added readme_0, background reiew, and research question

0_domain_study/README.md

@@ -1 +1,40 @@
-# Domain Research
+# README
+
+Throughout my research with microscopy images of gastrointestinal
+tissues, I have recognized an important methodological limitation:
+the lack of a reproducible and scalable method to measure the
+localization of nanoparticles inside cells. In experiments on
+nanoparticle exposure, it is common for investigators to use manual
+segmentation of cells, followed by visual inspection to determine
+the localization of nanoparticles inside or outside the boundaries
+of cells. This approach may be acceptable for a limited number of
+experiments, but it is soon recognized as inadequate when the scale
+of experiments is increased.
+
+This manual approach is, by its very nature, time-consuming and
+requires considerable human effort, especially when working with
+high-resolution microscopy images. More importantly, it also injects
+an element of subjectivity into the research workflow, where the
+determination of the boundaries of cells and the localization of
+nanoparticles may vary between different investigators. This is
+especially true even within the same research group, where different
+investigators may have different interpretations, resulting in
+inconsistencies in results. This inter-observer variability is an
+important limitation, especially when the number of experiments is
+increased to hundreds or even thousands of images.
+
+In fact, nanoparticle exposure studies in modern times commonly
+result in the production of large volumes of imaging data that are
+complex in nature. However, the analytical process that follows the
+experiment has not always kept pace with the developments made in
+experimentation, creating a gap between biological experimentation
+and quantitative reproducibility.
+
+It was out of this gap that the present project was formulated. In
+essence, the problem that was identified was the requirement of a
+computational framework that would be objective, reproducible, and
+efficient enough to quantify nanoparticle accumulation in
+gastrointestinal tissue. In other words, the question that was asked
+was: how can nanoparticle localization in cells be measured
+efficiently, ensuring reproducibility, transparency, and biological
+accuracy?

0_domain_study/actionable_research_question.md

@@ -0,0 +1,35 @@
+# Actionable Research Question
+
+In the context of the stated constraints, the underlying biological  
+problem was reformulated into a computationally tractable  
+research question.
+
+---
+
+## Main Research Question
+
+Can a machine learning-based segmentation pipeline be developed  
+to reproducibly quantify the localization of particles in  
+gastrointestinal tissue cells?
+
+---
+
+## Specific Research Questions
+
+1. Can the Cellpose model be trained to reproducibly segment  
+   adipocytes from colon tissue images?
+
+2. Can particle localization be classified as inside or outside  
+   segmented cells?
+
+3. Can image-level quantitative metrics be extracted  
+   reproducibly?
+
+---
+
+## Stated Constraints
+
+- Limited availability of annotated image masks
+- Lack of a ground-truth nanoparticle image set
+- Project time constraints
+- Requirement to ensure reproducibility

0_domain_study/background_review.md

@@ -0,0 +1,162 @@
+
+# Background Review
+
+## 1.Industrial Production, Nanoscale Transition, Emerging Biological Question
+
+In the last ten years, the use and consumption of nanomaterials
+have seen an increase. This is particularly true in the fields of
+biotechnology, pharmacology, food technology, and medical
+industries. Among the engineered nanomaterials, titanium dioxide
+(TiO₂) is one of the most synthesized nanomaterials. It is
+responsible for about 70% of the total pigment production worldwide
+and is produced at the rate of 10 million tons annually. A
+significant fraction of the total production is at the nanoscale.
+
+Nanoparticles with diameters smaller than 100 nanometers,
+particularly those in the 5-50 nanometer range, have fundamentally
+different physicochemical properties. This is particularly true
+because the surface area to volume ratio is dramatically increased.
+This increase is responsible for the enhanced reactivity of the
+nanomaterial. The transition from bulk material to nanoscale is
+responsible for the new biological questions. The exposure of the
+nanomaterials for an extended time, particularly through ingestion,
+can lead to the development of new interactions between the
+epithelial tissue and the nanoparticles. The transition from bulk
+material to nanoscale is responsible for the new biological
+questions. The exposure of the nanomaterials for an extended time,
+particularly through ingestion, can lead to the development of new
+interactions between the epithelial tissue and the nanoparticles.
+The transition from bulk material to nanoscale is responsible for
+the new biological questions. The exposure of the nanomaterials for
+an extended time, particularly through ingestion, can lead to the
+development of new interactions between the epithelial tissue and
+the nanoparticles.
+
+The scientific challenge is no longer the characterization of the
+material but the spatially localized biological location. Where are
+the nanoparticles? Are they outside the cells, or are they inside
+the cells? Can the location of the nanoparticles be quantitatively
+determined?
+
+---
+
+## 2. Biological Effects and Tissue Level Complexity
+
+Titanium dioxide and other metal oxide nanocomponents have
+traditionally been considered biologically inert in bulk form.
+However, the biological effects of nanoscale titanium dioxide
+particles have been found to differ. Experimental studies have
+shown that these nanoparticles are capable of crossing epithelial
+membranes, penetrating the tissue structures of the
+gastrointestinal tract, and directly interacting with intracellular
+structures. This can lead to the disruption of cellular signaling
+pathways and an increase in oxidative stress, which is the product
+of the reaction of reactive oxygen species. Oxidative stress is an
+important component in the pathogenesis of connective tissue and
+gastrointestinal diseases. It can lead to inflammation, tissue
+remodeling, genotoxicity, and apoptosis. This is particularly true
+in tissues with high rates of turnover and remodeling.
+
+The gastrointestinal tract is the primary target for ingested
+nanoparticles. This area is particularly vulnerable to the
+cumulative effects of exposure. The epithelial and sub-epithelial
+structures are particularly vulnerable in the context of systemic
+and connective tissue diseases, as has been shown by clinical and
+experimental studies.
+
+Despite the accumulation of biological data on the effects of
+nanoparticles, an important gap remains. The conclusions are based
+on the qualitative analysis of microscopic structures.
+Quantitative spatial analysis of the localization of nanoparticles
+at the cellular level is not well developed. To bridge the gap, not
+only biological expertise is necessary, but also the application of
+computational techniques capable of converting visual data into
+quantifiable results.
+
+---
+
+## 3. Deep Learning-Based Segmentation and New Models
+
+Histological imaging is an important instrument for studying
+tissue-level effects. Microscopy helps to visualize the shape of
+the cells and the distribution of nanoparticles. Manual
+segmentation is subjective and not very efficient.
+
+Rule-based segmentation models face challenges related to the
+complexity of histological images due to heterogeneity in staining
+and cell morphology.
+
+Machine learning models for segmentation are an important
+breakthrough in methodology. Machine learning models learn spatial
+representations of cell shape and texture.
+
+CellPose[5] is a generalist deep learning model for segmentation of
+cell images that has shown good cross-domain performance. It uses
+vector flow representations to segment cell images.
+
+Recently, CellPose-SAM has been developed by combining the
+generalist segmentation performance of CellPose with the
+promptable segmentation architecture of the Segment Anything Model
+(SAM) [6]. This new model has shown better generalization and
+flexibility to unseen data, allowing for superhuman performance
+across a range of biological data.
+
+The above discussion provides an opportunity to delve deeper into
+the learning objectives of deep learning models and train new
+models for segmentation while studying the behavior of the models
+from a biological perspective.
+
+---
+
+## 4. From Segmentation to Integrated Scientific Software
+
+The scientific problem, therefore, goes beyond the simple detection
+of cells. It requires the development of a comprehensive scientific
+problem-solving process that incorporates image preprocessing,
+deep learning-based segmentation, nanoparticle localization
+analysis, and the creation of structured statistical output. Rather
+than the segmentation tool standing alone, the goal here is to
+integrate the tool within a comprehensive scientific system.
+
+This paradigm aligns perfectly with the project’s learning goals,
+which include adapting Cellpose models to colon histology,
+exploring the capabilities of Cellpose-SAM, using cutting-edge
+deep learning to solve an open biomedical problem, and creating
+independent scientific software that unifies segmentation,
+simulation, and quantification within a reproducible paradigm.
+
+In this paradigm, deep learning becomes a bridge between the visual
+information of microscopy and the structured scientific process.
+
+---
+
+## References
+
+1. Chen P, Zhao J, Zhang H, et al. Tangshen Formula Attenuates  
+   Colonic Structure Remodeling in Type 2 Diabetic Rats.  
+   Evid Based Complement Alternat Med. 2017;4064156.
+
+2. Vona R, Giovannetti A, Gambardella L, et al. Oxidative stress  
+   in the pathogenesis of systemic scleroderma.  
+   J Cell Mol Med. 2018;22:3308–3314.
+
+3. Walecka I. Systemic sclerosis and the gastrointestinal tract.  
+   Gastroenterology Rev. 2017;12(3):163–168.
+
+4. Nica AE, Alexa LM, Ionescu AO, et al. Esophageal disorders in  
+   mixed connective tissue diseases.  
+   J Med Life. 2016;9(2):141–143.
+
+5. Pachitariu M, Rariden M, Stringer C.
+   Cellpose-SAM: superhuman generalization for cellular
+   segmentation.
+   bioRxiv. 2025;651001.
+   doi:10.1101/2025.04.28.651001.
+
+6. Stringer C, Wang T, Michaelos M, Pachitariu M.
+   Cellpose: a generalist algorithm for cellular
+   segmentation.
+   Nat Methods. 2021;18:100–106.
+   doi:10.1038/s41592-020-01018-x.
+   (Originally published as bioRxiv preprint 2020;
+   doi:10.1101/2020.02.02.931238.)