Machine Learning-Based Skin Tumor Delineation Combining Photoacoustic And Ultrusound Imaging
(2026) In Master’s Theses in Mathematical Sciences BERM01 20261Mathematics (Faculty of Sciences)
- Abstract (Swedish)
- Skin cancer is one of the most common forms of cancer worldwide and
its prevalence is increasing. Current assessment of tumor extent relies
mainly on histopathology after excision, and repeated surgeries are often
required to achieve complete tumor removal. This leads to high medical
costs and unnecessary patient discomfort. To reduce the pressure on the
healthcare system, there is a strong need for non-invasive methods capable
of estimating tumor boundaries before surgery.
This thesis investigates individualized delineation of basal cell carci
noma by combining ultrasound (US), photoacoustic imaging (PAI), and
machine learning (ML). The proposed framework uses US images to iden
tify conservative tumor regions and define reliable... (More) - Skin cancer is one of the most common forms of cancer worldwide and
its prevalence is increasing. Current assessment of tumor extent relies
mainly on histopathology after excision, and repeated surgeries are often
required to achieve complete tumor removal. This leads to high medical
costs and unnecessary patient discomfort. To reduce the pressure on the
healthcare system, there is a strong need for non-invasive methods capable
of estimating tumor boundaries before surgery.
This thesis investigates individualized delineation of basal cell carci
noma by combining ultrasound (US), photoacoustic imaging (PAI), and
machine learning (ML). The proposed framework uses US images to iden
tify conservative tumor regions and define reliable training data, while
multispectral PAI is used for pixel-wise tumor classification based on
spectral information. The resulting tumor probability maps are further
refined using contour-based post-processing to obtain final tumor delin
eations. The results showed good agreement between the image-based
delineations and histopathology.
The framework was also extended to three-dimensional (3D) tumor re
construction using volumetric image stacks. Although no fully registered
3D histopathological reference was available, the results indicate that com
bining structural information from US with spectral information from PAI
is a promising approach for individualized 3D skin tumor delineation. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9239219
- author
- Larsson, Otto LU
- supervisor
- organization
- course
- BERM01 20261
- year
- 2026
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Basal cell carcinoma (BCC), skin cancer, photoacoustic imag ing (PAI), ultrasound imaging (US), machine learning, tumor delineation, ac tive contour, Chan–Vese segmentation, multilayer perceptron (MLP), three dimensional (3D) reconstruction
- publication/series
- Master’s Theses in Mathematical Sciences
- report number
- LUNFBV-3019-2026
- ISSN
- 1404-6342
- other publication id
- 2026:E77
- language
- English
- id
- 9239219
- date added to LUP
- 2026-07-27 10:57:55
- date last changed
- 2026-07-27 10:57:55
@misc{9239219,
abstract = {{Skin cancer is one of the most common forms of cancer worldwide and
its prevalence is increasing. Current assessment of tumor extent relies
mainly on histopathology after excision, and repeated surgeries are often
required to achieve complete tumor removal. This leads to high medical
costs and unnecessary patient discomfort. To reduce the pressure on the
healthcare system, there is a strong need for non-invasive methods capable
of estimating tumor boundaries before surgery.
This thesis investigates individualized delineation of basal cell carci
noma by combining ultrasound (US), photoacoustic imaging (PAI), and
machine learning (ML). The proposed framework uses US images to iden
tify conservative tumor regions and define reliable training data, while
multispectral PAI is used for pixel-wise tumor classification based on
spectral information. The resulting tumor probability maps are further
refined using contour-based post-processing to obtain final tumor delin
eations. The results showed good agreement between the image-based
delineations and histopathology.
The framework was also extended to three-dimensional (3D) tumor re
construction using volumetric image stacks. Although no fully registered
3D histopathological reference was available, the results indicate that com
bining structural information from US with spectral information from PAI
is a promising approach for individualized 3D skin tumor delineation.}},
author = {{Larsson, Otto}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master’s Theses in Mathematical Sciences}},
title = {{Machine Learning-Based Skin Tumor Delineation Combining Photoacoustic And Ultrusound Imaging}},
year = {{2026}},
}