Novel, noninvasive skin cancer detection device shows promise

 

DALLAS – An investigational device that couples laser spectroscopy with a machine-learning algorithm demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions in real time, results from a single-center study showed.

“More than 5.4 million cases of nonmelanoma skin cancer were treated in 2012, but the accuracy of skin cancer screening prior to biopsy is pretty low, about 70%, and is individual dependent,” lead study author Sung Hyun Pyun, PhD, said at the annual conference of the American Society for Laser Medicine and Surgery. “There have been several in vivo skin cancer screening devices based on noninvasive techniques such as multispectral imaging, Raman spectroscopy, and electrical impedance spectroscopy, but their diagnostic accuracies were not sufficient for clinical use and could not be applied in real time.”

Dr. Pyun, founder and CEO of Sunnyvale, Calif.–based Speclipse, and his associates have developed a novel skin cancer diagnostic device based on laser spectroscopy and machine-learning algorithms that can be mounted on any kind of commercially available, short-pulsed aesthetic laser systems that are used in clinics. “When we irradiate the laser on skin, the patients don’t feel anything,” he said. “But since the energy is focused spatially and temporally, a trace amount of tissue is ablated, and microplasma plume is formed.” Next, the analysis module of the device examines the plasma light spectrally to extract the elemental and molecular information from the skin lesion. “Especially trace elements play key roles in cell proliferation and apoptosis, which is directly related to development of cancer cells,” he said. “We preprocess this raw spectrum to extract the most effective wavelength features. Finally, we train the deep neural network with spectral data labeled with biopsy results to construct a classification model. This classification algorithm generates the probability of the malignancy of the target skin lesion as an output based on the emission spectra as an input.”

For the single-site study, carried out in Australia, the researchers collected 502 emission spectra from skin cancers confirmed with biopsy results. They also collected 1,429 emission spectra from benign lesions. They achieved a sensitivity of 92% and a specificity of 90% out of 1,931 spectral data sets. No adverse events occurred and no microscopic damage of the irradiated skin was observed.

“Pathologic diagnosis-based cancer detection is considered to be time- and labor-consuming, and can sometimes be individual dependent,” Dr. Pyun said. “Our real-time, noninvasive, in vivo skin cancer detection device demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions.” He added that the device could be helpful in office-based cancer screening and real-time, on-site cancer detection during skin cancer surgeries.

Larger, multicenter studies of the device are being planned. Dr. Pyun holds ownership interests with Speclipse, and is an employee of the company.

[email protected]

 

DALLAS – An investigational device that couples laser spectroscopy with a machine-learning algorithm demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions in real time, results from a single-center study showed.

“More than 5.4 million cases of nonmelanoma skin cancer were treated in 2012, but the accuracy of skin cancer screening prior to biopsy is pretty low, about 70%, and is individual dependent,” lead study author Sung Hyun Pyun, PhD, said at the annual conference of the American Society for Laser Medicine and Surgery. “There have been several in vivo skin cancer screening devices based on noninvasive techniques such as multispectral imaging, Raman spectroscopy, and electrical impedance spectroscopy, but their diagnostic accuracies were not sufficient for clinical use and could not be applied in real time.”

Dr. Pyun, founder and CEO of Sunnyvale, Calif.–based Speclipse, and his associates have developed a novel skin cancer diagnostic device based on laser spectroscopy and machine-learning algorithms that can be mounted on any kind of commercially available, short-pulsed aesthetic laser systems that are used in clinics. “When we irradiate the laser on skin, the patients don’t feel anything,” he said. “But since the energy is focused spatially and temporally, a trace amount of tissue is ablated, and microplasma plume is formed.” Next, the analysis module of the device examines the plasma light spectrally to extract the elemental and molecular information from the skin lesion. “Especially trace elements play key roles in cell proliferation and apoptosis, which is directly related to development of cancer cells,” he said. “We preprocess this raw spectrum to extract the most effective wavelength features. Finally, we train the deep neural network with spectral data labeled with biopsy results to construct a classification model. This classification algorithm generates the probability of the malignancy of the target skin lesion as an output based on the emission spectra as an input.”

For the single-site study, carried out in Australia, the researchers collected 502 emission spectra from skin cancers confirmed with biopsy results. They also collected 1,429 emission spectra from benign lesions. They achieved a sensitivity of 92% and a specificity of 90% out of 1,931 spectral data sets. No adverse events occurred and no microscopic damage of the irradiated skin was observed.

“Pathologic diagnosis-based cancer detection is considered to be time- and labor-consuming, and can sometimes be individual dependent,” Dr. Pyun said. “Our real-time, noninvasive, in vivo skin cancer detection device demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions.” He added that the device could be helpful in office-based cancer screening and real-time, on-site cancer detection during skin cancer surgeries.

Larger, multicenter studies of the device are being planned. Dr. Pyun holds ownership interests with Speclipse, and is an employee of the company.

[email protected]

 

DALLAS – An investigational device that couples laser spectroscopy with a machine-learning algorithm demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions in real time, results from a single-center study showed.

“More than 5.4 million cases of nonmelanoma skin cancer were treated in 2012, but the accuracy of skin cancer screening prior to biopsy is pretty low, about 70%, and is individual dependent,” lead study author Sung Hyun Pyun, PhD, said at the annual conference of the American Society for Laser Medicine and Surgery. “There have been several in vivo skin cancer screening devices based on noninvasive techniques such as multispectral imaging, Raman spectroscopy, and electrical impedance spectroscopy, but their diagnostic accuracies were not sufficient for clinical use and could not be applied in real time.”

Dr. Pyun, founder and CEO of Sunnyvale, Calif.–based Speclipse, and his associates have developed a novel skin cancer diagnostic device based on laser spectroscopy and machine-learning algorithms that can be mounted on any kind of commercially available, short-pulsed aesthetic laser systems that are used in clinics. “When we irradiate the laser on skin, the patients don’t feel anything,” he said. “But since the energy is focused spatially and temporally, a trace amount of tissue is ablated, and microplasma plume is formed.” Next, the analysis module of the device examines the plasma light spectrally to extract the elemental and molecular information from the skin lesion. “Especially trace elements play key roles in cell proliferation and apoptosis, which is directly related to development of cancer cells,” he said. “We preprocess this raw spectrum to extract the most effective wavelength features. Finally, we train the deep neural network with spectral data labeled with biopsy results to construct a classification model. This classification algorithm generates the probability of the malignancy of the target skin lesion as an output based on the emission spectra as an input.”

For the single-site study, carried out in Australia, the researchers collected 502 emission spectra from skin cancers confirmed with biopsy results. They also collected 1,429 emission spectra from benign lesions. They achieved a sensitivity of 92% and a specificity of 90% out of 1,931 spectral data sets. No adverse events occurred and no microscopic damage of the irradiated skin was observed.

“Pathologic diagnosis-based cancer detection is considered to be time- and labor-consuming, and can sometimes be individual dependent,” Dr. Pyun said. “Our real-time, noninvasive, in vivo skin cancer detection device demonstrated a high sensitivity and specificity for discriminating skin cancers from benign lesions.” He added that the device could be helpful in office-based cancer screening and real-time, on-site cancer detection during skin cancer surgeries.

Larger, multicenter studies of the device are being planned. Dr. Pyun holds ownership interests with Speclipse, and is an employee of the company.

[email protected]