Diagnosis with near infrared spectroscopy during minimally invasive procedures

The goal of this dissertation is to present the potential of diffuse optical spectroscopy technique to characterize and differentiate types of tissue, including dysplastic and cancerous tissues, when measuring the tissue spectra during a surgical or an interventional procedure under medical image guidance. This dissertation begins with a chapter that describes the different mathematical modeling of light transport in scattering media such as tissue. Each of the existing models used in literature is described including the way to extract the optical properties by applying it to the tissue measurements performed with fiber-optic handheld probes. An overview of the clinical applications investigated by the research groups is given as well as the performance of the diagnosis in discriminating different types of tissue based on the derived parameters. Chapter 2 and 3 corresponds to the validation of the diffusion theory model applied to optical spectroscopy measurements performed in a wide wavelength range as compared to what has been already presented in literature. In fact, chapter 2 presents the very first validation available in literature of the applicability of diffusion theory approximation model to measurements performed from 900 to 1600 nm. Water and lipid absorption coefficients were measured for different temperatures and used to in the model to derive the concentration of these chromophores. The validation was performed by recovering the actual water and lipid content in custom made emulsions with known lipid content. The validation of the reduced scattering estimation was performed by correlating the estimated parameters related to the reduced scattering with the particles size of the emulsions after blending and investigating the particle size distribution. Chapter 3 presents the advantage of extending the commonly-used wavelength range from 400 to 1000 nm up to 1600 nm. Although water and lipid can be estimated from diffuse optical spectroscopy measurements up to 1000 nm, this chapter shows that the extension of the wavelength significantly improves the accuracy of the optical properties extraction and that the lack of spectral feature of water and lipid when measuring up to 1000 nm can not only yield inaccurate water and lipid content but as well influence the estimated blood concentrations and reduced scattering parameters. Chapters 4 through 6 present the application of diffuse optical spectroscopy for diagnosis related to liver diseases. Chapter 4 is a benchmarking of optical spectroscopy with other techniques such as magnetic resonance spectroscopy (MRS), nuclear magnetic resonance spectroscopy (NMR), high performance thin layer chromatography (HPTLC) and histopathology for hepatic lipid quantification in mice. The derived hepatic fat fractions in the mice liver did not show any significant differences between the various techniques. Furthermore, it was shown that it was possible to clearly distinguish the group of mice on chow diet from the group of mice on high fat diet. The potential of diffuseoptical spectroscopy in quantifying hepatic lipid is of great interest for diagnosis of fatty liver disease where it is considered to be positive in patients for hepatic lipid fractions as low as 5%. Chapter 5 presents optical spectra acquired ex vivo on metastasis in liver and the surrounding healthy liver tissue in 14 patients. This chapter demonstrates the importance of including bile absorption coefficients to the model in addition to oxygenated-hemoglobin (HbO2), deoxygenated hemoglobin (Hb), water and lipid as livers are rich in bile ducts. This study shows that it is possible to discriminate tumors from the surrounding healthy liver tissue based on the amount of bile, water and the reduced scattering amplitude. Chapter 6 describes the diagnosis performance of diffuse optical spectroscopy in discriminating the tumors from the healthy liver samples with two different methods: classifying the types of tissues using the derived clinical parameters from fitting the diffusion theory mathematical model to the measurements as well as applying a statistical method to classify the raw optical measurements. In this chapter, in addition to discriminating tumors from healthy tissue, the lipid content estimated with diffuse optical spectroscopy showed a strong correlation with hepatic fat estimation from the histological slides. These findings ultimately have impact on detecting tumors when performing a biopsy as well as defining the steatosis level in liver. Chapter 7 demonstrates the capability of diffuse optical spectroscopy in discriminating tumor sites from the surrounding healthy lung tissues. An ex vivo study was conducted in samples excised from 10 patients with lung cancer. This study showed that hemoglobin volume fraction and the reduced scattering amplitude showed significant difference in both type of tissue by being lower in tumors as compared to the healthy lung sites. Additionally, the performance of diagnosis to discriminate the tumors from the healthy lung samples was evaluated and yielded sensitivity and specificity up to 86% and 85%, respectively. Chapters 8 and 9 demonstrate the potential of diffuse optical spectroscopy to classify several types of breast tissues including malignant types. Chapter 8 corresponds to an ex vivo study on 54 excised breast samples that were measured at 5 different sites, namely adipose, glandular, fibroadenoma, invasive carcinoma and ductal carcinoma in situ. From the various optical parameters that are derived from the measurements and the chromophores volume fractions, statistical tests were performed to investigate which parameter shows significant difference between pairwise types of tissue. Furthermore, the performance of diagnosis in discriminating the 5 types of tissue yielded area under receiver operator curve (AUC) ranging from 86% to 100%. Additionally, the performance of classifying benign and malignant samples was made with different types of classification methods that were already applied by various research groups that conducted optical spectroscopy measurements on breast samples. The different classification schemes were compared and it was shown that the performance of the diagnosis can vary a lot depending on the type of classification that is used. Therefore this chapter emphasizes the importance of being very critical when selecting the classification scheme. Chapter 9 shows the differences in investigating the difference in optical properties and measurements between malignant and non-malignant tissue by comparing optical spectroscopy in quantifying hepatic lipid is of great interest for diagnosis of fatty liver disease where it is considered to be positive in patients for hepatic lipid fractions as low as 5%. Chapter 5 presents optical spectra acquired ex vivo on metastasis in liver and the surrounding healthy liver tissue in 14 patients. This chapter demonstrates the importance of including bile absorption coefficients to the model in addition to oxygenated-hemoglobin (HbO2), deoxygenated hemoglobin (Hb), water and lipid as livers are rich in bile ducts. This study shows that it is possible to discriminate tumors from the surrounding healthy liver tissue based on the amount of bile, water and the reduced scattering amplitude. Chapter 6 describes the diagnosis performance of diffuse optical spectroscopy in discriminating the tumors from the healthy liver samples with two different methods: classifying the types of tissues using the derived clinical parameters from fitting the diffusion theory mathematical model to the measurements as well as applying a statistical method to classify the raw optical measurements. In this chapter, in addition to discriminating tumors from healthy tissue, the lipid content estimated with diffuse optical spectroscopy showed a strong correlation with hepatic fat estimation from the histological slides. These findings ultimately have impact on detecting tumors when performing a biopsy as well as defining the steatosis level in liver. Chapter 7 demonstrates the capability of diffuse optical spectroscopy in discriminating tumor sites from the surrounding healthy lung tissues. An ex vivo study was conducted in samples excised from 10 patients with lung cancer. This study showed that hemoglobin volume fraction and the reduced scattering amplitude showed significant difference in both type of tissue by being lower in tumors as compared to the healthy lung sites. Additionally, the performance of diagnosis to discriminate the tumors from the healthy lung samples was evaluated and yielded sensitivity and specificity up to 86% and 85%, respectively. Chapters 8 and 9 demonstrate the potential of diffuse optical spectroscopy to classify several types of breast tissues including malignant types. Chapter 8 corresponds to an ex vivo study on 54 excised breast samples that were measured at 5 different sites, namely adipose, glandular, fibroadenoma, invasive carcinoma and ductal carcinoma in situ. From the various optical parameters that are derived from the measurements and the chromophores volume fractions, statistical tests were performed to investigate which parameter shows significant difference between pairwise types of tissue. Furthermore, the performance of diagnosis in discriminating the 5 types of tissue yielded area under receiver operator curve (AUC) ranging from 86% to 100%. Additionally, the performance of classifying benign and malignant samples was made with different types of classification methods that were already applied by various research groups that conducted optical spectroscopy measurements on breast samples. The different classification schemes were compared and it was shown that the performance of the diagnosis can vary a lot depending on the type of classification that is used. Therefore this chapter emphasizes the importance of being very critical when selecting the classification scheme. Chapter 9 shows the differences in investigating the difference in optical properties and measurements between malignant and non-malignant tissue by comparing intra and inter-patients variations. It was concluded that the diagnosis performance is best when comparing the tissues within single patients as compared to when all data from all patients are compared. Chapter 10 presents the feasibility of real-time tissue characterization during needle insertions from healthy liver to hepatocellular carcinoma tumor where medical imaging such as 3D fluoroscopy and ultrasound were used as reference. Whereas in literature point measurements in healthy and in tumors are compared, this study shows that continuous diffuse reflectance measurements while advancing the needle enables the identification of the tumor boundaries based on the derived clinical parameter, namely blood oxygenation and volume fraction as well as the scattering amplitude.

• View at Worldcat

Free Full Text ( Final Version , 6mb )