconstruction. There have been reports of results for photo-realistic image synthesis using deep learning techniques, especially generative adversarial networks (GANs).,,, Various methods and approaches using deep learning have also been reported for brain segmentation.
Parkinson's and Alzheimer's diseases are neurological disorders. Neurological testing methods like Mini Mental State Examination (MMSE) and (Unified Parkinson's Disease Rating Scale) UPDRS and brain scans are routinely used to determine the diagnosis of these diseases. It is reported in the work of Sarraf and Tofigh that the scale and shift-invariant based features like shape of data, mean and standard deviation using CNN model (LeNet-5), has carried out on functional MRI 4D. The proposed system was trained on 270900 images and validated and tested on 90300 images in fMRI. The authors obtained 96.86% accuracy for detection of affected brains by Alzheimer disease. Different CNN based Deep learning models are used for features extraction and for tumor detection for cancerous and non-cancerous tumor. [Table 1] gives a list of the DL methods and its uses.
Plasma technologies developed by IPR
At Institute for Plasma Research, an Atmospheric pressure plasma jet (APPJ) working at the flow rates of 5 LPM with Helium gas was developed for in-vitro studies to study the interaction of plasma jet on oral cancer cells (gingivobuccal squamous cell carcinoma (GB-SCC) [Figure 1]. These studies were carried out in collaboration with ACTREC, Mumbai, India. Initially, the plasma jet parameters were optimized for in-vitro condition in terms of voltage, and duration of treatments. In this study, plasma jet treatment was done on oral cancer (ITOC-03), breast cancer (MCF7) and HEK293 cell line using biological assays—MTT and Clonogenicity and Raman spectroscopy. On the basis of our in vitro findings preliminary in-vivo studies were initiated on tumors generated in hamster buccal pouch (HBP) model. It was found that plasma treatment at 5 minutes led to tumor regression. Further studies to understand the mechanism of interaction is in progress. IPR is also studying the interaction of these jets with tumor tissues resected from patients with tumor associated dug-resistant epilepsy (DRE) Pathologies in collaboration with AIIMS (New Delhi).
In the preliminary studies in collaboration with ACTREC Mumbai, SERS studies on MCF-7 breast cancer cells were performed. With the help of nanopatterned surfaces produced by low energy argon ions and Physical vapor Deposition (PVD) growth of silver nanoparticles, the breast cancer was detected and the detection limit was reduced from 105 to 500 cells. Detection of Chromatin and Histones was also detected from a breast cancer cell. In [Figure 2]a, the atomic force microscope (AFM) images of nanopatterns produced by low energy ions at 500 eV ion energy are shown. [Figure 2]b shows the image of nanoparticles grown on pattern substrate for the SERS study.
[Figure 3], shows MCF-7 breast cancer cell lines, as reported in the literature were also detected using silver nanoparticles. It can be seen that without Ag nanoparticles peak intensity is lower and other vibrations peaks are not visible, while with nanoparticles additional vibrations modes are visible with higher peak intensity.
Using nanoparticles Chromatin and Histones extracted from breast cancer cell were also detected using the nanoparticles as shown in [Figure 4], these are normally undetectable with sample Raman measurement. Research is being done to detect traces cancer in a very few cancerous cells or in only single cancerous cell.
IPR has made an attempt to generate a new alternative to disinfectants in the form of plasma-activated water (PAW). When plasma interacts with water directly or indirectly, it brings physico-chemical changes in it. These physico-chemical changes make water more reactive due to the formation of stable reactive oxygen and nitrogen species (RONS). Plasma-activated water has been produced in IPR for microbial disinfection, with potential applications in the health sector.
Institute for Plasma Research (IPR) is actively involved in the development of Artificial Intelligence based software for various applications. IPR has developed AI based model for detection of footprints of chest ailments from chest X-ray images. Inception based pre-trained model has been used for retraining the model with pre-processed chest X-ray images as input to the model.
Most of the real-life chest X-ray (CXR) images obtained for Indian sub-continent like scenarios have body parts such as full length arm and foreign objects like pins, coins etc., In our dataset, we have manually identified and removed all such images. [Figure 5]a shows a poor-quality representative image from the dataset. [Figure 5]b shows one such random images from containing foreign objects the categories. The digitized chest X rays as obtained is initially pre-processed using Contrast limited adaptive histogram equalization (CLAHE) to enhance the image [Figure 5]c quality. It operates on small data regions (tiles), rather than the entire image at one time. Here, each region contrasts in enhanced, so that the histogram of final output region nearly matches the desired value.
The enhanced image contains unnecessary body parts such as hands, shoulders, part of necks, etc., In few CXR images striking differences are seen in the background of X rays. These vary on individual cases e.g. hands too close to body, poor background resolution, the body parts such as thighs and neck, etc., The unwanted pixel value needs to be carefully removed from the images as they may hinder the efficiency of neural net. We have developed an algorithm to crop the region of interest from individual CXR [Figure 5]d.
A total of 6000–30000 images have been used for training/validating the model. Training/testing accuracies are obtained in the range of 0.92 and specificity/sensitivity are obtained in the range of 0.93 and 0.97 respectively on the test dataset. Below, we present Receiver operating characteristic (ROC)/Precision Recall (PR) curve for our model [Figure 6]a and b].
In the second stage of confirmation using sputum test requires at least 100 fields to be scanned for confirming presence of TB bacilli. A trained technician takes a minimum of 5–10 minutes for scanning the fields. This process can be automated within a couple of minutes using our software (AIBacilli). AIBacilli is trained on 338 images with more than 2200 individual single/multiple bacilli in the dataset. Mean average precision and average recall for a medium-size area is in the range of 0.884 and 0.921 respectively on the test dataset. [Figure 7] shows couple of predictions on the sputum test image.
Both the AI-based software are designed with an objective of keeping rural India as prime beneficiaries, as neither of them requires any internet connectivity and can work on a standalone desktop/laptop system.
An electroencephalogram (EEG) indirectly measures brain activity by recording electrical activity along the scalp. The EEG is an excellent tool for probing neural function due to its low cost, non-invasive nature, and pervasiveness. EEG recordings are used by neurologists to diagnose common abnormalities such as epileptic events, seizures, and strokes. The International 10–20 system is an internationally recognized method which is used to describe the locations of scalp electrodes in the context of an EEG examination, as shown below in [Figure 8].
The electrical impulses in an EEG recording usually are seen as wavy lines with peaks and valleys, as shown below in [Figure 9].
Clustering, a fundamental data analysis method, is widely used for pattern recognition. Clustering methods can be based on statistical model identification or competitive learning. Here we have used competitive learning-based clustering neural networks such as the self-organizing map (SOM). SOM is a type of artificial neural network that uses unsupervised learning to build a two-dimensional map of the input space. A self-organizing map differs from other approaches during problem-solving as it uses competitive learning rather than error-correction learning.
EEG signals are analyzed with popular and efficient signal processing technique like Wavelet transform. Different sub-bands of EEG are obtained and studied for detection of seizure. The extracted feature is then applied to Unsupervised Learning for clustering of the EEG signals. IPR presently is working towards analyzing EEG data for Seizure Detection, using Dataset from TUH (Temple University Hospital) EEG Seizure Corpus. At present we found that with unbalanced data, accuracy is in the range of 73–76, precision is in the range of around 86, recall is in the range of 80s and F-measure is in the range of 76–82. With balanced data the accuracy is in the range of 77–78, precision is in the range of around 92, recall is in the range of 80s, F-measure is in the range of 75–80. The results show that the score is improved by 2 to 5 percentage if balanced data is used, i.e. equal number of seizure and non-seizure records.
Deep learning model training and testing require a large amount of dataset. Accuracy of deep learning classifier depends on the quality and size of the dataset. One of the biggest obstacles in the success of deep learning in medical imaging is the lack of availability of dataset. On the other hand, it is quite challenging to develop large medical imaging data as annotation requires extensive time from the medical experts. Moreover, it also requires multiple expert opinions to overcome any human error. Furthermore, the annotation will not always be possible due to the unavailability of qualified expert for certain cases of rare diseases. Another major issue is unbalancing of data due to the presence of a rare disease, which is very common in the health sector as they are underrepresented in the data sets.
The availability of big data, deep learning algorithms, and pro-processing power are the major factors that are driving deep learning revolution. Despite great effort and predictions about the growth of deep learning in medical imaging, however, in its current state, it is still not clear whether deep learning can ultimately become a substitute for doctors/clinicians in medical diagnosis. However, deep learning has proved to be a good support for experts in the medical field as it has already achieved remarkable results in medical image analysis particularly for image-based cancer detection, TB detection and other medical diagnosis including computer-aided diagnosis (CAD). As a result, the efficiency as well as quality of healthcare will be improved in the long-run, thus reducing the risk of late-diagnosis of serious diseases. According to the gathered data, the most widely used deep learning method is convolutional neural networks (CNNs). In CNN segmentation, feature extraction is the most represented. Therefore, a collaboration developed between hospitals, vendors/government and AI developers/scientists will certainly resolve the issue of unavailability of data to the machine learning researcher.
Hence, plasma-based technologies represent a novel treatment and offer a lot of potential for targeted and specific anti-cancer therapy. Though most of the results show a clear pathway for translation from the laboratory to the clinic, it is yet to be used in humans directly. Also, a lot of research is to be done to understand the mechanisms behind the plasma jet to establish the clear role of ROS species. Both globally and also from the studies conducted in IPR, the results indicate that further investigation is needed to establish plasma as a promising emerging therapy in medicine.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8], [Figure 9]